Keith Dunnell

Safety and security can be looking at the very same machine and neither can see what the other has done to it, because the work order and the vulnerability list sit in different systems.

• Dunnell's reasoning is that this is not a technology gap that a better tool closes — it is a structural gap in how information about a piece of equipment is recorded and shared.
• The posts support the point with two outside figures: manufacturing as the most-attacked industry five years running, and a Siemens estimate of $2.3 million per hour for an idle automotive line.

A controller fails on a CNC machine.

• Maintenance swaps it out and has the line running before the shift ends, but the replacement part is two firmware versions behind and carries a flaw already used in attacks on other plants.
• Nobody catches it, and six weeks later a scan flags the exposure — at which point fixing it means stopping a production line that is running full.
• Dunnell calls it an architecture failure rather than a technology failure.

Read original

The same fragmentation shows up as a decision standoff, where exposure, maintenance schedules, and customer deadlines are each owned by a different function and nobody moves first. The delay is manufactured by the structure, not by any one team being unreasonable, so a shared view of the machine and its state is what shortens it.

Security flags an exposure on a production line.

• Maintenance says the line cannot be touched for three weeks.
• Production says the customer needs parts Friday.
• Nobody moves.
• Dunnell frames this as fragmentation rather than a strategy problem and points to a recorded conversation at IMTS 2026 about why data split across OT, IT, security, and maintenance costs weeks of delay, and about one manufacturing leader who reportedly fixed it in 90 days.
• Ask for the podcast details to assess that fix.

The framing is consistent across the posts, but the emphasis shifts: earlier posts state the fragmentation thesis in a sentence, while later ones walk through the operational scene before naming the cause.

Read original · Original 2

Shared from the IMTS 2026 show in Chicago, the point is that AI agents have always been accessing data, but now they are also making decisions.

• Manufacturers need to monitor the agents and keep security in place while AI adds value — oversight running alongside deployment rather than after it.
• This is the emerging-technology edge of the same visibility argument Dunnell makes about plant equipment.

The remark is short and direct: agents are always accessing data, but now they're also making decisions. The post frames the need to monitor agents and keep security in place while AI delivers value, and links to a video.

Read original

Across five transformations, the near-failure point was the same every time, and it was never the strategy, the technology, or the organizational design — it was the cadence: the weekly sales reviews, biweekly supply chain reviews, and monthly budget accountability sessions that nobody puts on a roadmap or celebrates, but which hold the effort together. These routines feel boring and redundant precisely when they matter most; sustaining them is the actual work of transformation.

The examples are deliberately ordinary — weekly sales reviews, biweekly supply chain reviews, monthly budget accountability sessions. Remove them, Dunnell argues, and the whole program drifts even though the plan was sound.

Read original

Announcing a move into a global manufacturing go-to-market role at ServiceNow, Dunnell set out the themes underpinning the work — AI, IT/OT convergence, supply chain resilience, and connecting plant-floor performance to enterprise outcomes — and described ServiceNow as an AI control tower that connects and orchestrates across the manufacturing value chain in ways no point solution can.

• They describe the job as building the commercial architecture that turns agentic AI and platform capability into measurable business outcomes for large manufacturers.
• Weigh the security and convergence arguments with that frame in view: Dunnell works for a company that sells the orchestration layer.
• The cadence argument is the exception; it does not depend on a platform.

The post names the manufacturing shift, positions ServiceNow against point solutions, and states the intent to lead the global go-to-market motion for the vertical and turn platform capability into measurable business outcomes. It thanks named ServiceNow colleagues and others for the opportunity and support.

The ServiceNow announcement, which says the role began on Monday, July 13, is the earliest vantage point. The security, AI-governance, and event posts that follow express the role's themes in public.

Read original

Start here for the mechanism, not the slogan.

A controller fails on a CNC machine.

• Maintenance swaps it and restarts the line before the shift ends, but the replacement part runs firmware two versions behind and carries a flaw already used in attacks on other plants.
• Nobody notices because the work order and vulnerability list sit in different systems.
• Six weeks later, a scan finds the exposure, and fixing it means stopping a line running at full capacity.
• Dunnell calls this an architecture failure, not a technology failure: safety and security are looking at the same machine but cannot see what the other has done to it.
• The post notes that manufacturing has been the most-attacked industry for five years running and cites a Siemens figure of $2.3 million per hour for an idle automotive line.

