How it works
From one line to a plan of tests you can start today.
Step 1
You give one line
You write one line about the agent you want to build.
It looks wider than your first idea.
You see places to build start to appear.
We reply within a day. Your first board is free.
Step 2
Where you could build
You read each place, with the honest case against it.
You take one. It starts the research from there.
SampleHandovers that close wrong between ops and finance
- Who is stuck
- Ops leads who close loads that finance invoice a week later.
- Money
- Teams already pay for close-out tools.
- The honest case against it
- The platform they already pay for could add this in one release.
Depot capacity nobody can see until Friday
- Who is stuck
- Planners who quote on Monday and learn the cost on Friday.
- Money
- Not yet. Nobody pays for this today.
- The honest case against it
- Depots run on habit. A new screen is one more thing to ignore.
Step 3
Who is stuck, in their own words
Inside the place you took, you see the struggles people voice.
Each one says who said it, or that nothing backs it yet.
You pick the one to pursue.
SampleFinance finds the wrong step a week late
“I keep my own sheet because the system is a week behind by the time finance looks at it.”
A finance lead, in a conversation Nobody owns the handover between two teams
Nothing backs this one yet.
Step 4
The agent worth building
You pick the agent to build first.
It comes as a blueprint with five plain parts.
You also see who is already there, and the gap each one leaves.
SampleWhat it remembers. Every handover, and who closed it.
How it decides. It flags a line when ops and finance disagree.
What it may touch. It reads the handover. It never changes it.
How you know it did well. Finance accepts the close the first time.
What happens when it fails. It holds the handover and asks a person.
Step 5
A plan of tests
You say what you want to build.
It writes the tests, the questions, who to ask and the order, riskiest first.
You see a plan you can start today, and the date you will know by. On that day, it checks in on how it went.
SampleThree tests, about two weeks. Riskiest first.
Will they hand over real handovers, and act on what comes back?
5 to 7 days
Does it name the right line, often enough?
2 to 3 days
Will they pay for a 45-day pilot?
5 to 7 days
On the day you named, it checks in.
Step 6
A test, with a bar you set first
Before you ask anyone, you write the result that would make you believe it.
You can try your questions on simulated people first, to catch the leading ones.
Nothing a simulated person says counts as evidence.
Then you ask real people and keep what they said, in their own words.
SampleWhat result would make you believe it is real?
Three of five forward a real handover in week one and act on the line we send back.
Set before the first conversation.
Rehearse your questions
Three simulated people answer first, and the leading questions are flagged.
Nothing a simulated person says counts as evidence.
Step 7
What happened
You say how it went: it worked, mixed, or it did not.
You say how sure you are.
Then you take one of four roads.
SampleHow did it go?
- It worked
- Mixed
- It did not
How sure are you?
- Very sure
- Quite sure
- Not so sure
- Not at all
Then take one road
- Test it again, properly
- Change one thing
- Right people, wrong idea
- Stop here
Step 8
What comes next
If it held, you hand the blueprint to whoever builds it.
If it did not, you change one thing, or stop here.
Either way, you checked before you built.
SampleIf it held
Hand the blueprint to a builder.
If it did not
Change one thing, or stop here.