The AX Benchmark Step 1 of 7
The auxfirst AX Benchmark

How far does your agency actually trust its AI agents?

Take the 8-minute benchmark and get your agency's Agent Trust Index — a confidential scorecard across the five dimensions of trustworthy agents, with how you compare to the rest of the industry.

No email, no account, no sign-up. Your answers stay in this browser and your score is computed here — nothing is stored and nothing is sent back to us except the anonymous score itself. Eight minutes.

What is 'agent trust' and why measure it? →
Time~8–10 min
Format21 questions · 5 dimensions
OutputIndex 0–100 + scorecard
PrivacyAnonymous · GDPR
Screen 01 · About you

First, a little context.

So we can benchmark you against the right peer group.

Q1.1
Which best describes your organisation?
Q1.2
Your role / seniority
Q1.3
Your function
Q1.4
Agency / team size
Q1.5
Primary region of operation
Please answer all five questions to continue.
Screen 02 · Adoption baseline

Where are AI agents actually used?

Systems that take actions, not just generate drafts.

Q2.1 · Select all that apply
Where are AI agents in use today?
Q2.2
Your agency's overall stage with AI agents?
Please answer both questions to continue.
Dimension 01 · Authority & Autonomy

How much are your agents allowed to do on their own?

Q3.1
When an agent can take an action (adjust a bid, shift budget, push creative), what typically happens?
Q3.2
Do you have written, role-specific limits on what an agent may do with a client budget?
Q3.3
How confident are you that your agents stay within their intended scope?
Please answer all three questions to continue.
Dimension 02 · Oversight & Escalation

When something is high-stakes or unusual, what happens?

Q4.1
For high-stakes agent actions, which is closest to your setup?
Q4.2
How does an agent behave when it hits something it's unsure about?
Q4.3
How confident are you a human would catch a bad agent decision before it affects a client?
Please answer all three questions to continue.
Dimension 03 · Transparency & Traceability

If a client asked 'what did your AI do, and why?' — could you answer?

Q5.1
Can you produce a record of what an agent did and why?
Q5.2
Are clients told when AI agents are involved in work delivered to them?
Q5.3
How confident are you that you could explain an agent's decision to a sceptical client?
Please answer all three questions to continue.
Dimension 04 · Safety & Failure Design

What stops a small agent error from becoming a big client problem?

Q6.1 · Select all that apply
Which safeguards are actually in place for your agents?
Score = number selected, capped at 4. "None" clears the rest.
Q6.2
Has an agent ever done something you didn't intend, with real consequences?
Q6.3
How prepared do you feel for an agent-caused incident in a regulated category?
Please answer all three questions to continue.
Dimension 05 · Accountability & Ownership

When an agent acts, who owns the outcome?

Q7.1
Is there a named person or team accountable for how your agents behave?
Q7.2
Does your client contracting / SOWs address what happens if an agent causes harm?
Please answer the two required questions to continue.
Your Agent Trust Index
out of 100

Five-dimension breakdown

Questions about your score? Ask us.

The number is the easy part. If you want to know what it means for your agency — which gap to close first, what it costs, whether the score is even the right question — write to us. We read every one, and a reply is not a sales call.

Contact us →

Prefer to read first? The AI agent operating model covers ownership, mandate and shutdown; the Action Heat Ladder is how to get dimensions one and two of your score right.

Edition 01 · The auxfirst AX Benchmark
Your answers stayed in this browser. Only the anonymous score is counted, so screenshot it if you want to keep it.