For brands planning a campaign

A strong opening hook for OT/IT convergence or plant-security campaigns: it reframes the pitch as a records-and-coordination problem, not a product gap. A vendor whose message is "give safety and security one record to read" gets a ready-made narrative setup.

The three-way stalemate clearly shows how delay takes hold inside a plant.

Security flags an exposure on a production line.

• Maintenance says it cannot be touched for three weeks.
• Production says the customer needs parts Friday.
• Nobody moves.
• Dunnell frames this as fragmentation, not a strategy problem, and points readers to a recorded conversation with Tom Wilk at IMTS 2026 on the Great Question: A Manufacturing Podcast.
• They discuss why data split across OT, IT, security and maintenance costs weeks of delay, and what one manufacturing leader did to fix it in 90 days.

For brands planning a campaign

Useful for podcast, event or sponsored-conversation formats; the "fixed in 90 days" detail suggests a case-study angle. The outcome claim comes from the post's description of the episode, so ask for the recording before repeating it as fact.

A good first read for newcomers who want the thesis before the detail.

When maintenance fixes a machine, security does not see it.

• When security isolates a compromised device, safety does not know.
• One piece of equipment, two systems, zero visibility—and that fragmentation is where production shutdowns begin.
• The post links to a longer piece on how companies are closing the gap.

For brands planning a campaign

Works as a pull-quote or social teaser, compressing the argument into a few lines. On its own, it is thin; pair it with the controller-failure post, which supplies the mechanism.

It clearly signals that the AI argument is about governance and monitoring, not adoption enthusiasm.

The post describes IMTS 2026 in Chicago, where Keith Dunnell is quoted saying that AI agents are always accessing data, but now they are also making decisions. It argues that manufacturers need to monitor agents and maintain security while AI adds value to the organisation, and links to a video.

For brands planning a campaign

Relevant for agentic-AI or AI-governance campaigns, where the differentiator is oversight rather than capability. Ask for the linked video before relying on claims about what was said on camera.

It is the execution-side counterweight to the technology posts.

Across five transformations, the near-failure point was never the strategy, technology or org design, but the cadence: weekly sales reviews, biweekly supply chain reviews and monthly budget accountability sessions.

• None appears on a transformation roadmap or gets celebrated, but remove them and the whole effort drifts.
• Dunnell concludes that transformation is a rhythm to hold, especially when it feels boring and redundant—and that holding it is the job.

For brands planning a campaign

A keynote-ready theme for change-management, operating-model or execution-focused briefs—and one of the few items here that does not rely on a security or platform pitch.

It establishes the professional vantage point behind the other posts.

An announcement that the role begins on Monday, July 13th.

• The post describes manufacturing as being in the middle of the biggest operational shift in a generation, driven by AI, IT/OT convergence, supply chain resilience and pressure to connect plant-floor performance to enterprise outcomes.
• It presents ServiceNow as an AI control tower that orchestrates across the manufacturing value chain in ways point solutions cannot.
• Dunnell says they intend to lead the global go-to-market motion for the manufacturing vertical and build the commercial architecture that turns agentic AI and platform capability into measurable business outcomes, thanking named colleagues from ServiceNow and elsewhere.

For brands planning a campaign

Essential context for any speaker or advisor brief: the perspectives on plant security and transformation come with a vendor's commercial frame, so build a neutral-advisor positioning deliberately rather than assuming it. Also useful for campaigns seeking the platform-side view of agentic AI in manufacturing.

One post walks through a CNC controller failure: maintenance swaps the part and has the line running before the shift ends, but the replacement is two firmware versions behind and carries a flaw already used against other plants.

• The work order and vulnerability list sit in different systems, so a scan flags the exposure only six weeks later, when fixing it means stopping a full production line.
• They call this an architecture failure rather than a technology failure: safety and security are looking at the same machine but cannot see what the other has done to it.
• The posts support the point with two cited figures — manufacturing as the most-attacked industry five years running, and Siemens' estimate of $2.3 million per hour for an idle automotive line.
• The stated fix is giving both functions one record to read rather than merging the disciplines.

Opens with a concrete operational scene, withholds the consequence until the middle, then names the architectural cause.

For brands planning a campaign

Strong material for OT security, IT/OT convergence, and plant resilience briefs. Expect recommendations to land on a shared data or records layer rather than a discipline-level or tool-level fix, so campaign assets should show cross-team visibility, not another point product.

