By Bill Murphy  ·  Powered by Colony Spark

THE OPERATIONS
BRIEF

ISSUE #19  ·  SEPTEMBER 18, 2026

Four data points behind the shift

· Oracle's own guidance warns that ambiguous terms like open or delayed can lead to unexpected results, so Ask Oracle appends an interpretation note when it is uncertain (Oracle Ask Oracle documentation).

· NetSuite's August 2026 release lets admins manage Ask Oracle access by role, use permitted saved searches as answer sources, and preview NetSuite Next on a copy of live account data (Oracle NetSuite 2026.2 release notes).

· NIST's July 2026 smart-manufacturing roadmap names industrial data management, integration across heterogeneous systems, and trustworthy, explainable, reliable operation as critical barriers to AI deployment on the floor (NIST).

· Deloitte's 2026 survey of more than 140 manufacturers found 84% seeing measurable value from AI, while only about 20% of use cases are actually scaled, citing robust data and governance as the gap (Deloitte).

The fix is not fewer questions. It is fewer ungoverned answers.

“I think you really have to show the proof before you get into the reasoning. You have to illustrate the why. Otherwise, it's really hard to get everyone on board.”

The takeaway

Before an AI-generated ERP answer changes a plan, a schedule, or a customer commitment, the operator needs to see what counts, where the data came from, and a way to check it against a report they already trust. Confidence is not proof. Reproducibility is.

1. Name and plain-English definition. One sentence an operator would say out loud, not a formula.

2. Owner. A person, not a department. They settle disputes about this number.

3. Calculation and grain. The math, and whether it counts orders, lines, units, or dollars.

4. Time dimension. Which date field, and why that one instead of the other three.

5. Filters and exclusions. Written down. Samples, intercompany, cancellations, rework, test orders.

6. Source of record. Which system and which table the number comes from when two disagree.

7. Refresh timing. How current the data is at the moment someone asks, stated in hours.

8. Reconciles against. The one certified report this metric has to match, every time.

1. Define. Build the operational metric in dbt: calculation, time dimension, entities, grain, filters, exclusions. Test it against known results and deploy it to the Semantic Layer.

2. Expose. Configure dbt MCP in the approved agent environment. Turn on the Semantic Layer and only the metadata tools the workflow needs, on the narrowest practical account and token permissions.

3. Ask. Pose the operating question. Require the response to show the metric name and definition, filters, grain, time range, and result, not just a number.

4. Reconcile. Check the result against the certified report or a known query. If the numbers differ, the metric owner resolves the definition, source, or filter mismatch before anyone acts on it.

Ask, why did on-time delivery fall at the East plant last month? The agent queries the governed on-time-delivery metric, states exactly how on time is defined, shows the plant and date filters and the result grain, then breaks the result down only through approved dimensions. The operator compares the total against the existing report before using the explanation for anything.

Availability

The dbt Semantic Layer is available on dbt Platform Starter, Enterprise, and Enterprise+ accounts. The self-hosted dbt MCP server can run locally, but its Semantic Layer toolset needs dbt Platform credentials and a production environment ID. Starter pricing runs $100 per user per month with 5,000 queried metrics included (dbt Semantic Layer docs, dbt pricing).

Guardrails

Keep the agent scoped to the Semantic Layer and the metadata tools the workflow actually needs. Do not enable SQL execution, administrative, job-management, or write-capable tools unless the use case and permissions explicitly call for them (dbt MCP setup).

The interface can be conversational. The number it returns still needs a constitution behind it.

By Bill Murphy  ·  Powered by Colony Spark

THE OPERATIONS
BRIEF

ISSUE #19  ·  SEPTEMBER 11, 2026

Three recognizable AI workflow outputs

· Production schedule. The planner receives a recommended sequence reorder, but the artifact must be import-ready, include change summary and reason for each reordered operation, attach the source constraints, show customer impact for affected orders, and validate against the ERP before release.

· Disposition record. Quality receives an AI recommendation to scrap, rework, or release with conditions, but the record must include reference photos, test data, audit trail, accountable approver, escalation trigger if rework time exceeds threshold, and archive proof that the decision was logged and verified.

· Supplier exception. Procurement receives a sourcing recommendation, but it must be routed to the category manager, include approved-vendor status and pricing tier, attach the exception justification and price comparison, and retain the last three quotes for audit.

Delivering an answer is step one. Delivering an artifact is the workflow.

“Agents can perceive conditions, reason over context, execute approved workflows, interact with other agents, applications and operational systems, and verify results.”

The takeaway

The test of whether an AI workflow is operating or advisory is whether it verifies what happened next. Verify means the workflow confirms the artifact was received, acted on, the action succeeded or failed, and the fallback executed if it did.

· Decision or action. The exact operational move the output must enable.

· Required artifact. The deliverable the next person can use without rebuilding it.

· Required format. The template, fields, file type, or system shape that makes the artifact usable.

· Source evidence. The records, constraints, references, and change history that must travel with the result.

· Destination. The queue, record, inbox, system, or workspace where the delivery becomes actionable.

· Accountable owner. The role that accepts, rejects, acts, and resolves exceptions.

· Acceptance test and fallback. The pass condition, rejection path, safe fallback, escalation trigger, and stop condition.

1. Choose one bounded AI workflow and name the next person or system that must act.

2. Rate each of the seven fields as Not defined, Partly defined, or Defined and tested, then add any useful notes.

3. The tool scores the handoff, flags missing critical fields, identifies the weakest field, and recommends the first repair.

4. You copy or print the completed contract and give it to the workflow owner or builder before the next pilot or automation change.

5. After a controlled test, the team updates the contract with any exception, ownership, evidence, or fallback requirement the test exposed.

Before the contract:

The AI recommends a production resequence. The planner receives the recommendation and still has to rebuild the work. They extract it into a spreadsheet, check it against open orders, flag which customers get impacted, rebuild the constraint logic to make sure nothing is missed, create a change summary for the finance team, and verify it will load into the ERP without errors. The recommendation was useful, but the workflow is not closed until the planner has rebuilt the work by hand.

After the contract:

The workflow delivers an import-ready schedule. The artifact includes the reordered operations, a change summary for each affected order with reason codes, the source constraints, customer impact flagged by order, ERP-format export ready to load, and a pre-load validation showing pass or fail. If validation fails, the fallback is to retain the last approved schedule and escalate to the planning manager. The planner receives a complete, verified artifact and makes the acceptance decision. The hidden work is now visible and happens in the workflow, not in the planner's inbox.

The handoff is finished when the next person can act without rebuilding the work.

By Bill Murphy  ·  Powered by Colony Spark

THE OPERATIONS
BRIEF

ISSUE #19  ·  SEPTEMBER 4, 2026

The adoption gap, by the numbers

· 34% of operations are already AI-augmented, but only 43% of collected data is used effectively, per Rockwell Automation's 2026 State of Smart Manufacturing survey of 1,560 manufacturing respondents across 17 countries. Read this as a scale signal, not a benchmark: the study is vendor-sponsored and spans companies from $100 million to more than $30 billion in revenue (Rockwell Automation, May 19, 2026).

· The hard deployment problems are operational, not algorithmic. NIST's 2026 roadmap for AI in smart manufacturing names effective data management, integration with heterogeneous sensing and control systems, and trustworthy, explainable operation in high-stakes environments as the critical challenges (NIST, 2026 Roadmap).

· Industrial AI only has meaning inside a defined system and user context. NIST's Industrial AI Management and Metrology program says an AI application should fulfill an explicit system need and stay bounded by the system's capabilities, and flags decision, planning, and control work as needing clearer standard operating procedures (NIST Industrial AI Management and Metrology).

· The operating context has to be documented before performance can be judged. NIST's AI Risk Management Framework calls for teams to define business value, users, system requirements, the specific task, model limits, human oversight, and deployment scope, with domain experts informing evaluation in the real operating context (NIST AI RMF Core).

If two experienced employees disagree about the source of truth, the exception owner, or the evidence of completion, the workflow is not ready for AI.

“Have we thought through our entire process, from design to build and order to cash? Have we ironed out the basics? Have we even mapped them out?”

The takeaway

A model can summarize a process it can see. It cannot settle a process the organization has never made explicit. The operating work is mapping the decisions, exceptions, and ownership before asking technology to accelerate them.

More from the roundtable

· Matt Gargas of RDT: “Broad AI rollouts designed at the executive level often miss the day-to-day work. The useful scope emerges when leaders walk the floor and talk with managers, engineers, operators, and quality staff about what should actually change.”

