Vendor diligence

About to hire an AI vendor? Vet them in 5 minutes.

Twelve questions turn vendor-pitch answers into a verdict, a 7-category scorecard, top concerns, customised questions to ask, and a contract clauses cheat sheet. Free, methodology open and auditable.

No login for the free preview. Work email reveals the full scorecard + PDF.

Why it matters

AI procurement fails differently than SaaS procurement.

The standard vendor checklist - SOC 2, references, financial stability - misses the patterns that actually end AI engagements badly. Token cost trap, runtime model swaps, “proprietary” wrappers, and IP that is “unclear” before signature are 2026-specific failure modes that demand a 2026-specific scorecard.

Cost trap

Per-token pricing without a cap

The single most expensive 2026 procurement antipattern. Buyer has zero leverage once usage scales. Caps + quarterly review are non-negotiable.

Model risk

"Proprietary AI" at premium price

Often a thin wrapper around GPT-4 or Claude, marked up. Disclosure clauses + benchmark vs. the underlying API are how you tell.

IP risk

IP that is "unclear" before signature

Reliably becomes a dispute later. Cost of resolving it after launch usually exceeds the original deal value.

Red flag taxonomy

25 patterns we have seen end engagements badly.

Every flag in the rule library is mapped to one of seven scorecard categories. The wizard surfaces only the flags that actually triggered for your specific pitch.

  • 4 flags

    Technical credibility

    Vague on the model, "proprietary" claims at premium price, runtime model swaps without a switching policy, "our AI" hand-waving.

  • 3 flags

    Pricing transparency

    All-inclusive pricing too good to be true, ongoing pricing not discussed, undisclosed cost assumptions.

  • 6 flags

    Customer proof

    No working example, slides-only pitch, generic demo, LinkedIn-only references, anonymized case studies, six-figure budget with zero references.

  • 3 flags

    IP protection

    Vendor owns the model and outputs, IP "depends on lawyers", no clause for derivative works.

  • 1 flag

    Cost predictability

    Per-token pass-through with no monthly cap - the single most expensive 2026 procurement antipattern.

  • 5 flags

    Operational competence

    Open-ended timeline, large budget without milestones, no named delivery lead, vague data-handling answer.

  • 3 flags

    Relationship trust

    Sales-led pitch with no proof, looks like a thin wrapper, freelancer at six-figure budget, staff aug without named engineers.

How we score

From pitch deck to verdict in 5 minutes.

  1. 01

    Twelve questions

    Answer how the vendor pitched. We adapt the wording for consultancies, agencies, SaaS vendors, freelancers, and staff-aug shops.

  2. 02

    Rule engine evaluates

    Your answers are cross-referenced against 25 procurement-failure patterns and rolled up into a 7-category trust score. Same answers, same verdict, every time.

  3. 03

    You get a verdict + PDF

    Verdict, customised questions to ask, contract clauses to negotiate, a 12-question reference-call template, and a board-ready PDF.

7 risk categories

The full scorecard surface area.

  • Technical credibility

    Did they name a specific model? Was the data-flow concrete or hand-waved?

  • Pricing transparency

    Are ongoing costs itemised, all-inclusive, or "your responsibility"?

  • Customer proof

    Live demo? Anonymised case study? Or just a slide deck?

  • IP protection

    Who owns the model, the code, the outputs, the fine-tunes?

  • Cost predictability

    Token consumption pricing without a monthly cap is a CFO trap.

  • Operational competence

    Milestone payments? Named delivery lead? Concrete timeline?

  • Relationship trust

    Sales-led pitch with no proof and an open-ended timeline = walk away.

Who this is for

Built for the buyer, not the vendor.

CFOs sanity-checking the contract. Heads of procurement asking the right questions on the second call. Founders deciding whether to write a six-figure check.

  • CFO

    Cost predictability, contract clauses, walk-away signals.

  • Head of Procurement

    Customised questions, reference-call template, vendor scorecard.

  • Founder / CEO

    Quick verdict on whether to keep evaluating or move on.

Pillar topics

Ten deep-dives on the topics that matter.

Each pillar is a self-contained reference for one part of the scorecard. Use them on the second vendor call - the questions and clauses are linked back to the rule library.

Methodology

Transparent, deterministic, refreshed quarterly.

Every flag pattern, every clause, and the verdict cascade is published. buzzi.ai scores green on its own methodology - we run the self-pass test in CI. Read the methodology to see exactly how a verdict is computed from your 12 answers.

Read the methodology

Frequently asked

Questions buyers ask.

What is the 12-Point AI Vendor Trust Score?

A 12-question evaluation that turns vendor-pitch answers into a green / yellow / red verdict, a 7-category scorecard, top concerns, customised questions to ask, and a contract-clause cheat sheet. Methodology is open and auditable.

How accurate is the verdict?

The engine is rule-based: 25 declarative flag patterns derived from procurement-failure case studies. There is no LLM scoring loop, so the same answers always produce the same verdict. Re-run the wizard whenever your vendor changes their pitch.

Who is this for?

CFOs, COOs, heads of procurement, and founders evaluating their first or fifth AI vendor. The scorecard is most useful for engagements above $50K, but the methodology applies at any budget.

Does the tool work for SaaS vendors and consultancies?

Yes. The wizard adapts question wording for six vendor archetypes: boutique consultancy, agency, SaaS vendor, freelancer, staff-augmentation, and mixed engagements.

Is the free preview enough?

The preview gives you the verdict and your top 2 concerns. The full scorecard, customised questions, contract clauses, reference-call template, and PDF require a work email.

How long does the scorecard take?

Five minutes for the 12 questions. The free preview is instant; the full scorecard returns within seconds of submitting.

Where does buzzi.ai stand on this scorecard?

buzzi.ai answers each of the 12 questions in writing before any engagement. We score green on our own methodology - we run the self-pass test in CI.

Can I use the contract clauses with my own counsel?

Yes. The clause library ships with US-default jurisdiction copy. Have your counsel adapt them to your jurisdiction before signing.

What if my vendor refuses to amend the contract?

Three or more red flags plus refusal to negotiate is the walk-away signal. The PDF report includes a "when to walk away" page with the specific patterns we have seen end badly.

Can buzzi help me run a vendor evaluation?

Yes - book a 30-min call from the results page. Free, no pitch, just a sanity-check on the contract before you sign.

Does the methodology refresh?

Quarterly. New procurement-failure patterns get added; outdated flags get retired. The methodology page shows the version + change log.

Is my data shared with vendors?

No. Submissions are private to your account. Vendor names and emails never leave your row.

Want a second opinion?

Run the proposal past buzzi - free 30 min.

No pitch. Senior delivery consultants will sanity-check the contract, flag the three things to negotiate hardest, and tell you when to walk away.

Book the call