Examples: Source 1 · Source 2

The post describes ServiceNow as an AI control tower that connects and orchestrates across the manufacturing value chain in ways no point solution can, and describes the job as building the commercial architecture that turns agentic AI and platform capability into measurable business outcomes for large manufacturers.

For brands planning a campaign

This is an explicit vendor position, not neutral commentary.

• A brand assessing Dunnell for independent advisory work should expect ServiceNow-shaped recommendations.
• The strongest fit is a platform-aligned campaign, partner content, or sponsored industry conversation; category comparisons that require vendor neutrality are a weaker fit.

Examples: Source 1

Their framing is that monitoring agents and maintaining security have to happen while AI is adding value, not after. This is the emerging-technology edge of the same argument about fragmentation: visibility and oversight have to keep pace with what the machines and software are doing.

For brands planning a campaign

Usable for campaigns on AI governance, agent monitoring, or industrial AI adoption where the audience is operations and security leadership rather than data scientists. The message pairs naturally with a control, oversight, or auditability product story.

Examples: Source 1

The examples are deliberately ordinary: weekly sales reviews, biweekly supply chain reviews, monthly budget accountability sessions.

• None appears on a transformation roadmap or gets celebrated, but removing these sessions lets the whole program drift.
• The conclusion is that transformation is a rhythm that has to be held, especially when it feels boring or redundant.

Argues from a repeated pattern across unnamed engagements, then reduces it to a single unglamorous mechanism.

For brands planning a campaign

A useful counterweight when a brief needs execution credibility rather than product messaging. Fits change-management, operations-excellence, and transformation-program campaigns aimed at leaders skeptical of large platform narratives.

Examples: Source 1

Dunnell calls this fragmentation rather than a strategy problem. The post points to a podcast conversation recorded at IMTS 2026 covering why fragmented data across OT, IT, security, and maintenance costs weeks of delay, and what one manufacturing leader did to fix it in 90 days.

Ask for the linked podcast conversation and the longer article referenced in a separate post; the posts do not describe the specific 90-day method or details in those pieces.

For brands planning a campaign

Best used as a hook for a problem-first campaign asset about downtime, escalation, and cross-team decision rights. Because the resolution sits in a linked piece, ask for the specifics of the 90-day example before building creative around it.

Examples: Source 1

Statistics are borrowed from outside sources rather than original measurement. Public appearances at IMTS 2026 include live-audience and podcast formats.

Scene, failure, reframe, sometimes closing with a link to longer work.

For brands planning a campaign

A brand can expect short, quotable, operations-flavored posts and event or podcast appearances rather than long-form written analysis. Before committing to a multi-part content series, check the creator's publishing history; the longer formats referenced here sit off-platform and were not reviewed.

Examples: Source 1 · Source 2 · Source 3 · Source 4 · Source 5 · Source 6

They name Monday, July 13, as their start date. A second post from the same account also identifies them as ServiceNow's VP and global head of manufacturing industry and describes their comments on AI agents in manufacturing.

View example

Their posts document attendance and speaking, not a commercial agreement with the event.

View example

They discussed why fragmented data across OT, IT, security, and maintenance costs manufacturers weeks of delay, and how one manufacturing leader addressed the problem in 90 days. The post documents a guest appearance; it does not state terms or payment.

View example

Their IMTS 2026 reporting works like Dunnell's: name a real operating constraint, then explain what it costs a team. Both use the show to gather problems rather than announce products.

Gafvert's constraint is physical and rooted in materials and additive manufacturing: gripper geometry, durometer, tooling lead times, and getting a part from CAD to a testable, engineering-grade part in a day.

• Dunnell's is informational, centered on records and visibility across safety, security, and maintenance.
• Follow Gafvert for hardware and production-iteration content; follow Dunnell for OT/IT security and cross-team decision structure.

Read original · Original 2

Both make their case through numbered operational lessons, stress context around machine data, and urge leaders to start with a business problem rather than a tool.

Khan stays inside the data and integration chain, connecting ERP, MES, planning, quality, and finance so AI can act on operations.

• They write as a data-engineering practitioner at Fuzzitech.
• Dunnell approaches the same plant from cybersecurity and records architecture, focusing less on data pipelines and more on who can see what a machine has done.
• Choose Khan for data foundations and AI value realization, Dunnell for security, governance, and operating rhythm.

Read original · Original 2

Mohan's framework requires human sign-off and sets explicit trust rules around agent output, echoing Dunnell's call to monitor agents as they add value. Both also use IMTS 2026 and the Google Cloud partnership as the setting for that conversation.

