Tribal Knowledge to AI-Guided Execution: What We're Building

At Automate in Chicago, our Chief Product Officer, Rutherford Wilson, took the Innovation Stage to talk about the problem every multi-site manufacturer knows, and few have closed: the gap between what the schedule promised and what the floor actually did. The room was filled past capacity. The heads kept nodding. The execution gap is an expensive problem that most operations have learned to live with, and the floor already knows it.

Watch the full session: MachineMetrics CPO Rutherford Wilson on closing the manufacturing execution gap, from the Automate Innovation Stage.

The schedule is the hypothesis

Every production schedule is a best guess. The plan leaves your ERP looking perfect: every job released, every due date met, the day ending on time. Then it meets the floor. Two machines go down for maintenance you did not plan for. An operator on a critical cell calls out. A machine has to come down because maintenance knew about a job that operations did not.

Your ERP is optimistic. Your machines are honest.

The plan was right until the moment it was not. That space between the plan and reality is the execution gap, and it is where on-time delivery slips, capacity disappears, and margin leaks out in increments no one captures. It is rarely one big failure. It is a hundred small ones a week.

Your ERP has the plan. Your floor has the truth. The work is closing the distance between them while the shift is still running.

Your factory already has a brain

The knowledge that closes the execution gap already exists inside your operation. It lives in the heads of your senior operators.

It is the person who can hear a machine and tell you it will fail in two weeks. It is the scheduler who knows a part can only run on one machine because of its length, a fact that lives in no BOM and no system anywhere. That is tribal knowledge. It is some of the most valuable operational data you have, and it works right up until the day Bob retires.

The question Rutherford put to the room: how do you get that knowledge out of your operators' heads and into a system that can surface it to the next operator at the exact moment they need it?

Closing the gap with real-time data

MachineMetrics is a manufacturing operations platform. We are not an MES, though we share capabilities with one. We work across four areas: connecting to your machines and systems, executing work on the floor, optimizing how that work runs, and building custom applications for the problems unique to your operation.

The foundation is data. We pull real-time telemetry off your machines and combine it with the plan in your ERP. That puts the plan and the reality side by side, in the moment, instead of in a morning-after report that only tells you what already went wrong. By then, it is too late to act. AI-guided execution puts the right information and the right next step in front of the right person while they can still do something about it.

The small fixes that compound

Closing the execution gap is rarely one big thing. It is a lot of small, accurate fixes that add up. A few Rutherford walked through:

Dispatch scheduling on real cycle times. The cycle time in your ERP is often wrong. We calculate completion using the actual cycle time measured on the machine, updated every cycle. Jobs show as on time, at risk, or already late, and you reschedule by dragging a card. Your ERP still owns the plan. This gives the floor a live view of how to hit it.

Job tracking that matches reality. When a strong operator running three machines gets everything started and logs the work 90 minutes later, the timestamp punishes them and corrupts your OEE. We anchor the record to when the machine actually started, so the numbers reflect what happened.

Reclaiming idle time at inspection. First article inspection often leaves a machine idle while the part sits in the QA queue, and the operator moves on. The platform flags that idle time so you recover it instead of paying for it.

Digital shift handover. The black book and the sticky notes become a spoken note. Max, the AI built into the platform, cleans it up, ties it to the jobs that actually ran, and surfaces it to the next operator at startup. Nothing gets lost between shifts.

From surfacing knowledge to acting on it

Max sits underneath all of it. Max knows how your machines are running, who is running efficiently, what is scheduled, and where the day is at risk. Ask Max where a late start is coming from, and it walks you through how to close it.

The next piece is the Knowledge Hub: a place to capture the operator know-how that lives nowhere else today. Part ABC runs best on machine five. This is how you clear that recurring fault. Once that knowledge is in the system, Max can surface it to any operator on any shift, in their display, the moment the relevant job starts. The expert who only works first shift stops being the only person who can keep the job moving.

Building for problems that are one of a kind

No platform ships an out-of-the-box answer for every problem, and any vendor who tells you otherwise is selling you something. Some of your problems are specific to your operation. Some are temporary. Solving them used to mean a long, expensive integration project.

We are building toward a different model. With Carbide, the MachineMetrics application builder, you describe the application you need in plain language and build it on your own live production and ERP data, without the six-figure integration bill.

Rutherford demoed one he built himself: a cycle time validator that profiles production over the last 30, 90, or 180 days and shows exactly where actual cycle times drift from the ERP standard, with a recommendation to update. About twenty minutes to build. If the first version is wrong, you describe it again and rebuild. Building software is cheap now. The value is in describing the problem clearly.

This is not a trick that only works for a CPO on a stage. In March, we brought five customer teams to Production Lab, a two-day build event at our Massachusetts headquarters, and asked each to solve a real problem on their own live production data. Every team shipped a working application by the end of day two. The people who built them were not software engineers. They were a continuous improvement specialist with no coding background, a director of training, a director of manufacturing analytics, an IT manager, and a director of operations across five different companies. One built a forklift dispatch system from scratch. Another built the first tool offset analytics application on the platform, almost entirely on his own. Here is what they built at Production Lab.

The point is not that the building looks impressive. It is that the people closest to the problem can solve it themselves, in days, without waiting for a roadmap.

The gap shows up in the numbers

The execution gap is not abstract. It lands in on-time delivery, overtime, consumable waste, utilization, and cost per part. The floor touches every number that matters, and most of those numbers stay invisible until the month-end report, long after anyone could act on them.

That is the case for leaders as much as operators. Operators get live guidance at the machine. Leaders get a real view of production across every machine, cell, and plant. Everyone works from the same truth, while there is still time to change the outcome.

The early results we are seeing

We do not just argue that the legacy approach is too slow. We show the better way and prove it.

On scheduling, Harvey Performance improved schedule attainment by more than 25 percent.

On tooling, one customer shortened tool changes by building an application that predicts when each tool will hit its preset. Operators stage the next tool before the machine stops instead of scrambling after it does. They went further and built one screen showing the upcoming changes for every machine on the floor, so a small team works the changes like a pit crew. Tool changes happen every shift, every day. Time recovered their compounds fast.

What we are building toward

A few of the pieces coming next:

Tool intelligence. Tool-life prediction extended with tool cost and machine-fit data, so Max can recommend not just when to change a tool, but how to sequence changes to protect a lights-out shift. Coming later this year.

The MachineMetrics MCP Server. A secure way to bring your MachineMetrics context, machine status, downtime, and the ERP plan into the AI tools your team already uses. Ask "what is at risk today" inside your own workflow and get an answer grounded in live shop floor data.

A starter application library. Well-built templates you copy and adapt to your own operation, so no one starts from a blank screen.

These are the pieces we will put in front of manufacturers at IMTS 2026 in Chicago, September 14 to 19, booth 133100.

See what is happening. Know what matters. Act before production slips. If the execution gap is costing you on-time delivery and capacity, come see how AI-guided execution closes it.

Frequently asked questions

What is the manufacturing execution gap? The execution gap is the space between what the business planned and what the floor actually did. Machines go down, operators are out, priorities shift. The gap is where on-time delivery and capacity quietly erode.

What is tribal knowledge in manufacturing? Tribal knowledge is the operational know-how that lives in experienced operators' heads rather than in any system: which machine a part runs best on, how to clear a specific fault, when a machine is about to fail. It is valuable and easily lost when people retire.

How does AI help with shop floor execution? AI-guided execution combines real-time machine data with the ERP plan to flag risks as they happen, recommend the next best action, and surface captured operator knowledge to the person who needs it, in the moment they can still act on it.