CX Vision 2035 · Wembley · AI Perception Score™

Companies think their AI is working. Their customers don't.

We asked 1,984 US customers of 38 brands about their last conversation with the company's AI agent: would you recommend it, do you trust it, and what should it fix first. Here is what they said.

Book a free 30-minute call Get the free report

Both are free. The report arrives in your inbox within a minute. On the call, Robert looks at your own AI agent first; there are 20 slots before December 1.

The advocacy gap

Customers who would recommend the brand would not recommend its AI agent.

Verizon
−28
its AI agent
+14the brand, same customers
Chase
−10
its AI agent
+14the brand, same customers
Walmart
−24
its AI agent
+24the brand, same customers

Net score, −100 to +100: customers who would recommend, minus those who would not. Each brand scored by its own customers. Chase on a smaller sample, read as directional.

Key finding 1

Customers love the brand. They don't love its AI agent.

Across all four industries, the same 1,984 customers rated the brand they use, then the AI agent that served them. 38 points apart on recommendation, 18 on trust, in the same customer's head.

The brand
Its AI agent
Gap
Would recommend
+29
−9
38points
Trust
+6
−12
18points

Net score, −100 to +100 · 1,984 US customers · 38 brands in telecoms, banking, retail and payments

Key finding 2

Your AI agent meets your best customers. Six in ten wanted a person.

46% of the customers who used a brand's AI agent already recommend the brand, against 38% of customers who had no AI conversation. The people you can least afford to lose are the ones it serves.

They tell you what to fix, in this order. Ten things to choose from; the first one is the same in every industry.

6 in 10
wanted a person at some point in the conversation
7 in 10 in telecoms
#1
A human within reach when I need one
49%
#2
Answers that are right the first time
35%
#3
Understand what I actually mean
34%

What AI agent users ticked, out of ten things to fix. Share of 1,984.

Key finding 3

Your dashboard counts a win. Your customer counts a loss.

Dashboards count finished conversations. Customers judge the AI agent by whether they got what they came for.

Share of conversations
AIPS™
Would use it again
It helped
46%
+19
80%
Helped partly
37%
−18
44%
Did not help
13%
−64
9%

1,984 AI agent conversations · AIPS™, the AI Perception Score™, −100 to +100 · 4% of conversations were still pending when they ended

Key finding 4

Tell them it's AI. It costs nothing, and it changes the verdict.

Customers told "this is AI" would recommend the AI agent at +10. Told "automated" or "virtual assistant", −9. Told nothing, −30. With the brand and the outcome held constant, the honest word is still worth 22 points. In the EU it has been the law since 2 August 2026.

Would recommend the AI agent, by what the customer was told

"This is AI"
+10
"Automated"
−9
Told nothing
−30
−40−200+20+40

Net score, −100 to +100, by what the customer says the assistant told them. With the brand, the outcome, the hand-over and the journey held constant, "AI" against nothing is +22 points.

Key finding 5

No industry scores above zero. Telecoms is last.

−14
Telecoms
−5
Banking
−5
Retail
0
Payments

AIPS™, the AI Perception Score™: the three customer questions combined into one number, −100 to +100. 1,984 US customers.

The free report covers your industry. The score, the brands that sit above and below it, and the three things customers want fixed first. Eight pages, in your inbox within a minute.

Get the free report
What the report does for you

One number tells you, journey by journey, whether to scale, fix first, or keep a person.

Every journey is read the same way: would customers use the AI agent again, and did they want a person. US telecoms, as an example.

Scale
Choosing a plan or an upgrade
63% would use the AI agent again
Fix first
Bills, charges and refunds
46% would use it again · 74% wanted a person
Keep a person
Cancellation and switching
37% would use it again · 81% wanted a person
Small sample

The same read for your industry is in the free report. For your own journeys, it is what a baseline gives you.

Who ran this
Robert Hackl

Robert Hackl spent 30 years running the companies that answer your calls. Now he measures what AI agents do to their customers.

McKinsey10 years, junior partner
Tele2 AustriaCEO
Vodafone GermanyCOO, 11,000 frontline telesales and service, and Digital Transformation
T-Mobile USSVP Channel Management, member of the SLT
SprintSVP GM Handset Leasing and Insurance, member of the SLT
Awards InternationalChair judge for AI
$50B+of shareholder value created by Robert's teams
Two ways to start

What do your customers think of yours?

Start with the industry picture, or have Robert look at your own AI agent. Both are free.

Option 1Free · instant

The free industry report

The numbers for your industry in eight pages, or all four industries in one report.

  • Your industry's score, against the other three
  • The brands in your industry, and where their customers put them
  • The three things customers want fixed first

Name, work email and industry. In your inbox within a minute. Nothing else follows unless you book a call.

Get the free report
Option 2 · recommendedFree · personal to you

A 30-minute call with Robert

If a chatbot or AI agent already talks to your customers, or will soon.

  • A first outside-in look at your AI agents before the call
  • Your starting position against your industry, and the trust gaps that keep AI agents from scaling
  • A practical discussion of the status, challenges and opportunities of AI in your industry
  • How to do a baseline the best way, guided by your priorities

20 calls before December 1. We respond personally, within 24 hours. You also get all the reports you would like to benchmark with.

Book the free call

Questions

Can't we just survey our own customers?

You can, and we encourage it. What you cannot build alone is the benchmark: a carefully curated set of the same questions, asked the same way, to customers of 38 brands, and growing every month. And the other half of the picture: senior judgment, and an outside-in read of the judgment embedded in your agents, which in our experience is where the biggest root cause usually lies.

Can we get our own score?

Yes. A first survey of your own customers with the three questions gives you a number next to the benchmark. We recommend one step more: a fuller questionnaire that captures the dynamics between your AI agents and your people and the resolution rates by channel and journey, plus an evaluation of how mature the judgment embedded in your AI journeys is and how your agents' guardrails are set. A baseline cannot be failed; it sets the line.

What happens on the 30-minute call?

Before the call, Robert takes a first outside-in look at your agents. On the call you discuss your starting position, the status, challenges and opportunities of AI in your industry, and how to do a baseline evaluation the best way: guided by your priorities, informed by our experience of the trust gaps that keep AI agents from scaling.

What does it cost?

The report and the call are free. A baseline measurement is about $10,000 with your own customers, or about $15,000 through a public panel that covers your brand. The judgment-maturity evaluation is interview-based and includes an exercise in breaking your chatbot the way a consumer would; it is a separate, small project. Once the baseline stands, we can talk about an operating rhythm: continuous measurement, not by survey but as a dynamic sidekick to your agents.

Will our number be public?

No. Your number stays private, and you decide who sees it. With your permission, we fold your data into the industry benchmark, without any personal identifiers of respondents, purely to make the benchmark richer for everyone.

We are replacing our AI agents next quarter. Should we wait?

That is the best time to measure. Today's AI agent becomes your before, the new one your after, and you can show the difference to your leadership or board.

"We don't know what our AI agents do to our customers" stops being an acceptable answer.

Book a free call
Book a free call Get the report