Mohan explains a specific open framework, the OKF Industrial Profile, and a vendor partnership that pairs Gemini reasoning with machine telemetry.

• The posts are closer to product and framework announcements.
• Dunnell makes a narrower argument about visibility between disciplines and does not present a technical standard.
• Follow Mohan for how industrial AI context is being packaged and shipped; follow Dunnell for why fragmented operational records cause failures in the first place.

Read original · Original 2

Gafvert's two posts work through one concrete problem: a robotics team needs a better gripper or a redesigned bracket; a new geometry or durometer normally means a new tool and weeks of waiting.

• The pitch is that a part can go from CAD to a testable, engineering-grade part in the same day, on the same material, without redesign.
• A brief seeking a speaker or advisor to explain why tooling lead times stall a robotics programme, or to walk an audience through material selection under heat, saltwater, and load, gets a tested narrative here.
• The second post extends that argument to defence programmes, construction robotics, drone delivery, underwater drones, and race cars, all with the same requirement: parts with no porosity that hold up under stress and extremes.
• For robotics hardware development or additive materials, Gafvert is closer to the buyer's question than Dunnell.

The tradeoff is topic and framing.

• Neither reviewed Gafvert post covers OT/IT convergence, the split between work orders and vulnerability lists, or the review cadence that keeps a transformation on track—the territory of Dunnell's material.
• Gafvert's argument points to a specific materials-and-software combination, so a brief needing a category-neutral supplier comparison is not a fit.
• One post closes with an invitation to connect at a show, making it a lead-generation pitch rather than an editorial argument.
• For a security or transformation keynote, choosing Gafvert means giving up the incident scene, the idle-line cost figure, and the execution-rhythm argument.

Examples: Source 1 · Source 2

Khan's two closing takeaway posts from a manufacturing technology show offer a repeatable decision path: shop-floor signals become context, insight, a recommended action, execution, and a business result.

• That chain has to reach ERP, planning, inventory, purchasing, supply chain, cloud, and finance because the business problem usually spans the operation.
• The advice is practical and specific: start from the outcome rather than what AI can do, use process knowledge, machine history, quality patterns, and operator experience as competitive advantage, and finish a small number of high-value use cases instead of attempting everything.
• Khan also raises a question Dunnell only touches: who owns the operational data and institutional knowledge converted into AI systems.
• For a campaign about use-case selection, data foundations, or where intelligence belongs at the edge versus the cloud, Khan offers a coherent, teachable structure.

Khan's posts are structured lists of lessons rather than scenes.

• There is no controller swap, three-way standoff between security, maintenance, and production, or named cost of an idle line, so a keynote built around a vivid incident loses its strongest asset.
• The posts also end with a company position and an invitation to message, the same vendor-aligned posture as Dunnell's, rather than a neutral comparison of approaches.
• Khan's focus is mid-size manufacturers and operational excellence; a brief centred on OT/IT security architecture should look to Dunnell or another specialist.
• Khan addresses context, integration, action, and ownership, but does not work through a security exposure end to end.

Examples: Source 1 · Source 2

Mohan's posts describe an open framework that turns thirty years of operator judgement into plain files a process engineer can review: operator heuristics with human sign-off, trust rules for incoming readings, and cause-and-effect relationships.

• The problem is precise and quotable for the right audience: intuition about how a line behaves on a cold start, or what a specific pressure signature meant the last three times it spiked, exists but is trapped in email threads and people's heads—and retires when they do.
• A second post pairs reasoning models with machine telemetry and operational context to deliver role-based agents to manufacturing teams.
• A brief seeking a speaker on governing AI agents, capturing retiring expertise, or why vertical-specific context is where foundation models earn their value will find more usable substance here than in Dunnell's posts, which raise agent decisions once without developing them.

Both posts are company announcements: one launches a framework; the other congratulates a partner on a platform milestone and points to a booth demo.

• They offer a position and a partnership, not a close reading of a plant-floor failure.
• There is no security incident, maintenance-versus-production conflict, or transformation-cadence argument—the core of Dunnell's published material.
• A campaign that needs an advisor to open with a recognisable operational breakdown and move from it to a recommendation is better served by Dunnell.
• For agent governance and knowledge-capture expertise, choose Mohan, but verify the framework's results before repeating any claim about it in campaign copy; the announcement-heavy framing offers little evidence of outcomes.

Examples: Source 1 · Source 2