Practical AI Advisory
Signal by FMD

Even when AI output looks finished, teams still need a clear way to judge the right level of review.

FMD Signal helps teams decide how much review AI-assisted work actually needs.

AI-assisted output can look finished before it has been properly checked.

That creates a common problem: people may trust it too quickly, and teams are often unclear on what level of review is actually required.

Signal helps make that clearer. It gives teams a more practical way to match review effort to risk, so checking is more consistent and more proportionate.

See the example View the Signal model
Clearer review decisions Proportionate governance

Starts with

AI influence

Checks

Data and reliance

Outputs

Review expectation

A

AI influence

How much did AI shape the output?

From light drafting help to work that shaped meaning, options, or recommendations, start by judging the level of AI influence.

Input

D

Data sensitivity

What kind of information was involved?

Personal, confidential, sensitive, or restricted information should raise the level of care.

Context

U

Use

What will the output be used for?

A rough draft, an internal summary, and decision support may look similar, but they do not carry the same review expectation.

Reliance

S

Signal

What review expectation is returned?

Signal brings those earlier judgements together and returns the review expectation: what level of checking, ownership, and evidence is needed.

Output

How Signal Works

See how a team would use Signal in practice.

Move through the judgements a team would make before relying on AI-assisted work.

Step 1

How much did AI shape the output?

From light drafting help to work that shaped meaning, options, or recommendations, start by judging the level of AI influence.

Step 2

What kind of information was involved?

Personal, confidential, sensitive, or restricted information should raise the level of care.

Step 3

What will this output be used for?

A rough draft, an internal summary, and decision support may look similar, but they do not carry the same review expectation.

Step 4

What review expectation does Signal return?

Signal brings those earlier judgements together and returns the review expectation: what level of checking, ownership, and evidence is needed.

The Problem

AI output often looks more dependable than it is.

That polished finish changes behaviour. People trust outputs too quickly, review varies from person to person, and similar cases get handled differently.

01

Finished-looking output attracts over-trust.

Fluent language and tidy formatting can make weak work feel ready before it deserves that confidence.

02

Review becomes inconsistent.

One person treats it as a rough draft. Another treats it as decision support. The control response drifts.

03

Teams over-check some work and under-check other work.

The result is unnecessary friction in some places and weak assurance where it matters more.

The Idea

Match review effort to risk instead of treating every AI use the same.

Signal makes the judgement clearer. It gives teams enough structure to decide what needs light review, what needs stronger checking, and why.

Signal helps teams ask

  • How much did AI shape this output?
  • What kind of information was involved?
  • What will this output be used for?
  • What level of review should that trigger?

Working rule

More AI influence, more sensitive information, and heavier reliance should lead to stronger review.

Four Dimensions

The four judgements behind the Signal output.

A

AI influence

How much did AI shape the output?

D

Data sensitivity

What kind of information was involved?

U

Use

What will this output be used for?

S

Signal

What review expectation is returned?

Simple Example

Not every AI-assisted output needs the same review response.

Worked example

Meeting summary for internal planning

AI helps draft a summary for an internal team update. The output supports planning, but it is not the final basis for a regulated or high-consequence decision.

AI role

Light drafting help

Data

Routine internal information

Use

Internal planning

Review

Proportionate review before sharing

The review is still real, but proportionate.

Signal keeps the public example focused on judgement and review.

The Outcome

What clearer review looks like in practice.

Clearer decisions

Teams can explain why one use is treated lightly and another gets stronger checking.

Proportionate governance

Review effort rises where reliance, sensitivity, or consequence rises.

Less over- and under-checking

The model reduces unnecessary friction in some places and weak assurance in others.