AI Trust: how PROOF uses AI, and where it stops

Demonstration authority path

AI can prepare the work. People own the decision.

Source retained

01 · Authorized inputBuyer email

“Security review is booked for 24 September.”

Sender, source and timing retained

02 · AI assistanceEvidence prepared

Extract the commercial fact
Map it to the Method
Show uncertainty and source

03 · Human authoritySeller reviews

Accept, correct or reject the proposed evidence before it can affect the commercial record.

Decision and accountable owner retained

Protected CRM fieldForecast category

AI can't write outside the configured authority and approval path

Demonstration deal · Northstar. Use AI to move faster without giving it the final say.

The working boundary

Use AI for speed without handing it the decision.

PROOF keeps four things visible: the source, the machine contribution, the human judgment and the action that followed.

AI helps with

Preparing work for review

Extract facts from authorized evidence inputs
Structure them against the customer's Method
Surface gaps, staleness and conflicting evidence
Prepare drafts and explanations with sources attached

People retain

Commercial authority

Confirm, correct or reject proposed evidence
Commit the forecast and accountable override
Approve consequential actions and protected writes
Challenge a conclusion using the underlying source

Not a score. A trace.

Never take an AI's word for your number.

Most AI tools ask you to trust a prediction. PROOF asks what the prediction is built on. That's the difference between a score and a trace.

01Evidence before prediction

02Source before summary

03Human judgment before automation

04Transparency before confidence

05Authority before action

When procurement asks

Answer the question without slowing the deal.

Show exactly how AI helped, who made the decision and whether anything was allowed to change.

What did AI do?Prepared the evidence

The source, extraction and uncertainty remain visible.

Who made the decision?A named person

The review, correction and accountable judgment remain attached.

Could it write back?Only with authority

Protected CRM fields follow the configured role and approval path.

Give Security, RevOps and Revenue Leadership the same answer without asking them to trust a black box.

Analyst prediction

Forrester predicts that untested generative AI functionality, combined with lagging AI user skills, will lead to incidents costing more than $10 billion in enterprise value through falling share prices, legal settlements and fines.

Forrester, 2026 B2B Marketing, Sales, and Product Predictions, 28 October 2025. This is an analyst prediction for 2026, not a survey finding. Read the source

A prediction isn't a measurement, and PROOF won't present it as one. It's worth reading anyway, because the mechanism is specific: the damage comes from AI output nobody could check before acting on it. PROOF is built so that every machine contribution names its source, its age and the human who accepted it, which is what makes an incident answerable afterwards rather than deniable.

See the commitments PROOF is willing to sign

What PROOF refuses to do

Things PROOF will never do.

Each refusal protects the commercial record and the people accountable for it.

Invent evidence or a source.

Turn missing information into confidence.

Hide human judgment behind an AI score.

Rewrite an earlier decision with later facts.

Rank Sellers through surveillance.

Change protected CRM fields without authority.

The standing commitments

Four things that do not change.

Whatever the model does, these hold. They are the reason the rest of this page can be trusted.

01Show uncertainty

Missing stays missing.

02Preserve judgment

People own the decision.

03Respect Sellers

No surveillance scores.

04Keep history

Corrections never erase.

Security · Privacy · Operations

Looking for the procurement position?

Review data residency, access, retention, sub-processors, incident response and the document library on the separate Trust page.

Data, security and privacy positions

That is what the model may do.

The next question is what happens to your data while it does it.

See Trust

Inspect the boundary

See AI assistance and human authority on the same decision.

See PROOF in action Four questions to ask first

PROOF provides clarity. You provide judgment.