AI Vendor Evaluation

Score the shortlist before you sign.

Two weeks, $7,500, fixed. Your AI vendors scored against your actual requirements: capability on your real cases, security and HIPAA posture, what happens to your data, integration cost, and the build-versus-buy math. A recommendation with the reasoning shown.

What you get

Four artifacts, two weeks.

1
Week 1

Requirements matrix

What the tool must do, drawn from your workflows and two or three staff interviews, ranked must, should, nice. Written before any vendor is scored, so the scoring can't be steered by the best demo.

2
Week 1–2

Scored comparison

Up to five vendors against the matrix: capability on de-identified cases from your organization, security and HIPAA posture (BAA terms, subprocessors, retention, whether your data trains their models), integration effort, and true cost over three years.

3
Week 2

Recommendation memo

Buy X, build instead, or wait, with the build-versus-buy math in dollars. If the honest answer is "none of these," that is the deliverable, and it is written to stand on its own if you take it to another consultant or your board.

4
Readout

Negotiation points

The contract terms to demand before signing: data-use limits, AI training carve-outs, exit and export rights, service-level teeth. Aligned with Kent Applied's own AI Use Addendum so your paper and the vendor's can coexist.

  • Fixed price, 50% to start, 50% on delivery
  • Independent: no software sold, no referral fees
  • No PHI required

Built for

  • Organizations mid-decision, with sales decks piling up and nobody whose job is to check them
  • Hospitals and medical groups evaluating ambient documentation, prior-auth automation, patient messaging, or revenue-cycle AI
  • Digital health and payer teams choosing a model provider or an AI platform
  • Anyone whose counsel has asked "what exactly happens to our data" and gotten a slide

Why an outside score

The principal has sat on both sides: shipping production AI products end to end, and evaluating what vendors claimed against what their systems did. Demos are optimized to be believed. A requirements matrix written before the demos, and tests on your own cases, are how you buy the thing that works on Tuesday morning.

Every evaluation states which AI tools touched which data and keeps the whole engagement PHI-free.

Questions

What buyers ask first.

Do you resell or get referral fees from any AI vendor?

No. Kent Applied sells no software, resells nothing, and takes no referral fees. The recommendation has exactly one customer, and sometimes it is "none of these" or "build it."

How many vendors can be on the shortlist?

Up to five in the two-week engagement. A longer list gets cut to five in week one against the requirements matrix, which is usually the most useful thing that happens all engagement.

Do you need our patient data to test the vendors?

No. Capability tests run on de-identified or synthetic cases you approve. A vendor's handling of real PHI is evaluated on paper: their BAA, subprocessors, retention, and training terms.

What if the right answer is to build instead of buy?

Then the memo says so, with the cost of building shown next to the three-year cost of buying. Kent Applied can build it as a fixed-price Workflow Build, but the memo is written to stand on its own if you take it elsewhere.

Can you join the vendor negotiation?

The deliverable includes the contract terms to demand: data-use limits, AI training carve-outs, exit and export rights, service-level teeth. Joining calls is available as a small add-on if your counsel wants a technical voice in the room.

Contact

Send the shortlist.

Twenty minutes. Bring the vendors, the deadline, and what you're afraid of. You'll get a fixed-price proposal within 48 hours.

Email to book a call →

Or connect on LinkedIn.