AI READINESS ASSESSMENT

AI Readiness Assessment for Manufacturers

By Jason Osajima — former VP of AI at a $250M manufacturer ·
Quick answer

An AI readiness assessment for manufacturers from an operator who shipped it: the 4 pillars, a scoring rubric, and the data/access gaps that kill projects.

An AI readiness assessment for manufacturers should answer one blunt question: if you green-light an AI project Monday, will it ship or will it stall? Most won't, and the reason almost never shows up in the vendor pitch. It shows up in your item master, your ERP access rights, and whether anyone on the floor actually owns the workflow. I ran AI into production at a $250M manufacturer, and I've watched plenty of plants spend six figures discovering after the fact what a one-week readiness assessment would have told them up front. This is the assessment I'd run before spending a dollar.

What an AI readiness assessment actually checks

Forget maturity-model wall charts with 47 dimensions. For a mid-market manufacturer, AI readiness comes down to four pillars. Score each one honestly, because the lowest score is your real ceiling. A 9/10 on strategy and a 2/10 on data access gets you a 2.

Pillar 1: Data readiness

AI agents are only as good as the data they read. The assessment checks:

Pillar 2: Systems and integration access

This is the pillar that quietly fails the most projects, and the one vendors gloss over. The assessment checks whether you can read AND write to your systems of record:

If an agent can't write its decision back into the system of record, it's a research tool and a human still re-keys everything. No write-back, no production. Settle this in week one.

Pillar 3: Process fit

Not every workflow is worth automating. The assessment ranks candidate workflows by automation value:

The sweet spot: high volume, high handle time, exception-rich, with decision logic you can describe.

Pillar 4: People and ownership

The pillar everyone skips. The assessment checks:

Agents without owners drift. This pillar is why IT-led pilots get adopted by nobody.

A scoring rubric you can run this week

Score each pillar 1-5. Your readiness is gated by the minimum, not the average. Here's the rubric I use.

Score Data Integration Process fit People
1 Master data a mess, no idea of duplicate rate No API, no path, locked systems Can't name a high-volume workflow No owner, no bandwidth
3 Master data usable, some cleanup needed API exists, write-back access takes weeks Candidate workflows identified, logic partly tribal Owner exists but stretched
5 Clean master, cross-refs mapped, unstructured mapped Read/write access available now Top workflow baselined, logic documented Named owner, willing team, bandwidth allocated

How to read your score:

The three gaps that actually kill projects

Across the plants I've seen, the same three readiness gaps account for most failures:

  1. Dirty master data. The agent surfaces every wrong lead time and duplicate SKU. "AI failed" is often "your data was always wrong and now it's visible."
  2. No write-back access. The integration to write decisions back into the ERP is the hard, unglamorous part. Teams discover the permission wall in month three instead of week one.
  3. No owner. A workflow with no human owner gets an agent nobody tunes, trusts, or adopts.

Notice that none of these are about the AI model. The model is the easy part now. Readiness is about your data, your access, and your people.

Readiness assessment vs. jumping straight to a pilot

Skip the assessment Run the assessment first
Time to start Faster (days) 1-2 weeks slower
Risk of stall High Low
Cost of failure Six-figure dead pilot One week of analysis
Knows its first agent Guesses Ranked, evidence-based
Finance buy-in Hard (no baseline) Easier (gated plan)

The one to two weeks an assessment costs is the cheapest insurance you'll buy on the whole program.

Get your readiness scored for free

If you run ops or IT at a $100M-1B manufacturer, the smartest first move isn't picking a vendor, it's scoring your readiness across these four pillars so you know whether your next AI dollar will ship or stall. Our free First 5 Agents teardown does exactly that: we run the readiness assessment on your real data, systems, and workflows, then name the five agents that pay back first for your operation. Book a 30-minute call and bring a sample of your item master and one workflow's exception report. You'll leave with a readiness score and a ranked shortlist, not a sales pitch.

Let's see what's worth building first.

A 15-minute call: tell me where your AI or planning is stuck, and I'll tell you the one thing worth building first — and whether it's worth doing at all.

More field notes

An AI Strategy Playbook for the Manufacturing COOHow to Prioritize Your First AI Use CaseAI Change Management for Plant and Ops TeamsThe AI Maturity Model for Manufacturing Ops