World Model Readiness
Engraved functions instrument

For your company

AI in the Functions

Module · What customers learn changes what they feel

The Customer AI Trust Check

Customers increasingly assume AI is somewhere in the room; the question is whether you tell them or let them find out. Trust is cheap to keep and expensive to rebuild, and the way you handle disclosure, consent and recourse decides which way it moves. This module checks the five practices that hold customer trust as AI spreads through the relationship: honest disclosure, real consent for AI processing, a complaint channel that works, a right to a human, and some way of knowing whether trust is holding at all.

Question 1 of 5 · Disclosure is honest

Do customers know when AI is involved in what they receive from you?

AI writing the email, scoring the application, or answering the chat is invisible unless you say so. Customers who discover undisclosed AI feel handled, not served, and the trust cost lands all at once. Honest disclosure sets expectations and, increasingly, meets a legal duty.

Question 2 of 5 · Consent is real

Do customers meaningfully consent to AI processing their data?

Feeding customer data into AI systems for personalisation, scoring or training is a use they may not have agreed to. Consent buried in a privacy policy nobody reads is compliance theatre. Real consent is specific, informed and revocable, and it is the difference between a partner and a subject.

Question 3 of 5 · Complaints have a channel

If a customer is unhappy with an AI-driven outcome, where do they go?

An AI decision a customer cannot contest is a wall, not a service. Whether it is a declined application, a wrong recommendation or a bad automated reply, there needs to be a named, reachable channel to raise it, and a person on the other end who can actually change the outcome.

Question 4 of 5 · Human fallback exists

Can a customer get a human decision when the AI one is not good enough?

For consequential outcomes, being able to reach a person is not a courtesy, it is what keeps automation legitimate. A customer stuck with an AI verdict and no human recourse experiences your efficiency as their powerlessness, and remembers it that way.

Question 5 of 5 · Trust is measured

Do you actually know whether AI is helping or eroding customer trust?

Trust erodes silently: no complaint, just a slow drift in loyalty, referrals and tolerance. If you are not measuring how AI in the relationship affects what customers feel, you will learn about the damage from the churn report, long after you could have fixed it.

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.

Do you tell customers when AI is involved in what they receive?

Always, proactively
For some interactions
Only if they ask
No
We are not sure

Can customers reach a human when an AI outcome is not good enough?

Yes, guaranteed
On escalation
Difficult in practice
No
AI makes no such decisions

Do you measure how AI affects customer trust?

Yes, and we act on it
Occasionally
No
We assume it is fine
Too early to tell

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.

Disclosure is honest

  1. 1Deliberately hidden
  2. 2Never mentioned
  3. 3Disclosed if asked
  4. 4Proactively disclosed
  5. 5Clear and consistent

At the low end: Hiding AI involvement is a bet that customers never find out, and it is a bet you lose loudly. Decide where AI touches the customer and disclose it plainly before someone else reveals it for you. What good looks like: Clear, consistent disclosure across touchpoints is what lets customers trust the parts that are automated. Keep it current as AI spreads into new interactions; yesterday's disclosure does not cover today's new use.

Consent is real

  1. 1No consent sought
  2. 2Buried in the fine print
  3. 3Blanket opt-in
  4. 4Specific and informed
  5. 5Specific and revocable

At the low end: Processing customer data through AI without consent is a breach of trust and, in many places, of law. Map where customer data feeds AI and get specific, informed consent before it does. What good looks like: Specific, revocable consent treats customers as partners in how their data is used. Keep the revocation genuinely easy; consent you cannot withdraw is not consent, it is a formality.

Complaints have a channel

  1. 1Nowhere to complain
  2. 2Generic inbox
  3. 3Channel exists, unknown
  4. 4Known and reachable
  5. 5Known, reachable, closes loop

At the low end: An AI outcome customers cannot challenge tells them their case does not matter. Publish one clear channel for contesting AI-driven decisions, with a human empowered to act on it. What good looks like: A known channel that closes the loop turns a complaint into retained trust and a system improvement. Track how those complaints resolve; the pattern tells you where the AI is quietly failing people.

Human fallback exists

  1. 1No human option
  2. 2Human in theory
  3. 3Available on escalation
  4. 4Clear human fallback
  5. 5Guaranteed for key decisions

At the low end: No route to a human on a decision that matters is efficiency that reads as contempt. Guarantee a human fallback on consequential outcomes before customers learn to expect a wall. What good looks like: A guaranteed human fallback on key decisions is what keeps automation feeling like service rather than sentence. Keep it adequately staffed; a fallback nobody can reach in time is a promise you are breaking quietly.

Trust is measured

  1. 1Never measured
  2. 2Assumed fine
  3. 3Occasional pulse
  4. 4Tracked deliberately
  5. 5Tracked and acted on

At the low end: Trust you never measure is trust you find out about only when it is gone. Add one or two questions about the AI experience to a survey you already run; the signal is worth the two lines. What good looks like: Trust measured and acted on turns AI from a reputational gamble into a managed relationship. Keep the loop live; the technology and customer expectations both move faster than an annual survey can catch.