Pilot-to-scale AI roadmap
Decide which AI pilots are ready to scale, which to fix first and which to stop, with dated roll-out waves.
Ready to use. Buy it in your account: a one-time licence fee and a monthly fee, plus the AI it uses.
What it does
List the AI pilots you have run, what they have shown, who could own them and the teams to roll out to. The AI checks each pilot against eight scale gates (owner, measure, results, data, process, training, a person checking, support), quoting your own words as evidence; anything you did not say is marked not known, never assumed. We decide each verdict: ready to scale when every gate is met, stop only when your brief shows it, otherwise fix first. You get fixes for each gap, success measures, roll-out waves with dates team by team, a monthly steering meeting, what to track and a message to staff.
Who it is for
- Businesses with AI pilots that never quite became everyday work
- Managers asked to roll out what worked
- Steering groups that need a clear plan and measures
Quick start
- 01List your pilotsA line each: the name, then how it is going, such as who checks it and since when.
- 02Add what they have shownMeasures, feedback and problems: the more you say, the fewer gates are left as not known.
- 03Run itIt takes a minute. You can leave the page: the result waits in your run history.
- 04Agree it at your steering meetingCheck the verdicts and owners, then start the first wave.
What you give it
- Your AI pilotsRequired, up to 10 lines
- Your AI pilots, a line each: the name before a colon, then how it is going.
- What they have shownOptional, up to 3,000 characters
- What the pilots have shown: measures, feedback and problems. Optional.
- Who could own themOptional, up to 15 lines
- People who could own the pilots, with their role. Optional.
- Teams to roll out toOptional, up to 12 lines
- Teams to roll out to, with the number of people. Optional.
- Start date (YYYY-MM-DD)Optional
- When the roll-out starts. Optional: the next Monday.
- AI tools you useOptional, up to 8 lines
- The AI tools your business already uses, so the plan names only those. Optional.
- ConstraintsOptional, up to 600 characters
- Limits such as budget or where data must stay. Optional.
- Anything elseOptional, up to 1,500 characters
- Anything else the plan should know. Optional.
What you get
- Your pilots
- Each pilot's verdict, risk, the eight gates with your words as evidence, a success measure, a suggested owner and fixes for the gaps.
- Roll-out waves
- Ready pilots from week 1 team by team, fix-first pilots from week 7 once their gaps close, and a review at week 12, all dated.
- Message to staff
- What is rolling out and why, that a person stays responsible, and how to give feedback.
- Steering meeting and measures
- A fixed monthly agenda and what to track: adoption, quality, time, problems and feedback.
You can edit the result, then copy it, print it or download it as Word, Markdown, plain text or data. Every run is kept in your run history until you delete it.
An example
Calder Freight is a made-up business. Below is what was entered, then the tool's real result, unedited, made on 14 November 2026. Subscribe to run it with your own details.
What was entered
From Calder Freight’s Business profile
- What you sell
- Road haulage and warehousing for manufacturers in West Yorkshire, about 120 staff
- Your AI pilots
- Invoice reading: running in accounts since August; Tom Reid owns it and checks every invoice before it is posted; it reads only our own supplier invoices, which our data policy allows; the process is written in the accounts handbook and the three accounts staff were trained in July
- Customer email drafts: tried by two people in customer service for three weeks; they like it but nobody has measured anything yet
- Route planning suggestions: tested in October, drivers found the routes ignored low bridges, so nobody uses it now
- What they have shown
Invoice reading: since August every supplier invoice has gone through it, with corrections on few of them; target agreed in July was fewer than one correction in twenty invoices, and it has met it each month. If it goes wrong, Tom Reid switches back to manual entry and tells the supplier desk.
- Who could own them
- Tom Reid, finance manager
- Aisha Patel, customer service lead
- Gary Holt, transport manager
- Teams to roll out to
- Accounts (3)
- Customer service (8)
- Warehouse office (5)
- Start date (YYYY-MM-DD)
2026-11-30
- AI tools you use
- Microsoft Copilot
2 optional details were left empty.
In your account this is a form you fill in, and the tool runs once you subscribe.
What the tool returned
Pilot-to-scale AI roadmap
In short
Invoice reading looks ready to become everyday work in accounts, with an owner, an agreed target that it has met, written steps, trained staff, checks and a fallback in place. Customer email drafts are promising but lack a short enough trial, an agreed measure and most other basics. Route planning suggestions should stop because drivers found its routes unsafe and nobody uses it.
