Agent Journey Lab

Start a $199 pilot

Agent Journey Lab · Journey assurance

Your next buyer may be an AI agent.

AI agents now try to buy, book, and request service on the public web. When an agent stops, your analytics do not show the stop. We test one revenue journey with named agents, and we give you a dated record of the exact stop.

Public research stops before form entry. A deeper test needs a written scope that you approve.

Start a $199 pilotSee an example record

Our test sessions declare AgentJourneyLab/0.1. Site owners can stop a public test: read the research boundary or write to research@agentjourneylab.com.

Run record

AJL-FIXTURE-CAL-002

Internal fixture record — not a customer result.

Journey
ambiguous_label_records_friction
Configuration
Primary — deterministic fixture script
Runner
runner-v1.0.0
Build
calibration-runner-v1
Target
Local fixture site on a loopback address
Operator
  1. 01Entry page✓ reached
  2. 02Renew a plan✗ stopped
  3. 03Confirmation page· not reached

✗ stopped at checkpoint 2 of 3, and the control journey passed in the same batch.

One record from the calibration fixture in our own repository. It shows the field shape that every record carries. It is not a customer result.

  • Declared journey

    You give the goal and the completion criterion before the run.

  • Named agents

    Each attempt names the agent family that made it.

  • Dated record

    Every observation carries the date of the run that produced it.

The problem

Agents now take your customer journeys.

A five checkpoint journey that stops at checkpoint fourCheckpoints one to three are reached. Checkpoint four is the stop. Checkpoint five is not reached.EntryPlanOptionsRenewConfirmSTOPReachedNot reached
The shape of one recorded run: the checkpoints that the agent reached, the checkpoint where it stopped, and the checkpoints after the stop.

AI agents open your pages, compare your offers, and try to finish your forms. A scan reads your markup. A scan does not make an attempt. A run shows where the work stops. The record from a run tells your team which checkpoint the agent reached, and which one it did not.

The same stop, seen two waysWithout a run, the stop leaves one unnamed exit in your analytics. With a run, the stop is a named checkpoint on a dated record.WITHOUT A RUNAgent arrivesSession endsNo named causeWITH A RUNDeclared journeyStop at step 4Dated record
A stop that ends a session leaves no named cause. A stop inside a declared run is a named checkpoint on a dated record.

What you have today

  • An exit, with no checkpoint name beside it.
  • No record of what the agent read at the stop.
  • No way for the team that owns the fix to repeat the stop.

What a run gives you

  • The checkpoint that the agent reached, and the one that it did not.
  • The observation at the stop, in one dated document.
  • A self-check that your team can repeat in less than one minute.

See the control rule and the citations

Method

Declare, attempt, record, reproduce, watch.

We use these five states for every run. The order makes each stop checkable by a person who was not there.

  1. 01

    Declare

    You give the goal, the checkpoints, and the completion criterion.

  2. 02

    Attempt

    A named agent family runs the declared journey.

  3. 03

    Record

    Each checkpoint is marked reached or stopped, with the observation.

  4. 04

    Reproduce

    We repeat the run beside a control journey.

  5. 05

    Watch

    One run each week for one month, and each record carries its date.

Read the full method

Journeys

Select one journey for the pilot.

Each journey page states the declared goal, the checkpoints, the stop rule, and the teams that usually hold the fix.

Record

What you receive.

You get one dated document for each run. Give it to the team that owns the fix. These are the pages as they stand today.

A screenshot of the example record page on this site. It shows the header field table of one run: record ID AJL-FIXTURE-CAL-002, the journey ID, the goal, the target, the attempt count, the configuration, the runner and build identity, the environment fields, and the control journey result.
The header of a finished record, photographed from our own example page. Every value in it comes from the calibration fixture in this repository, so it shows the field shape of a record and no customer work.
A screenshot of the method page on this site. It shows the five run states in order — Declare, Attempt, Record, Reproduce, Watch — each with the one sentence that states what goes in and what is written down.
The five states as the method page states them: what each state takes in, and what each state writes down.
Record fields
GoalThe declared journey, in your words.
Named agentThe agent family that made the attempt.
TimestampThe start of the run.
CheckpointsEach checkpoint and its state, reached or stopped.
Final stateThe place where the run stopped, and the observation that decided it.
Likely fix ownerThe team that holds the change: content, front end, platform, or policy.
Self-checkThe steps that let your team repeat the stop in less than one minute.
Change since last runWhat moved since the last dated record for the same journey.

See an example record

Price

The fee schedule.

Founding Journey Pilot

$199

one time, one site, one journey

  • Three baseline attempts with the primary configuration, and one with the secondary configuration.
  • One dated evidence record with a self-check.
  • One month of weekly monitoring.
  • Up to two retests after you fix an issue.

Start a $199 pilot

Journey Monitoring

$199

per site, each month, optional

  • The same declared journey each week.
  • One dated record for each run.
  • A statement of what changed since the last record.
  • Stop it by email at the end of the paid month.

See the monitoring terms

Multi-site Program

Scoped

after a pilot, per agreement

  • Many sites or locations with one shared journey model.
  • One authorization document for the group.
  • A record for each site and a summary of the differences.
  • Agreed data handling, retention, and deletion.

Read about a program

How you buy

  1. You send the detailsEmail the site and the journey that you want us to test.
  2. We send a written scopeWe reply with the exact scope, the boundary, and payment instructions.
  3. You approve and payYou confirm the scope in writing, you confirm your authority for the domain, and you pay $199.
  4. Testing startsWe run the baseline attempts and we send the dated record.

No checkout exists on this site. We send payment instructions with the written scope.

Read the pilot scope

See all three offers

Read about a multi-site program

Limits

The boundaries that we keep.

  • Public research

    We read public pages only. We stop before we enter data in a form.

  • Authorized test

    We go deeper only inside the written scope that you approve. You state the limits.

  • Never without approval

    We do not submit forms, use accounts, make payments, hold bookings, or file claims outside the scope.

We keep your results private. We do not publish them without your written permission.

To exclude your domain from public research, email the domain to us. We confirm in writing.

Questions

Common questions

What do I get for $199?

You get three baseline attempts with the primary configuration and one attempt with the secondary configuration. You also get one dated evidence record, four dated weekly monitoring records across one month, and up to two retests after a fix.

Is this a software product?

No. This is an operator-led service. We do the work and we send you the record by email. There is no dashboard and no account.

Do you need access to my systems?

No. We use the public journey. If the scope needs a test account, you supply it and you set the limits in writing.

How do I pay?

We send payment instructions with the written scope. You pay before the test starts. This site holds no payment page and no form.

Do you publish my results?

No. Results stay private. We ask for written permission before we use any part of your results in public.

Can I stop the monthly renewal?

Yes. The renewal is optional. Tell us by email and the monitoring stops at the end of the paid month.

Read the full question list

Next step

Test the journey before the agent takes it.

Send the site and the journey. We reply with a written scope, the boundary, and payment instructions.

Start a $199 pilotRead the pilot scope