· Robert Wyse of Industronics Service Company: “Automate the lower-risk, repetitive work so people can focus their attention on the higher-risk decisions that cannot be allowed to slip.”

· Trigger and outcome. What starts the process, and what business result proves it finished correctly.

· Actual steps and systems. What people really do, including local spreadsheets, inboxes, sticky notes, and workarounds absent from the official procedure.

· Source of truth. Which field, record, or system is authoritative at each critical point.

· Decision and owner. What judgment is being made, who is accountable, and who has approval authority.

· Required evidence. What the next person must see before accepting the decision or handoff.

· Exceptions. What changes the normal route, who resolves it, where it escalates, and what happens if it is missed.

· Risk and consequence. Which errors are reversible, and which affect safety, quality, customer commitments, cash, or compliance.

· Potential AI role. Could a model retrieve, summarize, classify, detect, predict, recommend, or draft at this point.

· Human boundary and fallback. What requires human confirmation, and how the process continues when the model or system is unavailable.

· Baseline. Current cycle time, queue age, rework, error rate, manual touches, and decision outcome.

Eight steps, one bounded process

1. Select one bounded workflow and name its owner.

2. Observe the real work with the people who perform it. Record workarounds instead of copying the SOP.

3. Map the normal path and the exceptions that change sequence, owner, evidence, or risk.

4. At every decision point, name the authoritative input, accountable person, allowed action, and required proof.

5. Identify only the narrow steps where AI may assist.

6. Define the model's limit, the human review point, the escalation rule, and the non-AI fallback.

7. Score whether the process is ready based on agreement, source authority, exception coverage, ownership, and an available baseline.

8. Pilot the highest-readiness assistive step. Add newly discovered exceptions to the map before expanding.

· A current-state process and decision map.

· A source-of-truth table for critical inputs.

· An exception and workaround register.

· A decision-rights and escalation matrix.

· An AI-role map showing where a model may assist and where it must not decide.

· A readiness view showing what must be fixed before a pilot.

· Baseline measures for cycle time, queue age, rework, error, and decision outcomes.

· Map actual work, not only the official procedure.

· Keep high-consequence decisions under explicit human approval.

· Give every AI output an evidence view and a non-AI fallback.

· Version the map when systems, policy, ownership, or exceptions change.

· Update the process record with exceptions discovered during the pilot before expanding automation.

· Accept “do not automate yet” as a valid result when the source of truth, ownership, or process agreement is unresolved.

Path 1: Internal build

Start with a shared document, whiteboard, or process-mapping tool. Bring the process owner and two or three people who perform the work into a 60-minute session. Map one real order, exception, or decision from trigger to completion. Add the fields above, then test the map against the last three exceptions the team handled.

Path 2: Existing technology partner

Ask the ERP partner, systems integrator, or automation partner to document the same decision-ready fields before proposing an AI workflow. The deliverable should show source systems, decision owners, exception routes, evidence requirements, model boundaries, fallback, and baseline measures. The operating team should own the process map even if the partner builds the integration.

If the process is not clear enough for a new employee to follow, it is not clear enough for AI to improve.

By Bill Murphy  ·  Powered by Colony Spark

THE OPERATIONS
BRIEF

ISSUE #22  ·  AUGUST 28, 2026

Six signals from 2026

· 84% of manufacturers agreed executive leadership is more enthusiastic about AI than employees. In the same survey, 88% said they have the right in-house AI expertise, but only 22% said their AI investment is concentrated on upskilling their own people (RSM US, July 2026).

· Vendor surveys put manufacturing AI adoption near 88%. The federal survey puts actual business AI use at 19.8% as of May 2026. Both are accurate. They measure different things. One measures that a company has AI somewhere. The other measures whether the business used it (U.S. Census Bureau, May 2026).

· Usage splits hard by size. 37% at firms with 250+ employees, 32% at 100 to 249, under 20% at the smallest firms.

· The failure has a specific shape on the factory floor. Operators don't trust the recommendations. Supervisors revert to spreadsheets. Engineers quietly disconnect the system to keep lines running. The problem is not the algorithm. It is the absence of people in the process (Catalyst Connection, Oct 2025).

· When AI pilots underperform, manufacturers cite data quality (51%), integration (43%), security or compliance (34%), unclear ROI (30%), and adoption resistance (30%). Data quality and integration get budgets, owners, and dashboards. Adoption gets a training email.

· The measurement fix: Track engagement metrics, user confidence, decision consistency, and collaboration alongside technical KPIs like accuracy or uptime. The best models mean little if no one uses them (Catalyst Connection, Oct 2025).

For every tool you deployed this year, can you name the workaround it was supposed to replace, and can you prove that workaround is gone.

"I can't tell you how many VBA spreadsheets I built that looked good and calculated quickly but never got used. Everyone went back to their own spreadsheet."

The takeaway

The rollout is not the finish line. A training completion report is not evidence. If the old spreadsheet is still open on somebody's second monitor ninety days in, the deployment is a line item, not a capability.

· Target task. One specific task this tool performs. If you cannot name one, that is already the finding.

· The workaround it replaces. Name it exactly. "Manual process" is not an answer. "Denise's shipping spreadsheet on the shared drive" is an answer.

· Expected users and how often. Daily, per shift, per order, per exception.

· One operating metric. Cycle time, error rate, rework, or touches per order. Pick it before go-live.

· Usage this week against the expectation.

· Workaround status: retired, shrinking, unchanged, or grown.

· Exceptions with a reason code: too slow, did not trust it, could not access it, did not fit the process, faster my way, nobody showed me, tool was wrong.

· The operating metric reading.

· Unprompted advocacy: did anyone ask for it or recommend it without being asked.

One line per tool

Keep: usage meets expectation, workaround is shrinking or gone.

Change: usage is short and the reason codes point at a fixable cause. Fix the named cause, set a review date.

Stop: four straight weeks of unchanged workaround status with no fixable reason. Write down why, so the next vendor conversation starts from evidence.

AI quoting assistant.

Target task: first-pass quote build. Workaround: the estimator's personal pricing spreadsheet. Expected on every quote under $50K. Used on 9 of 22. Workaround unchanged. Top reason code: did not trust it on nonstandard parts. Decision: change. Pull nonstandard parts out of scope and re-baseline in three weeks.

Predictive maintenance alerts on line 3.

Workaround: monthly walk-around plus operator intuition. All 11 alerts triaged within a shift. Walk-around dropped to biweekly. Unplanned downtime down. The line 2 supervisor asked to be added. Decision: keep and expand.

AI meeting notes for production standup.

Workaround: the supervisor's notebook. Used 2 of 5 days. Workaround unchanged. Reason code: faster my way. No operating metric was ever defined at deployment. Decision: stop.

Path 1: Internal build

A shared sheet and a recurring fifteen minutes on Friday. Start with the three tools you spent the most on this year. Run it four weeks before judging anything. The discipline is the product, not the software.

Path 2: Through your partner

Ask the ERP partner or integrator to instrument usage and report override rate, abandonment rate, and exception reasons alongside uptime. Most can. Almost none do unless asked. Put it in the next statement of work.

A tool nobody uses is not a tool. It is a line item.

By Bill Murphy  ·  Powered by Colony Spark

THE OPERATIONS
BRIEF

ISSUE #22  ·  AUGUST 21, 2026

Evidence already in your system

· Organizations take 25 days on average to identify and onboard a replacement supplier, and 74% are exposed to shortages or operational disruption during that window. The cost is not just the disruption. It is all the customer promises that were made on top of dates that should not have been trusted.

· 76% of organizations still rely on manual processes for supplier due diligence and risk checks, and 58% say their reliance on spreadsheets exposes them to human error. Most never extract the supplier-reliability signals that are already sitting in acknowledgment logs and PO histories.

· Reliability now outweighs speed in keeping customers. Retail delivery leaders say: "It's when you're not reliable is when we're going to lose our customers." A 2024 McKinsey consumer survey found shoppers rank arriving within the promised window above speed, with speed falling from the top delivery priority in 2022 to fifth in 2024. Industrial buyers expect the same discipline from their own suppliers.

The promised date is not a fact. It is the output of a process that has its own track record, and that track record is knowable before the date is missed.

"The way this industry moves now, every order is a rush order. Once someone loses trust in that communication or delivery time, the customer won't come back."

The takeaway

The operators winning on customer retention are not the ones with the best forecasts. They are the ones who spot which supplier commitments are already at risk before the customer discovers it through a missed shipment.

