← All articles

· Crossover Software

Decisions First, Data Second: A Method for Starting with AI on the Factory Floor

On 10 November 2026, IIoT World published an interview with Keith Moore, CEO of AutoScheduler.AI: mapping decisions before cleaning up data avoids endless delays. The focus is warehousing, but the method is relevant to MES, OEE and line monitoring too.

Smart factoryIIoTData qualityExplainable AIOEE

Many industrial digitalization projects stall before they begin: data is incomplete, inconsistent and scattered across different systems, so the decision is made to fix it before introducing any artificial intelligence tool. On 10 November 2026, IIoT World published a summary of a video interview from its CxO series with Keith Moore, CEO of AutoScheduler.AI, which challenges this approach. According to IIoT World, the summary was produced with the help of AI tools.

The risk of endless data cleanup

The article argues that large multi-year data-cleaning programs often end up postponing operational improvements indefinitely. Moore's proposal is to reverse the order: first map the decisions the site has to make, then determine how good the data needs to be to support them.

"Your data only has to be as perfect as the decisions you're going to make on it."

The five steps of decision mapping

The method described follows this sequence:

  • catalog the data the plant already has;
  • document how decisions are made today;
  • trace how data feeds those decisions and which events cause them to change;
  • distinguish decisions that can be standardized across the network from those specific to a single site;
  • rank decisions by impact against data maturity, and start automating with the most significant ones.

The practical benefit is a clear scope: there is no need to make the entire information estate perfect, only the part that supports the highest-impact decisions.

Operator trust depends on explainability

An experienced operator will not blindly follow a machine's recommendation. For this reason, according to Moore, AI must behave as an explainable decision agent. Each application records the equations, constraints and trade-offs behind its choices. Before go-live, the platform shows its logic to the people who configured it and asks whether they would handle the trade-offs differently; their answers are stored alongside the rules.

On the shop floor, if an operator questions a recommendation, the system retraces the data along the decision flow and shows its reasoning. The operator can ask for the logic to be changed, resulting in a documented, verifiable trail.

Domain-specific AI versus general-purpose models

AutoScheduler.AI also argues for the superiority of AI built for a single domain. General-purpose models, the article claims, do not understand physical constraints: equipment handling times, restricted loading docks, the knock-on effect of a late truck on the plan. The platform was originally developed for Procter & Gamble.

What plant managers can take away

The interview is set in warehousing, not on the production line: applying the method to MES, OEE or line-monitoring projects, for example in poultry processing, is our own extrapolation. It should also be noted that the source is an interview with a vendor's CEO and reports no results or metrics. With these caveats, the idea is useful as a working outline:

For a production or IT/OT manager, a possible starting point is to:

  • list the recurring decisions (changeovers, order priorities, response to a stoppage) and who makes them;
  • check which data really supports them and which is already sufficient;
  • start automation or decision support with high-impact decisions where the data is already reliable;
  • build in, from the outset, the recording of the rules and reasons behind each recommendation, so operators are involved.

In short, data quality should be measured against how the data is used. Starting from what you have, with transparent logic that the shop floor can challenge, can shorten time to launch without waiting for a hypothetical perfection.