World Model Readiness
Engraved commercial instrument

For your company

Money & Vendors

Module · Test the claims before you sign

The Vendor & Pitch Check

Every vendor demo works. The question is whether it works on your data, your edge cases, and your terms. This module checks whether you can put a claim to the test before you commit: proof on your own data, an exit you can actually take, ownership of what you feed in, and references who will tell you the truth.

Question 1 of 5 · Proven on your data

Can you make a vendor prove their claims on your data before you sign?

A demo runs on data chosen to make it shine. A proof of concept on your messy, real data is the only test that predicts what you will actually get. Vendors who resist it are telling you something.

Question 2 of 5 · Exit is written in

If the tool underperforms, can you leave without a penalty that traps you?

The time to negotiate the exit is before you sign, when you still have leverage. Auto-renewals, multi-year lock-ins, and painful offboarding are how a disappointing tool becomes a permanent one.

Question 3 of 5 · You own your data

Is it clear, in writing, that your data and its outputs stay yours?

Read the clause on training and reuse. Some vendors treat the data you feed in, and the model improvements it produces, as theirs. Ownership decided after a dispute is ownership you have lost.

Question 4 of 5 · Lock-in is limited

If you wanted to switch vendors, how much would be built to trap you?

Proprietary formats, embedded workflows, and data you cannot export cleanly all raise the cost of leaving. Lock-in is not one decision: it accretes, until switching costs more than staying with a tool you have outgrown.

Question 5 of 5 · References were real

Have you spoken to customers the vendor did not hand-pick for you?

The three references a vendor offers are their happiest clients. The ones worth finding are the customers who churned, or who are your size in your industry. LinkedIn finds them; the vendor will not.

For the statistics · one click each

Three questions for the public picture

These do not affect your score. They feed the anonymised, aggregated statistics; groups under 8 respondents are never shown.

How does your AI vendor selection usually work?

Chosen on the pitch
Compared a few
Piloted before buying
Formal evaluation
No process yet

Do you test AI vendors on your own data before signing?

Never
Sometimes
Usually
Always
Not bought any yet

Have you been locked into an AI or software vendor before?

Not that we recall
Mildly
Yes, badly
Still trapped now
Not sure

Your context

Used to calibrate the report. Company size and sector remain in the anonymized dataset; your email does not.

What the five levels look like

Every dimension in this assessment is scored 1 to 5. This is what the levels mean, dimension by dimension. The graded report diagnoses where your own answers land and what to do about it.

Proven on your data

  1. 1Take claims on trust
  2. 2Demo only
  3. 3Reference data test
  4. 4POC on our data
  5. 5Paid pilot, our data

At the low end: Taking claims on trust means buying a demo, not a capability. Before you sign, require a proof of concept on your own data; a vendor who refuses is answering the question for you. What good looks like: A paid pilot on your data is the strongest pre-purchase test there is. Define success metrics up front so the pilot gives a verdict, not just a good feeling.

Exit is written in

  1. 1No exit terms
  2. 2Long lock-in
  3. 3Standard notice
  4. 4Clear exit clause
  5. 5Exit and transition rights

At the low end: A contract with no exit terms is one you will not be able to leave when the tool disappoints. Negotiate the exit now, while you still hold the leverage a signature gives you. What good looks like: A clear exit with transition rights keeps the vendor honest for the life of the deal. Test the offboarding path on paper before you need it, not during a dispute.

You own your data

  1. 1Never checked
  2. 2Vendor-favourable
  3. 3Ambiguous terms
  4. 4Ownership retained
  5. 5Retained, reuse blocked

At the low end: Unchecked ownership terms often default to the vendor's favour. Read the training and reuse clauses before signing; ownership contested after the fact is ownership already lost. What good looks like: Retained ownership with reuse explicitly blocked is the terms to hold. Re-check the clause on every renewal; vendors update terms in their favour and count on you not reading them.

Lock-in is limited

  1. 1Deeply locked in
  2. 2Hard to move
  3. 3Some friction
  4. 4Portable with effort
  5. 5Cleanly portable

At the low end: Being deeply locked in means the switching cost, not the tool's quality, decides whether you stay. Before adding another, map what already traps you and start reducing it. What good looks like: Clean portability is leverage you keep for the whole relationship. Preserve it deliberately: every proprietary workflow you adopt trades a little of it away.

References were real

  1. 1No references checked
  2. 2Vendor-supplied only
  3. 3Checked their list
  4. 4Found our own
  5. 5Spoke to churned users

At the low end: Checking no references, or only the vendor's, means hearing only the story they curated. Find customers they did not choose: your size, your industry, ideally ones who left. What good looks like: Speaking to churned and independently-found users is diligence most buyers skip. Ask them what they wish they had known; that is the intelligence the pitch is built to hide.