Your pilots
Pilot 1
Pilot: Invoice reading
Verdict: Ready to scale
Risk: High
Scale gates
Gate 1
Gate: Owner: A named person is responsible for it
Status: Met
Evidence: Tom Reid owns it
Gate 2
Gate: Measure: Success is measured with a number agreed in advance
Status: Met
Evidence: target agreed in July was fewer than one correction in twenty invoices
Gate 3
Gate: Results: It has results from at least four weeks of real use
Status: Met
Evidence: since August every supplier invoice has gone through it, with corrections on few of them
Gate 4
Gate: Data: The information it needs is available, accurate and allowed to be used
Status: Met
Evidence: it reads only our own supplier invoices, which our data policy allows
Gate 5
Gate: Process: The new way of working is written down
Status: Met
Evidence: the process is written in the accounts handbook
Gate 6
Gate: Training: The people who will use it have been shown how
Status: Met
Evidence: the three accounts staff were trained in July
Gate 7
Gate: Control: A person checks its output wherever a mistake would matter
Status: Met
Evidence: checks every invoice before it is posted
Gate 8
Gate: Support: Someone knows what to do when it goes wrong
Status: Met
Evidence: If it goes wrong, Tom Reid switches back to manual entry and tells the supplier desk.
Why
Invoice reading has run in accounts since August with every invoice checked before posting, and it has met its agreed correction target each month. All eight gates are met.
Success measure: Fewer than [target] supplier invoices in every twenty need a correction each month
Suggested owner: Tom Reid
Pilot 2
Pilot: Customer email drafts
Verdict: Fix first
Risk: Low
Scale gates
Gate 1
Gate: Owner: A named person is responsible for it
Status: Not known
Gate 2
Gate: Measure: Success is measured with a number agreed in advance
Status: Not met
Evidence: nobody has measured anything yet
Gate 3
Gate: Results: It has results from at least four weeks of real use
Status: Not met
Evidence: tried by two people in customer service for three weeks
Gate 4
Gate: Data: The information it needs is available, accurate and allowed to be used
Status: Not known
Gate 5
Gate: Process: The new way of working is written down
Status: Not known
Gate 6
Gate: Training: The people who will use it have been shown how
Status: Not known
Gate 7
Gate: Control: A person checks its output wherever a mistake would matter
Status: Not known
Gate 8
Gate: Support: Someone knows what to do when it goes wrong
Status: Not known
Why
Two people in customer service have tried it for three weeks and like it, but nothing has been measured and the trial is shorter than four weeks. Ownership, data use, written steps, training, checks and support are not described.
Success measure: Share of customer emails sent from a draft with only minor edits, at least [target]
Suggested owner: Aisha Patel
Close the gaps
Fix 1
Gate: Owner
Fix: Name one person, such as the customer service lead, as responsible for the pilot.
Fix 2
Gate: Measure
Fix: Agree a number to measure, such as the share of drafts sent with only minor edits, before continuing the trial.
Fix 3
Gate: Results
Fix: Keep the trial running in real use until it has at least four weeks of results against the agreed measure.
Fix 4
Gate: Data
Fix: Confirm which customer information the drafts use and that the data policy allows it.
Fix 5
Gate: Process
Fix: Write down the steps for creating, reviewing and sending a drafted email.
Fix 6
Gate: Training
Fix: Show each person who will use it how to use it and review drafts before anyone new starts.
Fix 7
Gate: Control
Fix: Make sure a person reads and approves every draft before it is sent to a customer.
Fix 8
Gate: Support
Fix: Decide who staff contact when a draft is wrong and how they go back to writing emails themselves.
Pilot 3
Pilot: Route planning suggestions
Verdict: Stop
Risk: Low
Scale gates
Gate 1
Gate: Owner: A named person is responsible for it
Status: Not known
Gate 2
Gate: Measure: Success is measured with a number agreed in advance
Status: Not known
Gate 3
Gate: Results: It has results from at least four weeks of real use
Status: Not known
Gate 4
Gate: Data: The information it needs is available, accurate and allowed to be used
Status: Not met
Evidence: drivers found the routes ignored low bridges
Gate 5
Gate: Process: The new way of working is written down
Status: Not known
Gate 6
Gate: Training: The people who will use it have been shown how
Status: Not known
Gate 7
Gate: Control: A person checks its output wherever a mistake would matter
Status: Not known
Gate 8
Gate: Support: Someone knows what to do when it goes wrong
Status: Not known
Why
Tested in October, its routes ignored low bridges, which is a safety problem, and nobody uses it now. It is not worth continuing in its current form, and route decisions stay with people.