· Supplier and part. Supplier name, part number or part family, and criticality tier (single-source, long lead-time, safety-related, or readily available with backup).

· Promised date. The current committed date as it stands today, and the original date first committed at PO placement.

· Acknowledgment lag. Days between PO issuance and supplier acknowledgment. Late or missing acknowledgment is an early confidence signal, before any date has moved.

· Date-change count. How many times this line's promised date has moved since the PO was placed, and the direction and size of each change.

· Partial-shipment history. Whether this supplier has shipped partial against comparable orders in the last 6-12 months, and how often a partial shipment preceded a final delay.

· Recent on-time rate. This supplier's on-time percentage over the trailing 90 and 365 days, weighted toward the same part family when volume allows.

· Order condition flags. Rush order, first order with this supplier, order placed during a known constraint, or unusually large quantity relative to supplier's typical run size.

· Confidence score. A single 1-100 score built from the fields above, weighted by part criticality so a low score on a single-source critical part surfaces above a low score on a readily available commodity part.

· Recommended action. Hold, monitor, contact supplier, expedite alternate sourcing, or escalate to customer communication.

Supplier / Part

Criticality

Promised (Current / Original)

Ack Lag

Date Changes

Score

Action

Supplier A / Single-source bearing

Critical

Sep 12 / Aug 28

6 days

2

34

Contact supplier, begin alternate check

Supplier B / Standard fastener

Low

Sep 3 / Sep 3

Same day

0

91

Hold, no action

Supplier C / Custom weldment

High

Sep 20 / Sep 5

4 days

3, partial shipped

41

Escalate, flag customer

Supplier D / Commodity resin

Medium

Aug 29 / Aug 25

2 days

1

68

Monitor

Daily: Automated fields (acknowledgment status, date changes, partial-shipment flags) update from ERP and supplier-portal data without manual entry.

Monday: Planner reviews the confidence heat map and movers list in a focused 15-20 minute session, not a full PO-by-PO review.

Midweek: Any commitment that drops below a defined threshold triggers an immediate supplier contact or customer-facing heads-up, not a wait for Friday.

Monthly: Compare predicted low-confidence commitments against what actually slipped. Use the miss rate to retune the scoring weights.

Quarterly: Review supplier-level trends. A supplier whose average confidence score is declining across multiple parts is a sourcing conversation, not just a planning adjustment.

· Score the commitment, not the supplier relationship. A good supplier can still have a low-confidence line on a specific constrained part.

· Keep the scoring logic visible to the planner. A black-box score that can't be explained in one sentence will get ignored the first time it's wrong.

· Weight by criticality first. A perfect score on a commodity part matters less than a mediocre score on a single-source part.

· Do not auto-communicate to customers. The score informs a human decision about when and how to raise exposure, it does not trigger the message itself.

· Feed it from data you already have. Acknowledgment timestamps, PO change logs, and shipment records already exist in most ERPs. This is an aggregation and weighting layer, not a new data-capture burden.

· Retune quarterly. Supplier behavior changes. A scoring model that never gets checked against actual outcomes becomes another number nobody trusts.

· A ranked view of which promised dates deserve trust and which deserve a buffer.

· Early visibility into suppliers whose reliability is quietly declining, before a single missed date makes it obvious.

· A clear reason to start an alternate-sourcing or customer conversation before the date slips, not after.

· Less time spent manually re-checking every open PO and more time spent on the handful that actually matter.

· A record of which planning assumptions were wrong, so the model gets sharper every cycle.

Path 1. Internal build

Start with a spreadsheet or Airtable base pulling the fields above from your ERP export: promised date, original date, acknowledgment timestamp, date-change history, and partial-shipment flags. Assign a criticality tier to your top 50-100 parts by hand. Build a simple weighted score, even a basic point-deduction model, and sort by score every Monday. Track your miss rate for one quarter before trusting the score to drive customer conversations.

Path 2. ERP or partner path

Ask your ERP provider or systems integrator whether commit-date history, acknowledgment timestamps, and partial-shipment data are already captured and exportable. Most are, they're just never aggregated into a single view. The deliverable should include a per-line confidence score, a criticality weighting scheme, a weekly ranked view, and change-tracking so a planner can see exactly why a score moved. The partner should leave the operating team owning the weekly review, not another dashboard that requires a login nobody remembers.

Build the confidence layer before the next missed date tells you it was already there to see.

By Bill Murphy  ·  Powered by Colony Spark

THE OPERATIONS
BRIEF

ISSUE #19  ·  AUGUST 14, 2026

Five signals from 2025 and 2026

· 16% of respondents were already using agents for multi-step workflows and multi-agent systems, while only 26% said leadership was clearly aligned on AI. Adoption is outpacing control (Microsoft 2026 Work Trend Index).

· 45 non-human identities exist for every human identity. These include service accounts, API keys, and secrets used by applications and automated systems. Agents use those identities to access sensitive systems and data (Cloud Security Alliance).

· Tool Misuse and Identity Abuse are now distinct risks in OWASP's Top 10 for Agentic Applications. An agent does not need an exotic exploit if it has legitimate access to export customer records, modify an order, send an email, or trigger a workflow (OWASP Top 10 for Agentic Applications).

· One compromised OAuth token associated with Salesloft Drift was used to export large volumes of Salesforce data. The same tokens provided access to connected Google Workspace accounts. One integration became the attack path across multiple systems (Google Threat Intelligence Group).

· 13% of breached organizations reported breaches of AI models or applications, and 97% of that group lacked proper AI access controls. Sixty percent of AI-related incidents led to compromised data and 31% caused operational disruption (IBM 2025 Cost of a Data Breach).

Model review asks what the AI said. Access review asks what the entire connected workflow can do if the agent, token, source document, or integration is wrong.

“Companies are already exposed to Agentic AI attacks. Often without realizing that agents are running in their environments.”

The takeaway

No single screen shows that an order-status agent can read margin data, retrieve files, send customer email, and update a case because each permission was approved in a different system. The operating risk is not visible until the permissions are mapped together.

· The business job. Agent name, owner, intended task, and named business context.

· Connected systems and actions. ERP, CRM, email, file repositories, workflow tools. For each, whether the agent can read, create, update, delete, approve, send, export, or trigger.

· Credentials and ownership. Service account, API key, OAuth grant, shared identity. Who issued it, when it was created, when it expires, who owns it now.

· Sensitive data reached. Customer records, order data, pricing, supplier information, financial commitments. Data classification and approval thresholds.

· Gaps and risks. Permissions beyond the named job, credentials with no expiration, shared identities, unnecessary system connections, write access without approval gates.

The mapping process

Step 1. Name the agent and the job. What is it called, who owns it, and what business work does it do. One agent, one job. If the agent touches multiple workflows, map them separately or consolidate the job definition first.

Step 2. List every system it touches. ERP, CRM, email, file storage, workflow automation, external APIs. For each system, record what permissions the agent holds: read, create, update, delete, approve, send, export, trigger.

Step 3. Trace the credentials. What identity does each connection use. Service account, API key, OAuth grant, shared account. Who owns it now, when was it created, when does it expire, what else can it reach in each system.

Step 4. Identify gaps. Compare actual access against the minimum needed to do the named job. Anything outside that path gets flagged: excessive read scope, write access without approval, unnecessary system, shared identity, credential with no expiration, inherited permissions from a previous workflow.

Path 1: Internal build

Start with the three agents or copilots closest to customer orders, purchasing, inventory, finance, or production scheduling. Export enterprise-application grants from the identity provider, connected-app permissions from CRM, integration users and roles from ERP, mailbox delegation from the email platform, and service credentials from the secrets manager. Join them in a controlled table using the agent or service identity as the key. Add a fixed action vocabulary: read, create, update, delete, approve, send, export, trigger. Have the business owner confirm the minimum actions needed for the named job, then flag every mismatch. A spreadsheet, diagram, and monthly credential review are enough for the first version if one owner is accountable for keeping them current.

Path 2: ERP or partner build

Ask the ERP partner, systems integrator, or managed service provider to map the agent's complete transaction path rather than reviewing only the ERP role. The partner can pull integration users, role permissions, API scopes, workflow scripts, connected applications, and audit history from the systems it already manages, then join those records with identity-provider and email permissions. The useful deliverable is not another security report. It is a live operating map tied to order entry, purchasing, inventory, production, and customer communication, with named shutoff steps and a permission-change queue the internal team can review.

