Illustrative Case · Medical Devices
A representative composite based on patterns we see across comparable mid-market manufacturers. Figures are illustrative. No individual client is depicted.
Seven dimension AI readiness index, out of 100.
Found in tail spend once it became visible.
New spend now classified automatically at 95% accuracy.
A planned AI analytics rollout, resequenced so the data could carry it.
The situation
An AI decision made before anyone checked the data.
Leadership had approved an AI spend analytics platform, with a license commitment of about $1.1M over three years. The vendor demo was impressive. The question nobody had asked was whether the company's own spend data could support it.
Procurement suspected it couldn't. Spend sat across two ERPs and a legacy purchasing tool, categories were coded differently by site, and the tail of small suppliers was effectively invisible.
The diagnosis
Ready on paper, not ready in practice.
We scored the organization across seven readiness dimensions before any contract was signed. Systems integration scored well. Everything the AI would actually depend on did not, and readiness is set by the weakest dimensions, not the average.
- Data governance scored 1.4 out of 5. No one owned the spend data, so no one fixed it.
- Category coding differed across five sites, so the same purchase landed in different categories.
- About 2,900 tail suppliers made up 9% of spend and almost none of the attention.
- No agreed use case. The platform had been bought for "visibility," with no decision it was meant to improve.
The dashed shape shows the starting profile, with the weakest dimension marked. The solid shape shows the profile after twelve weeks. The platform was deployed only once every dimension cleared the threshold.
What we did
Fix the data first, then let the AI do its job.
We paused the rollout, fixed the dimensions holding readiness down, and agreed one concrete use case before the platform went live.
Operating flow
Mapped how spend was coded at each site and rebuilt one shared category taxonomy, so every purchase lands in the same place wherever it is made.
Enabling technology
Cleaned and merged supplier records across both ERPs, then configured the platform's automated classification against the new taxonomy.
Adoption in practice
Named data owners for each category and agreed a first use case: tail spend consolidation, with a clear savings target.
The outcome
The AI went live on data it could trust.
Readiness
The readiness score moved from 41 to 78, with every dimension above the deployment threshold before go live.
Savings
With tail spend finally visible, the team identified $3.9M in consolidation and price savings in the first quarter.
Capacity
88% of new spend lines are now classified automatically at 95% accuracy, freeing the team from manual coding.
What made it stick
Owners, not just tools.
AI classification degrades quickly when nobody owns the data underneath it. Here, each category has a named data owner, exceptions are reviewed weekly, and the readiness index is rescored twice a year. The platform keeps working because someone is accountable for what it reads.
Next step
Considering AI for your spend? Check readiness first.
A short conversation is the fastest way to find out.