Success measure: Share of suggested routes accepted by the transport manager with no safety issues, at least [target]
Suggested owner: Gary Holt
Roll-out waves
Step 1
When: Week 1 (Mon, 30 Nov 2026)
What: Invoice reading: roll out to Accounts
Step 2
When: Week 1 (Mon, 30 Nov 2026)
What: Route planning suggestions: stop it, tell the people using it, and keep what was learned
Step 3
When: By week 6 (Mon, 4 Jan 2027)
What: Close the gaps: Customer email drafts
Step 4
When: Week 7 (Mon, 11 Jan 2027)
What: Customer email drafts: if every gate is met, roll out to Customer service; if not, decide at the review
Step 5
When: Week 12 (Mon, 15 Feb 2027)
What: Steering review: what scaled, what to fix, what to stop, and the next wave
Message to staff
Rolling out first: Invoice reading. Being improved before any roll-out: Customer email drafts. Stopping: Route planning suggestions.
We are moving the AI pilots that have proved useful into everyday work so that routine tasks take less effort and you have more time for the work that needs your judgement. AI only helps: a person always stays responsible for the work, checks the output where a mistake would matter, and makes the decisions about money, people and safety. Where something has not worked well enough, we will not use it. Your experience matters. If something looks wrong, is confusing, or could be done better, please tell your manager or the person responsible for that work so we can put it right.
The scale gates
- Owner: A named person is responsible for it.
- Measure: Success is measured with a number agreed in advance.
- Results: It has results from at least four weeks of real use.
- Data: The information it needs is available, accurate and allowed to be used.
- Process: The new way of working is written down.
- Training: The people who will use it have been shown how.
- Control: A person checks its output wherever a mistake would matter.
- Support: Someone knows what to do when it goes wrong.
Monthly steering meeting
- Each pilot and wave: on track, behind or stuck, from its owner in two minutes.
- The measures: adoption, quality and any problems reported since the last meeting.
- Gates: which fix-first pilots have closed their gaps and can join the next wave.
- Decisions: scale, fix or stop, with who does what by when.
What to track
- Adoption: how many of the people who could use it did so each week.
- Quality: how many outputs a person had to correct, and what kind of mistake.
- Time: the time a task takes before and after, timed on a small sample of real work.
- Problems: every issue reported, how fast it was fixed and whether it came back.
- Feedback: what the people using it say at each steering meeting.
Only for you: not included when you copy, print or download
Check before using
- Invoice reading is high risk by our rules: keep a person checking it as it scales.
What the AI does
It does
- Checks each pilot against eight gates using your words
- Writes success measures, fixes and the staff message
- Suggests owners and teams only from your lists
- Never states costs or savings
It does not
- Know anything about a pilot you did not tell it
- Measure the results for you
- Set up or change the AI tools
- Decide for your steering group
Please check itRead the gates marked not known: tell the tool more and run it again, or agree them at your steering meeting.
Limits
- Your AI pilots: up to 10 lines.
- What they have shown: up to 3,000 characters.
- Who could own them: up to 15 lines.
- Teams to roll out to: up to 12 lines.
- AI tools you use: up to 8 lines.
- Constraints: up to 600 characters.
- Anything else: up to 1,500 characters.
- Up to ten pilots per run.
- Your workspace's monthly cap: a run that would take you over it pauses, and carries on where it stopped when you raise the cap.
Your data
- What you enter, the files you choose and the result are stored encrypted with your workspace's own key. Only people in your workspace can open them.
- To produce the result, what you give the tool (and the parts of your Business profile it uses) is sent to our AI provider, Anthropic. It uses it only to answer, and its terms do not allow it to train its models on it. We do not train on it either.
- Files you upload are deleted after 30 days unless you choose to keep them. You can delete a run at any time: what you entered, the result and the progress notes go; only the usage record stays, because it is part of your billing.
- Our service is hosted in the United States.
The full details are in our privacy notice.
Questions
How are runs charged?
A one-time licence fee of $800 (£800 in the UK), then $200 a month (£150 in the UK), including any tax that applies. On top, you pay for the AI each run uses, billed each month with your invoice and never more than the monthly cap you set. From your fourth tool there is no licence fee.
Can I change the result?
Yes. Choose Edit this result to change any wording or figure, and Restore the original result to go back. Copy, print and downloads use your edited version.
Is it ready to use now?
Yes. Sign in, open Tools in your account and choose Get started for this tool.
Why is a gate marked not known?
Because nothing you wrote shows it. A gate counts as met only when your own words show it, so add the detail and run it again.
Who decides the verdict?
We do, from fixed rules: every gate met means ready to scale; stop only when what you wrote shows the pilot is not worth continuing; anything else is fix first.
If something goes wrong
- A pilot is fix first when you think it is ready.
- Check which gates are not known and add the evidence in What they have shown.
- No teams in the waves.
- Add the teams to roll out to, a line each with their size.
Still stuck? Use the enquiry form and say which tool and what you saw.