Treat every agent as its own worker identity. Give it only the systems and actions required for one named job. Make the combined reach visible before the workflow goes live.

By Bill Murphy  ·  Powered by Colony Spark

THE OPERATIONS
BRIEF

ISSUE #21  ·  AUGUST 7, 2026

Five signals from 2026

· The IIA Third-Party Topical Requirement takes effect September 15, 2026, and is mandatory for any internal audit engagement touching third-party risk. It sits inside a fast rollout: Cybersecurity effective February 5, 2026, Organizational Behavior effective December 15, 2026 (The IIA, Vendor Centric).

· Manufacturers pay an average $29,100 per employee per year on federal compliance. Small manufacturers under 50 employees pay $50,100 (NAM).

· 68.4% of total manufacturing compliance cost is labor, which is the whole point: compliance is a people-capacity line (NBER).

· 11% to 40% of third parties rate high-risk, yet most programs still assemble the supporting evidence reactively across email, shared drives, and disconnected systems (Gartner via Atlas Systems).

· Organizations that stay audit-ready year round report roughly 40% lower audit prep effort. That gap is the capacity story (Atlas Systems).

Where does compliance demand appear in your weekly capacity plan today?

Doing it properly can mean adding people and accepting a different cost structure. The training and coaching are not a one-time spend either.

The takeaway

If a recurring check needs production, quality, IT, or purchasing time, it is not an administrative footnote. It is scheduled work. Put it in the same conversation as labor coverage, maintenance, changeovers, cycle counts, and customer commitments.

· Obligation. Evidence request, supplier review, access recertification, training renewal, policy approval, customer audit pack, certification check, insurance document, remediation item, or regulatory filing.

· Requirement source. Regulation, customer contract, certification standard, insurer, internal policy, lender, prime contractor, or audit request.

· Primary owner and backup. One accountable person plus the person who covers an absence.

· Contributing functions. Quality, operations, purchasing, IT, finance, HR, engineering, warehouse, or outside partner.

· Source system. The exact ERP record, QMS folder, LMS course, IAM report, HRIS record, vendor portal, ticket queue, or controlled drive location that produces the evidence.

· Evidence output. The report, approval record, certificate, access list, training completion file, supplier document, control test, or signed review the obligation must produce.

· Frequency and trigger. Weekly, monthly, quarterly, annual, contract renewal, new vendor, employee role change, system change, customer request, or incident.

· Evidence freshness rule. How current the artifact must be when reviewed.

· Planned hours. Preparation, execution, review, correction, approval, and follow-up time by function.

· Interruption cost. Unplanned hours multiplied by loaded labor cost, plus documented overtime, expedite cost, delayed release, missed production, or late-shipment exposure.

· Capacity window. The week and time block in which the work will be completed.

· Status. Scheduled, at risk, blocked, complete, or overdue.

Simple enough to maintain

Planned hours. Preparation plus execution plus review plus correction plus follow-up.

Unplanned hours. Actual hours completed outside the reserved capacity window.

Interruption cost. Unplanned hours times loaded hourly cost, plus documented overtime, expedite, delay, or lost-production cost.

Forecast accuracy. Planned hours compared with actual hours. Use the variance to improve the next cycle.

Every Monday, one six-week view

1. Capacity demand by function. Total compliance hours required from quality, operations, IT, purchasing, finance, and HR this week.

2. Collision alerts. Obligations competing with production runs, inventory counts, month-end close, maintenance, customer shipments, vacations, or training windows.

3. Evidence risk. Missing, stale, inaccessible, or unapproved artifacts due in the next two, four, and six weeks.

4. Owner actions. The three to five decisions or escalations that cannot wait.

5. Actual versus planned. Hours used last week, interruptions created, overdue work, and the estimate that needs to change next cycle.

Supplier insurance and cybersecurity review

Purchasing manager  ·  Vendor portal and ERP vendor record  ·  Rolling annual renewal

2.5 planned hours  ·  Reserved Thursday, 2:00 PM

Exposure: extra chasing time and approval delay if documents expire.

Quarterly access recertification

IT manager plus department managers  ·  IAM and HRIS  ·  Quarterly and on role change

6 planned hours  ·  Reserved Tuesday morning

Exposure: manager pull-away time and elevated access remaining open.

Food-safety training evidence refresh

QA manager plus operations supervisors  ·  LMS and QMS  ·  Monthly and before audit

4 planned hours  ·  Reserved split across Tuesday shifts

Exposure: supervisor time removed from shift coverage.

Customer audit evidence pack

Quality lead plus finance and IT contributors  ·  QMS, ERP, controlled drive  ·  Monthly refresh and on customer request

8 planned hours  ·  Reserved Wednesday, 9:00 AM

Exposure: multi-function search time if the request arrives ad hoc.

Vendor corrective-action review

Quality manager  ·  QMS and supplier record  ·  Monthly until closure

3 planned hours  ·  Reserved Friday, 10:00 AM

Exposure: repeat defect, delayed approval, and added follow-up.

· Who owns this request?

· Which system holds the current evidence?

· Is the artifact still valid?

· Which functions will lose time this week?

· Does the work collide with production or close activities?

· Should the obligation be moved, staffed differently, or escalated?

· Which outside partner needs to act before the internal owner can finish?

· Is the recurring estimate wrong because the same interruption happens every cycle?

· Friday. Owners confirm next week’s hours, blockers, and evidence status. System data updates automatically where possible.

· Monday. Leadership runs a 20-minute capacity review. Resolve collisions, approve moved work, assign backup coverage.

· Midweek. Escalate only items that become blocked, exceed planned hours, or create a new operating risk.

· Month-end. Compare planned hours with actuals. Update the estimate, retire obsolete obligations, add new recurring work.

· Quarter-end. Review total compliance capacity by function. Decide whether to rebalance work, automate evidence creation, change the control, or add partner support.

· Link to evidence. Do not duplicate controlled documents in the calendar.

· Keep human approval on interpretations, attestations, and control conclusions.

· Restrict rows and evidence links by role. Compliance visibility should not create a new access problem.

· Require a backup owner for every material obligation.

· Record actual hours. A capacity plan that never learns from actuals becomes another stale checklist.

· Separate the obligation from the task. One requirement may generate several tasks across functions.

· Track recurring interruptions as defects in the operating design, not as proof that the team needs to work faster.

· A visible compliance load before it becomes urgent.

· A realistic view of which functions are overcommitted.

· A ranked list of stale evidence and blocked owners.

· Reserved time for work that was previously invisible.

· Better estimates based on actual hours.

· A record of where interruption cost keeps repeating.

· Fewer leadership decisions made from inboxes and memory.

Path 1: Internal build

Start with Microsoft Lists, Airtable, Smartsheet, or a shared spreadsheet. Import the next 90 days of obligations. Add the required fields above. Link every row to its source system. Ask each owner to estimate hours for one cycle, then replace estimates with actuals. Publish a six-week view and review it for 20 minutes every Monday. Automate reminders only after ownership, evidence location, and escalation rules are stable.

Path 2: ERP or existing partner

Ask the existing ERP, QMS, IT, or compliance partner to add the planning layer without creating a second system of record. The deliverable should include source-system links, owner and backup fields, recurring triggers, evidence-freshness rules, planned versus actual hours, role-based access, and a six-week capacity dashboard. Require the partner to show how vendor renewals, access changes, training completion, corrective actions, and audit requests enter the calendar automatically. The partner should leave the operating team with a weekly review rhythm, not another dashboard nobody owns.

Put compliance on the capacity calendar before the next request puts it there for you.

By Bill Murphy  ·  Powered by Colony Spark

THE OPERATIONS
BRIEF

ISSUE #20  ·  JULY 31, 2026

Three signals from 2026

· 1.6 million connected and reporting Caterpillar assets, tied to $24 billion in 2025 services revenue. The connected fleet creates an earlier view of asset need; the service network turns that view into planned parts, repairs, and lifecycle revenue (Caterpillar 2025 Annual Report Highlights).

· Predictive maintenance systems now combine IoT sensors, CMMS data, work orders, maintenance records, and environmental monitoring, and can auto-trigger work orders, inspections, and parts activity (IBM, June 2026).

· $13.89B in 2026, projected $23.79B by 2031 for the predictive maintenance market. Connected equipment and AI drive the growth. Data integration and accuracy remain the constraints (MarketsandMarkets, 2026).

The installed base was once the record of what had already been sold. It is becoming the earliest reliable view of what will be needed next.

