See how your customers use their agents with you.
We talk to your customers, collect the prompts they give their agents, and run those exact prompts on the major personal agents. You get one report: what they ask, what works, where it breaks, and what to change first.
You don't know what your customers ask their agents to do with you.
You're operating in the blind. How your customers use your product through their agents is a black box. No log shows what a customer typed into Muse or ChatGPT before the agent reached you.
Teams test with tasks they write themselves. Real customers write short prompts that leave out details, on agents your team has never tried.
Your customers ask in ways your team has never tested.
User research, re-thought for a person and their agent.
A user research study
The person at the screen
- 00:20Found Manage booking
- 01:45Missed the change fee inside a pop-up
- 03:10Changed the flight herself
“I didn't see the fee until the very end.”
You see the person. Their agent never shows up.
You learn
How people use your screens.
What they ask their agents stays out of sight.
A user and their agent's study
The person, their agent and your business
Maya · her own words
move my flight to friday, same time
Her agent · Muse
- Opened the booking
- Picked Friday at 6:05 am
- Dropped seat 14A
- About to confirm, without saying so
Meant
Same time, keep my seat
Got
6:05 am, seat dropped
“I stopped it at the last step.” Maya
You learn
What customers ask their agents, and what they get.
Where it breaks, who caused it, and what to fix first.
What's inside.
The study ends in one report. Here is a sample, for a made-up airline. Pick a chapter.
Summary
Five things to know
- 1
77 agent users joined the study. Most send their agent to change a flight.
- 2
Their own prompts finished right in 71 of 144 runs. Your team's wording of the same jobs finished in 112.
- 3
The biggest blocker is finding the booking: 49 of 77 never give the agent their booking code.
- 4
Your bot check stopped Muse and Dots at the payment page in every run that got that far.
- 5
Most people let agents search and pick a flight. Only 12 of 77 let it confirm alone.
Start with one question.
Tell us what you want to learn about your customers and their agents. We plan the study around it, with you.
For example
- “Which agents do our customers use?”
- “What do they send their agents to do with us?”
- “Can those agents finish the job on our site?”
- “Where do people step in and take over?”
- “What should we fix first?”
- “Did our last change help?”
Can't email your customers? We find agent users in your category instead.
Questions.
How do you find customers who use agents?
A short screening survey goes out from your own address, and only customers who use agents go on to the study. We also invite customers your systems flagged as coming through an agent.
How long does a study take?
Five weeks from the day the survey goes out. Setup and sign-off usually take one to three weeks before that.
What do you need from us?
One survey email, an invite to agent-flagged customers, sign-off on our consent form and data agreement, a few live bookings you can reset, and read access to your logs. About 6 to 10 hours of your team's time.
Do you use the real agents?
Yes. Every result in the report comes from the real agents, run by hand, or from your customers' own agents in live sessions.
What happens to our customers' data?
You stay in control of it, and we use it for your study only. People share only their chats about you. Names, booking codes and card details are removed before anything reaches your team. Originals are deleted 90 days after the readout.
Can we repeat it?
Yes. A repeat study shows what changed since the last one.
Do I need your other products?
No. It works on its own.