How this is scored
Each answer scores 0 to 4 and is multiplied by that question's weight. The data foundation questions carry the heaviest weight, because a plant with no readable data cannot judge an AI vendor no matter how sharp the leadership team is. The weights add up to 25, so the top possible score is 100.
There are three bands. Not ready to evaluate is 0 to 44, Nearly ready is 45 to 69, and Ready to run an RFP is 70 to 100. The band is what decides whether an evaluation is worth starting, and most plants that ask us this question land in the bottom two.
One rule sits on top of the arithmetic. Five questions decide whether an evaluation can even be run: how much data software can read, whether machine data leaves the PLC, whether your systems talk to each other, whether leadership agrees on the goal, and whether you have a baseline to judge a result against. If those five come back low, the band is held down to what they support, no matter what the total says. A plant can score well on budget and scope discipline and still have nothing an AI could read and no way to tell whether it worked. The band follows the foundation and the alignment, not the average, and when the rule moves your band the result says so on screen.
Why this comes before the RFP
An AI vendor evaluation is a test you give the vendors. The trouble is that a plant that is not ready gives the test to itself and fails without knowing it. Every demo runs on the vendor's clean sample data, every deck promises the outcome, and with no readable data of your own and no agreed definition of success, there is no way to separate the tool that would work on your floor from the one that only works in the demo.
The readiest buyers do an unglamorous thing first: they make sure the data an AI would read actually exists, they name a person who owns it, and they write down the single outcome and the number they will judge it by. Then the evaluation becomes a real test instead of a beauty contest. This scorecard is a five-minute version of that check.
Want the deeper version of the data side of this? Work through the AI Readiness Checklist, or have it scored for you with the AI Readiness Assessment.
What Harmony does about it
Harmony builds the readable data layer this scorecard is measuring for. The published pilot is $15-20K one time over 4-6 weeks, with forward-deployed engineers living in your plant and working software by the end of the pilot. Harmony is software and hardware agnostic and connects at the PLC over OPC UA, on Rockwell/Allen-Bradley, Siemens, Omron and Mitsubishi. That is Phase 1, Lay the Data Foundation, the work that has to exist before any AI vendor, including us, has something real to run on.