“We’ve managed field inventories tightly of new equipment and made significant progress on used. All the while, machine hours continue to accrue, aging out the fleet and driving a base-level need for replacement.”

The takeaway

The order book is a trailing measure of demand that already committed. The installed base is a leading measure of demand about to. Both belong in the same forecast, weighted differently.

· Asset register: installed equipment, age, and configuration.

· Service activity: calls, technician notes, work orders, parts consumed.

· Warranty: claims and coverage expiration dates.

· Condition and interval data: preventive-maintenance schedules and telemetry alerts.

· Commercial signals: open quotes and inbound operator questions.

Three layers

1. Join. Every event is tied to an asset, customer, product family, and component. No orphan events.

2. Rules. Known intervals and thresholds fire the obvious signals: warranty expiring, maintenance due, part reaching end of life.

3. Classification. AI reads unstructured notes and questions for failure symptoms, upgrade interest, obsolescence concerns, and replacement language. Each signal carries a demand type, time window, confidence level, evidence trail, and named owner.

· Parts staging list by component, location, demand window, and confidence.

· Asset watchlist for repeated service, accelerating parts use, approaching maintenance, warranty exposure.

· Customer-intent queue separating likely repair, upgrade, and replacement conversations.

· Warranty hotspot report by model, component, age band, and operating environment.

· Forecast-vs-actual log showing which signals produced a work order, quote, parts order, or replacement.

Path 1: Internal starting point

Start with one equipment family and two signals the team already trusts, such as service frequency and parts consumption. Back-test six to twelve months of history. Ask whether the radar would have identified the demand early enough to change a parts buy, a technician schedule, a warranty reserve, or a customer call. Add telemetry and free-text classification only after serial numbers, component codes, and outcomes reconcile.

Path 2: ERP or field-service partner

An existing ERP, field-service, IoT, or supply-chain partner can connect the source systems and automate the weekly exception run. The operator keeps ownership of signal definitions, confidence thresholds, planning windows, and the decision attached to each alert. No alert enters the forecast without its underlying evidence, and no AI-classified safety or warranty action bypasses human review.

The goal is not to predict every failure. It is to give parts, service, finance, and account teams the same early view of what the installed base is already saying.

By Bill Murphy  ·  Powered by Colony Spark

THE OPERATIONS
BRIEF

ISSUE #19  ·  JULY 30, 2026

Three signals from 2026

· 65% of large companies now rank third-party and supply-chain vulnerabilities as their greatest cyber-resilience challenge (up from 54% in 2025). Only 33% comprehensively map their supply-chain ecosystems (WEF Global Cybersecurity Outlook 2026).

· 83% of executives expect partner networks to expand over the next one to three years. Only 18% have third-party risk management fully integrated with enterprise risk management (KPMG Global TPRM Survey, March 2026).

· 16% of organizations don’t track the creation of new AI-related identities. Fewer than 25% have formal creation or removal policies. Service accounts, API keys, tokens, certificates, and AI-agent credentials now outnumber human identities (Cloud Security Alliance, 2026).

Steel, ERP, cloud hosting, an AI agent, and the only certified technician in the region belong on one list if any can halt the same order.

“Vendor lock-in is a real risk, and it’s not going away. The key is managing it through architecture.”

The takeaway

The distinction is sharper than “avoid lock-in.” Define the core, assign authority, preserve a bypass, and keep every high-impact decision reversible.

· The dependency and the workflows or sites it affects.

· Operational impact and recovery time. Scored against outcomes: degraded work, delayed shipment, stopped line, stopped order flow, customer-service failure.

· Business owner and technical owner. Two names, not a department.

· Fallback plan and last fallback test date. Untested does not count as tested.

· Contract, credential, license, or certification expiration. The dates that quietly turn a working dependency into a broken one.

Four outputs, every cycle

1. Heatmap. High-impact dependencies with no tested fallback, ranked.

2. Change log. New integrations, providers, credentials, contractors, expiring certifications.

3. Owner queue. Overdue fallback tests and unresolved single points of failure, by name.

4. Drill calendar. A short list of the few dependencies whose failure would stop orders, production, or shipping.

Path 1: Internal build

Start with one value stream, order entry through shipment. Put operations, procurement, IT, finance, HR, quality, and service in one working session. List every external dependency at each handoff and score it against the shared fields. Do not mark a fallback complete without an owner and a test date. A spreadsheet or lightweight database is enough for the first cycle.

Path 2: ERP or partner build

Use the existing ERP partner or supply-chain consultant to map integrations, scheduled jobs, connectors, custom workflows, privileged roles, support paths, and manual bypasses. Give that partner design authority inside the defined core (Bronson’s point). The operator retains ownership of impact scores, recovery targets, and risk acceptance. Strong partners add change alerts and fallback testing after go-live.

Put every dependency that can stop order flow on one list before the next outage writes it for you.

By Bill Murphy  ·  Powered by Colony Spark

THE OPERATIONS
BRIEF

ISSUE #18  ·  JULY 9, 2026

Three signals from 2025 and 2026

· 50 plus US tariff policy changes recorded since early 2025, pushing the effective tariff rate from 2.4% in 2024 to 7.7% in 2025 (highest reading since 1947), with 2026 estimated at 5.3% if temporary tariffs expire (Tax Foundation).

· 78% of manufacturing leaders now allocate more than 20% of their improvement budget to smart manufacturing initiatives, with scheduling, execution, and quality systems at the top of the priority list (Deloitte 2025 Smart Manufacturing Survey).

· $37.9 billion is NIST’s estimated annual potential impact of digital twin adoption in US manufacturing, with a median Monte Carlo scenario of $27.2 billion, tied to specific operational uses: predictive maintenance, business optimization, performance monitoring, and inventory management (NIST).

The room still makes the decision. The decision has just been rehearsed first.

“The team focused on what we could control and mitigated the risk of the things that were out of our control.”

The takeaway

The distributors and manufacturers moving through this tariff cycle best are not the ones with the best forecasts. They are the ones who have made “what we can control” a scenario the team can rehearse before the decision gets made.

· Demand. ERP open orders, backlog, forecast, customer mix and margin history.

· Supply. On-hand inventory, purchase orders, supplier lead times, BOMs, routings, current material quotes, tariff exposure by SKU or material.

· Capacity. Work-center calendar, labor schedule, overtime rules.

· Cost and commitment. Standard costs, freight assumptions, quote-expiration rules.

Same shape, different decision

Pricing meeting. Hold price, partial pass-through, full pass-through.

Capacity meeting. Accept the large order as requested, accept with revised ship date, or decline unless materials arrive by a threshold date.

Inventory meeting. Buy now, split purchase, or wait.

· Whether the data is current.

· Which SKUs are tariff-exposed.

· Whether capacity is already committed.

· What customers are affected.

· Whether the margin hit is real.

The meeting can focus on the actual trade-off: which risk the business is willing to take.

Path 1: Internal build

Start with one recurring decision, not the whole operation. Choose a decision with visible consequence: a tariff-exposed price increase, a large-order capacity commitment, or an inventory buy-ahead call. Export the required data from the ERP into a controlled spreadsheet or BI model. Build three scenarios with fixed formulas for margin, capacity, inventory, and customer impact. Assign an owner to refresh it before the weekly planning meeting. The first version does not need to be a full digital twin. It needs to stop the team from walking into the room with untested assumptions.

Path 2: ERP or planning partner

Ask the ERP partner or planning-system consultant to add a scenario layer on top of existing ERP data. The brief should be generated from saved ERP views or reports: open orders, inventory availability, PO status, routings, standard cost, and customer margin. The partner’s job is not to sell a new planning platform first. It is to connect the current system’s data to a repeatable scenario brief, then automate the refresh once the logic is trusted.

Make the meeting prove the decision before the business pays for it.

Suggest a topic

If you are running through a decision you would want to hear four other operators think out loud about, hit reply and tell us what it is. The topics that come from readers are the ones the room actually cares about.

By Bill Murphy  ·  Powered by Colony Spark

THE OPERATIONS
BRIEF

ISSUE #16  ·  JULY 3, 2026

The numbers behind the gap

· 36% of chief procurement officers say they’re very confident in their ability to redesign their function around AI (Gartner, May 2026, survey of 101 CPOs).

· 30% of supply chain organizations can point to measurable AI value in planning use cases, after all the investment (BCG Executive Perspectives).

· 47% of mid-sized businesses now use AI in operations. Almost none have built the governance structure to go with it (AllAboutAI Supply Chain Report).

The confidence is real. The autonomy isn’t there yet. And the gap between the two is where the risk lives.

“Procurement teams are seeing productivity gains from GenAI, but without intentional redesign of roles and processes, those gains remain confined to the individual level. To improve returns on their AI investments and unlock organizational gains, CPOs must design next-generation human roles focused on guiding AI toward achieving real financial outcomes, rather than mere efficiency gains.”

The challenge

Which three decisions in your operation touch customer commitments, and who, specifically, reviews the AI recommendation before it becomes a promise? If you can’t name the person for each one, that’s the design work waiting to be done.

Automate, flag, or escalate

1. Automate completely. Routine, rules-based, low-consequence. Carrier selection within approved parameters. Small inventory rebalances. Order prioritization inside predefined service rules. No human in the loop every time, just a human who set the rules and reviews the pattern monthly.

2. Flag and route. Decisions that cross a threshold. A forecast that deviates more than 15% from the prior week. A supplier lead time that jumps more than two days on a key SKU. A quote that would compress margin below the line. AI flags it; a specific, named person gets it, inside a defined window, with enough context to act: prior-period data, the reason for the deviation, and the downstream impact on customer commitments.

3. Human-required. Strategic calls. Major trade-offs. Anything touching customer commitments or multi-supplier relationships. AI provides the evidence; the human decides. Both the decision and the evidence trail get logged: the recommendation, the data behind it, the timestamp, and the name of the person who made the call. Not for compliance. For traceability. Ford couldn’t trace where its training data went thin. Don’t learn it the same way.

Path 1, Internal build

Map your top 20 recurring decisions. Assign each one to Category 1, 2, or 3. For every Category 2 decision, name the owner and document the context they need to act within 24 hours. Start there. The mapping alone exposes where the process is broken, most teams discover they can’t agree on who owns what.

Path 2, ERP or partner path

Ask your ERP partner or supply chain consultant how your current system handles exception routing today. Most platforms have the capability and it’s underused. A one-day process review focused on AI-output handoffs is usually enough to surface the top five places where recommendations land without a clear owner, and the consultants adding this layer to their post-go-live work are the ones clients keep calling back.

Build the gate before you need it. When the AI is wrong, the only question is whether you catch it before it hits the floor.

By Bill Murphy  ·  Powered by Colony Spark

THE OPERATIONS
BRIEF

ISSUE #15  ·  JUNE 26, 2026

The numbers behind the tax

· 1.9 million US manufacturing positions risk going unfilled by 2033, a 50% fulfillment gap on the 3.8M needed, with $1T in 2030 economic output on the line (Deloitte / Manufacturing Institute).

· 415,000 manufacturing job openings sitting unfilled as of mid-2025 (BLS / Manufacturing Lead Generation).

· 92% of organizations don’t consistently capture knowledge from soon-to-retire employees, even as 58% of C-suite leaders call the risk very serious (Emerj / APQC).

· 12% of annual revenue is the average enterprise loss to delayed productivity during the new-hire ramp period (SenseiLab, May 2026).

The ramp time is the capacity leak. Every week of it is productivity that was paid for and not received.

“Some of the older, more senior workforce were able to provide double the quantity in a shift as someone who was only doing it for a short period of time. They had developed their own best practices: they knew just because of quantity and time and hours on the equipment how they could produce almost no scrap, how they could produce higher throughput. But it’s very difficult to put that and ingrain that into someone who’s starting off.”

The takeaway

This is not a training problem. It is a knowledge access problem. The best practices that actually drive performance were never standardized, and the new hire has no way to reach them until they have spent months on the line absorbing the informal version one conversation at a time.

· Pre-loads role-specific context. Key accounts they will serve and the relationship history behind each, open opportunities and recent deals, who to call about what (the informal org chart, not the official one), and critical processes: what gets escalated, who approves exceptions, where the handoffs are.

· Surfaces tribal knowledge from the sources that actually hold it. Call transcripts and meeting notes from the past 12–24 months, deal history and CRM data (not just what is in the notes field, but what actually happened), past SOPs, project history, and internal communications where policy allows.

· Kills the “who do I ask?” loop. Instead of interrupting a senior colleague to ask about a customer’s last complaint, the new hire searches the brief. Instead of discovering three months in that one key account has a pricing exception, they know it on Day 1.

Four parts of the build

1. Inputs. CRM data (Salesforce, HubSpot, or equivalent), call transcript archive (Gong, Chorus, or any tool with transcript output), project records, SOPs and internal docs (even outdated ones, since the agent flags gaps), and internal comms where policy and access allow.

2. Processing. An AI agent ingests the data, synthesizes by role and account, and builds a structured brief. Not a data dump, but a curated context layer organized around what the new hire needs to function in their first 30 days.

3. Output. A personalized onboarding document delivered before or on Day 1: key accounts and relationships, critical processes and who owns what, tribal knowledge surfaced from transcripts and notes, and an “ask before you assume” layer that flags known exceptions.

4. Maintenance. The brief updates automatically as new calls are recorded, deals close or move, and the CRM is updated. It is not a document that goes stale. It is a living layer that gets more accurate over time.

Without the agent

The new hire spends two to three weeks in survival mode, asking the same questions the last person asked. Experienced colleagues get pulled out of their own work three to five times a day. The first customer call feels cold: the new hire knows the name, not the history. The first exception sends them on a two-hour expedition to figure out who handles it and why. The tribal knowledge the operation runs on stays in the heads of the people who are already overloaded. After 90 days, they might have 60% of what they need.

With the agent

The new hire arrives with context. Before the first customer call, they have read the relationship history, the last three CRM notes, and the summary of the last call transcript. They know this customer has a pricing exception on freight, that a competitor made a run at the account last year, and that the operations contact is the real decision-maker. The questions they ask are better questions. The interruptions to senior staff are fewer and more specific. After 30 days, they are contributing at a level that used to take 90.

· Shorter ramp. 30 days to meaningful contribution instead of 90, because the hire arrives with context instead of blank-slate questions.

· Fewer interruptions to senior staff. The institutional knowledge that used to live in their heads is now accessible through the brief.

· Institutional memory that survives turnover. When the next person leaves, the knowledge doesn’t go with them. It stays in the system.

· A system that compounds. Every call recorded, every note added, every deal closed makes the brief more accurate for the next hire.

Training teaches skills. Context makes them usable on Day 1.

By Bill Murphy  ·  Powered by Colony Spark

THE OPERATIONS
BRIEF

ISSUE #14  ·  JUNE 18, 2026

What moved on June 8

· Agricultural equipment (combines, harvesters) and residential HVAC: 25% → 15%

· Mobile industrial equipment (forklifts, bulldozers, backhoes, cranes): new 15% category under Annex I-C

· Steel racks & aluminum lithographic plates: added to Section 232 coverage at 25% for the first time

· US metal content threshold: 95% → 85% — more downstream manufacturers now qualify for the preferential rate without knowing to check

“Annual or quarterly price change cycles used to be sufficient… a process that took six to eight weeks (or longer) was acceptable. Today, market conditions move faster than ever. Pricing that takes too long can cause manufacturers to fall behind the cost or price curve, leading to smaller margins or money left on the table.”

IndustryWeek, January 16, 2026

The manufacturers still finishing their quarterly pricing review will arrive at the channel after their distributor has already built their own tracking page.

“I would not say we have recovered them retroactively. We absorb some of those costs until the effective date of price increases that will, among other things, serve to recover those higher costs… We honor the commitments we have made to customers; we are not operating on the basis of price in effect at time of shipment. The next orders are affected by price increases. That is the way the business is done, and that is how Insteel Industries, Inc. is operating.”

The takeaway

The repricing lag is not a failure of execution — it is the structure of the business. The operators who manage it well know exactly how long their window is and who gets called first when it opens.

· Tariff notices. Section 232 updates, HTSUS classification changes, USMCA qualification changes, proclamations, Federal Register publications.

· Supplier price announcements. Effective dates, SKU- or category-level changes, surcharges, raw-material cost adjustments from key supply partners.

· Freight rate changes. Lane-level changes, carrier surcharges, fuel adjustments, routing changes affecting landed cost on specific product lines.

· Purchase recency. Weighted toward the last 30 days. An account that ordered last week surfaces before one that last ordered six months ago.

· Deal size. Higher-value accounts appear first within the same recency tier.

· Commercial timing. Accounts with active quotes, renewals, or in-progress orders get priority regardless of size.

Four things in the packet

1. SKU-level cost delta. The specific product lines affected, with old landed cost and new landed cost side by side.

2. The recommended new price. Based on the updated cost model and target margin. Not a suggestion to “check with pricing” — a number.

3. A conversation brief. What changed, how it affects current orders or active quotes, what options are available. If June 8 created a rate reduction on the account’s primary product line, the brief says to lead with that number first.

4. A flag for accounts where cost went down. Those conversations are easier — they surface first in the queue.

Monday morning. Cost-structure inputs processed. Account queue generated and delivered to reps by 8am.

Monday–Wednesday. Account conversations happen with current pricing data. No rep says “I will have to get back to you on the new number.”

Thursday. Contract updates, quote revisions, and price confirmations go out before the commercial week closes.

Path 1: Internal build

Start with a spreadsheet that maps your cost-structure inputs to your account book. Pull 90 days of order data by product line. Define the trigger rule and the routing logic. Run it manually for 30 days before connecting any automation layer. The first value is not the automation — it is the agreement on which accounts get called first when a cost change happens. Once that logic is stable, the automation writes itself.

Path 2: ERP or partner

If your landed-cost model already lives in your ERP, the trigger and routing logic can be built inside that environment. A NetSuite or Acumatica partner with pricing-module experience can connect the tariff input to the account output in the same system your reps use for order management. The gap the agent fills is the rep-facing conversation brief, which the ERP alone will not produce. That brief is the piece worth building inside the relationship you already have.

Build the list now, while the window is still open.

By Bill Murphy  ·  Powered by Colony Spark

THE OPERATIONS
BRIEF

ISSUE #17  ·  JUNE 11, 2026

Where the agent breaks — three failure modes

· Incorrect ERP data. The agent acts on a master record nobody has cleaned in two years.

· Email-based coordination. The handoff lives in an inbox, not a system the agent can see.

· Unclear ownership. One person holds the logic for when to break the rules — and no one wrote it down.

The winners are not buying more AI. They are redesigning the work AI is supposed to touch.

“Persistent volatility is driving interest in evaluating AI-orchestrated capabilities, but investment remains constrained by foundational readiness. Even among leading supply chain organizations that have demonstrated success with performance gains and ROI on their AI investments, few have truly embedded AI into their core operations.”

The takeaway

The investment that pays off is not a bigger AI license. It is the unsexy cleanup work — documented workflows, clean master data, named owners — that lets the AI find solid ground when it lands.

Part 1  ·  Before You Automate

Fix the foundation. Then invite the AI in.

1. Scribe

Workflow Documentation / Workflow AI

Captures how work actually gets done — screen by screen — and turns it into step-by-step SOPs and process guides. Its newest products analyze those documented workflows and feed structured context directly to AI agents. Scribe’s benchmark data shows AI agents given unstructured knowledge performed 10% worse than baseline; agents given Scribe’s structured workflow data improved output quality by 71%.

Why it’s here: The workflow debt problem is literally their founding thesis.

Free tier available; Pro & Enterprise custom pricing. 600,000+ organizations, $1.3B valuation (Nov 2025).  ·  scribe.how

2. Trainual

SOP & Process Documentation

Turns your procedures, policies, and processes into searchable, assignable, trackable training manuals — with AI-assisted content building and built-in verification that people actually know the process.

Why it’s here: Before AI can touch a workflow, someone has to get it out of someone’s head. One of the most accessible on-ramps for mid-market teams to externalize tribal knowledge into role-assigned SOPs.

Starts ~$89/month for up to 25 users. SMB to mid-market.  ·  trainual.com

3. Celonis

Process Mining / Process Intelligence

Connects directly to your ERP, WMS, and transactional systems and reconstructs how processes actually run — not how you think they run — by mining event logs. Surfaces deviations, bottlenecks, and offline workarounds across procurement, order management, and production.

Why it’s here: If AI is supposed to automate a process, you first need to know what the process actually is — including every undocumented workaround. Customers like Molex use it to expose the gap between documented procedures and what’s actually happening in SAP.

Enterprise, custom pricing. #1 process mining vendor, 47.4% market share (2024).  ·  celonis.com

4. Syniti

ERP Data Quality / Master Data Management

Cleans, aligns, and governs master data across SAP and other ERP environments before AI systems are given access to it.

Why it’s here: AI is only as good as the data it touches. Syniti is the unsexy prerequisite for everything else on this list — SAP-recommended for manufacturers cleaning data before any AI migration.

Enterprise, custom pricing.  ·  syniti.com

Part 2  ·  What the Platforms Are Promising

Know what’s being sold before you buy it.

5. Kinaxis Maestro™

Supply Chain AI / Autonomous Planning

End-to-end supply chain orchestration with agentic AI — agents can autonomously monitor, predict, and act on inventory, disruptions, and demand balancing in real time, not just surface recommendations.

Why it’s here: Kinaxis is one of the loudest voices pushing the autonomous supply chain narrative. If you want to understand where the promise is being made — and where it’s currently bumping into workflow debt reality — start here.

Enterprise, custom pricing. $391M ARR (Q2 2025). Gartner Magic Quadrant Leader.  ·  kinaxis.com

6. o9 Solutions

AI-Powered Supply Chain & Enterprise Planning

The “Digital Brain” platform — connects demand planning, supply, procurement, and finance in a single AI-driven environment using agentic and generative AI across the full planning cycle.

Why it’s here: o9 is making the same autonomous supply chain bet as Kinaxis, built for the large enterprise that wants to consolidate everything into one planning model. Worth watching to understand where this market is heading.

Enterprise, custom pricing. Targets Fortune 500 manufacturers and distributors globally.  ·  o9solutions.com

Buy the platform last. Document the workflow, clean the data, name the owner first. The agents reward the operators who did that work.

By Bill Murphy  ·  Powered by Colony Spark

THE OPERATIONS
BRIEF

ISSUE #16  ·  JUNE 5, 2026

Three 2026 signals from the cost front

· ISM Prices Index hit 84.6 in April — its highest reading since April 2022, with raw material prices rising for the 19th straight month and slower supplier deliveries reported for the fifth consecutive month across 14 manufacturing industries (ISM April 2026 PMI).

· 34% of companies are now passing more than half of tariff costs to customers, up from 13% last May, and 55% of executives plan price increases of up to 15% within six months (Manufacturing Dive / KPMG).

· 87% of job shops say customers are demanding faster quote turnaround and greater order visibility than a year ago (Paperless Parts 2026 Supplier Readiness Report).

The advantage is not perfect forecasting. It is less time between a cost signal entering the business and a credible answer leaving it.

“We continue to manage the business with the goal of maintaining price/cost neutrality over time.”

The takeaway

Neutrality is not a pricing slogan — it is a stopwatch problem. The smaller the gap between supplier movement and customer-facing action, the less margin disappears between the two.

Inputs

· Supplier price notices, open RFQs, and active quotes not yet accepted

· Customer contracts with fixed shipping or pass-through limits

· SKU-level data — cost, country of origin, margin target, available inventory, lead time, customer tier

Three lanes, every morning

1. Quote now. Open RFQs or draft quotes where the cost basis changed enough to affect margin or lead time. Output: a refreshed quote packet with new landed cost, current margin, customer history, and recommended approval path.

2. Customer communication. Active customers tied to affected SKUs, contracts, or backorders. Output: a short account-impact note for the sales or service owner — what changed, whether the customer needs notice, and whether pricing language already exists.

3. Pricing review. Items that crossed a margin or cost threshold but are not tied to an immediate quote. Output: a repricing task with owner, deadline, affected customer count, and last price-change date.

How to start

The best version is not complicated. A spreadsheet, an ERP export, a supplier inbox rule, and a daily review is enough to begin. The point is cadence and ownership. When the business knows which cost changes affect which quotes today, quote speed stops being a heroic sales behavior and becomes an operating rhythm.

Quote speed stops being a heroic sales behavior and becomes an operating rhythm.

By Bill Murphy  ·  Powered by Colony Spark

THE OPERATIONS
BRIEF

ISSUE #15  ·  MAY 29, 2026

Where the market actually stands — MaintainX 2026 report

· 58% of maintenance and operations teams are already using AI.

· 75% report measurable ROI in under six months.

· 79% still saw unplanned downtime stay the same or increase (MaintainX).

Those numbers do not contradict each other. They describe the same problem. Teams have adopted the tools. The execution maturity to back them up has not kept pace.

The market is not waiting on AI to catch up. It is finding out, sometimes expensively, that the hard constraint is not the technology.

“Unplanned downtime impacts safety, productivity and performance across our operations. The financial and operational value lever is significant. A 10-15% reduction in mean-time-to-repair quickly translates into multimillion dollar savings when scaled across production lines and sites. With this agentic factory, we can help our teams respond faster to issues, capture operational knowledge, and continuously improve how our factories run.”

The takeaway

A 10-15% MTTR improvement only shows up if you were measuring MTTR to begin with. Before you bring an agent into production scheduling, write down the number it is supposed to move — otherwise you will never be able to prove it did.

Four checks, in sequence

1. Material availability. Pulls current inventory positions for every material and component the affected order needs. If there is a shortfall, it returns the quantity, the gap, and the open PO that is supposed to cover it.

2. Machine capacity & labor. Checks machine availability and labor coverage against the proposed shift window. If the resource is down, over-allocated, or missing skills coverage, it flags the constraint and the expected resolution date.

3. Customer commitments. Reviews ship date, order promise, and any priority flags already in the system before the promise changes.

4. Approval routing. Checks whether the change requires approval at the current risk level, based on order value, lead-time impact, or customer tier.

What it eliminates

The schedule change that looked valid at 8 AM and turned into a material shortage by noon. The shop floor that got a new work order without the materials to run it. The customer call where no one knew the ship date had moved until the customer asked.

What the operator gets

A documented preflight result for every AI-suggested change. A clear record of what was checked and when. A production schedule that reflects reality before it reaches the floor.

Path 1: Internal build

If your ERP exposes an API or scheduled-job capability, the preflight logic can run as a lightweight script that queries inventory positions, work-center calendars, and open customer orders before any planning-agent output gets written back. NetSuite, Dynamics, and SAP all have the data. The constraint is usually not the system — it is the decision to require the check before the action.

Path 2: ERP partner

The firms already working inside your planning and scheduling configuration are the right people to build the constraint-validation layer. The ones adding this kind of governance to their post-go-live work are the ones their clients are still on the phone with two years later, instead of filing implementation complaints.

Build the check before you need it. Once a bad AI-suggested schedule change makes it to the floor, the cost of the lesson is already committed.

By Bill Murphy  ·  Powered by Colony Spark

THE OPERATIONS
BRIEF

ISSUE #14  ·  MAY 21, 2026

Three 2026 signals that make the gap visible

· Over 40% of agentic AI projects will be canceled by 2027, not because the technology failed but because of governance gaps, unclear ownership, and inadequate risk controls. The pattern is consistent across industries: deploy faster than you build accountability and projects quietly get shelved (Gartner).

· Finance automation ownership consistently falls into the gap between finance and IT. As workflows stretch beyond the ERP into billing, payroll, and expense systems, no one owns exceptions, workflow changes, integration upkeep, or the operating model behind the automation. Result: manual workarounds, tribal knowledge, single points of failure (Accordion, April 2026).

· Industry analysts call it “automation drift” — processes run but no one owns end-state performance. Roles have to shift toward exception management, workflow ownership, and control design. Skip that step and you end up with automations that are technically functioning and operationally dangerous (TBlocks, February 2026).

The companies that will compound value from automation aren’t the ones with the most workflows running. They’re the ones where every automation has a name next to it, a review date, and a clear answer to “what happens when this breaks.”

“Without clear accountability, no one owns exceptions, workflow changes, integration upkeep, or the operating model behind the automation. Which means automation works… until it quietly breaks.”

The takeaway

The ERP partners and operations consultants who help clients build an ownership layer into their automation work — not just the initial deployment — are the ones their clients call back when the next project starts. If you’re the operator, the question is whether your partner sees the engagement as ending at go-live, or as ending the first time you review the exceptions together.

Inputs

· Exception logs from your ERP, workflow tools, CRM, or dispatch system

· The ownership map — active automations with assigned owners

· Business rule changes logged in the past 30 days (pricing, vendors, routing, staffing)

Four outputs, every week

1. Exception volume by automation. Which workflows generated the most exceptions? Rising volume on a stable automation usually means the business changed and the workflow didn’t.

2. Unresolved exceptions. Which were logged but never addressed? The clearest signal that ownership is missing. If nobody claims them, nobody owns the automation.

3. Repeat patterns. Which exceptions keep appearing? Three identical exceptions in three weeks means the rule needs updating, not the exception handling.

4. Value check. For every automation on the list, one question: is this workflow still saving more time than the exceptions it creates? If the answer is unclear, it needs a review, not more exceptions.

Path 1: Internal build

Export exception logs weekly. Tag each one to its source automation and assigned owner. Track volume, resolution rate, and repeat patterns in a shared spreadsheet or dashboard. Run a 30-minute review with the relevant process owners every week. Adjust rules when the same exception appears three times.

Path 2: ERP partner

Ask your ERP consultant or operations partner to build the exception audit into your post-implementation support. The best partners already track exception patterns across their client base and know which automations tend to drift first. This is a natural extension of ongoing support, not a new engagement.

The automation is already running. The question is whether anyone is watching.

By Bill Murphy  ·  Powered by Colony Spark

THE OPERATIONS
BRIEF

ISSUE #13  ·  MAY 14, 2026

Three 2026 signals that make the gap visible

· 75% of supply plan failures happen at the factory execution stage, not in the planning model. 73% say their ERP can show required materials but cannot prevent execution failures. 93% report difficulty getting ERP visibility into actual execution outcomes (LeanDNA via CFOtech).

· 64% of manufacturers spend 10%+ of their manufacturing budget reacting to disruption: premium freight, emergency sourcing, last-minute schedule changes. 84% hit inventory shortages multiple times last year. 80%+ are also carrying excess inventory at the same time.

· Plants generating 50–100+ weekly reports are absorbing 25–75 supervisor hours/week on manual compilation, with 1–3% transcription error rates and reports arriving 4–8 hours old (iFactory 2026 reporting guide).

The gap is not "do we have the data." The gap is "does anyone own the exception, reconcile the signal, update the record, and turn system output into a decision before trust erodes."

"Recently, we did an inventory check, and we were within 97 to 98% of what the system says we should have. Eliminating stockouts has been a godsend. Because if you have the order but cannot ship it, you are not going to make money."

The takeaway

98% inventory accuracy is not the headline metric. The headline is that the team now trusts the system enough to make decisions from it instead of working around it. Trust is not a soft outcome. It is the difference between shipping the order and missing the order.

Note: Light Composites story originally published by their software vendor and republished by Manufacturing.net. The quotes are from the operator about their own business, but the case study isn't independent reporting. Worth treating as directional, not benchmark.

What it pulls

· Support tickets: open count by account, aging by severity, reopened tickets, escalations, SLA misses

· Order behavior: order frequency vs. 13-week and 52-week baselines, missed reorder windows, partial shipments, backorders, canceled lines

· Payment timing: DSO by account, change vs. normal behavior, past due amount, short-pay or dispute notes

· Communication pattern: days since last meaningful touch, unanswered emails, negative sentiment in notes, new stakeholders appearing in threads, mentions of competitor, rebid, or vendor review

Trigger logic

· Red: severe escalation + missed reorder window + no customer reply in 7 days

· Red: payment delay 15+ days beyond account norm + unresolved billing or delivery dispute

· Yellow: order frequency down 25%+ from 13-week baseline + support tickets 2x baseline

· Yellow: expected reorder missed 10 business days + buyer communication volume down 50%

· Watch: new stakeholder on 3+ threads while order volume or payment timing is slipping

Weekly output, Monday morning

1. Top 10 accounts needing attention: name, status, owner, trigger reason, evidence, recommended action, deadline

2. Newly deteriorating accounts: what changed, and whether it's operational, financial, relationship, or mixed

3. Recovery accounts: which signal normalized, whether records were updated

4. Write-back queue: CRM tasks created, account status updated, AR notes added, owners notified

Path 1: Internal build

Start with exports from help desk, ERP, accounting, CRM, and email metadata. Define 8–12 triggers. Run it manually for 30 days before automating any write-backs. The first win is not the AI. It is agreement on which signals count and who owns each exception.

Path 2: ERP or partner

If customer, order, AR, and support data already live in the ERP ecosystem, build the pulse there. Use native alerts where possible, then add an agent layer for cross-system reasoning and task creation.

The operating question isn't how many reports you have. It's which reports are allowed to trigger action, and who is accountable when the report is wrong.