Planning Office · AI strategy & pilots
Pick the right workflow. Prove it with real data.
We map your workflows with your team, rank the use cases worth building and run a focused pilot on real data. Success measures are agreed before a line of code is written.
- 01 Start
- Discovery Sprint2–3 weeks
- 02 Prove
- Agent Pilot6–8 weeks
- 03 Decide
- Go, adjust or stopYour call
Opportunity Finder
Where should you start? Find out in a minute.
Pick your sector and tick the everyday pains you recognise. We place the candidate use cases on value and effort, the same way we do in a Discovery Sprint.
Tick two pains and watch where your use cases land.
Suggested first pilot
Price-list intake agent Pilot first
Reads supplier price lists in any format and prepares ERP updates for a person to approve.
Why this one: High value for lower effort: prove it in a pilot. It answers “We retype supplier price lists”.
Next in line: Bilingual customer replies · Pilot first
Indicative. We confirm it with your data in a Discovery Sprint.
The Discovery Sprint
Three weeks to a plan your leadership can approve.
Two to three weeks, depending on scope. This is the full three-week version: what you give and what you get, week by week.
01 Week 1
Map
Map the work as it really happens
Workshops and interviews with the people who do the work. We follow real cases from start to finish, exceptions included.
You give
- Time with the people who do the work
- Sample documents and data under NDA
- A contact in IT
You get
- Workflow map
- Data and systems inventory
02 Week 2
Prioritise
Score every candidate, honestly
Each candidate use case is scored for value, effort, data readiness and risk. We record how the work performs today as the baseline.
You give
- A scoring session with your team leads
- Today’s volumes and handling times, where you have them
You get
- Prioritised use-case shortlist
- A business case for each candidate
- Baseline measures
03 Week 3
Plan
Design the pilot, agree the measures
We design the pilot for the top use case: scope, data access, success measures and the weekly review rhythm.
You give
- A decision-maker at the read-out
- Confirmed data access for the pilot
You get
- Fixed-scope pilot plan
- Agreed success measures
- Leadership read-out
Your sprint pack
Yours to keep, whatever you decide next.
From sprint to pilot
One workflow, proven on real data. Then a clear decision.
The sprint ends with a pilot plan. The Agent Pilot runs it on your real data in a controlled environment, and your team reviews the results every week.
- Discovery Sprint2–3 weeks
- Pilot designScope, data, measures
- Agent Pilot6–8 weeks on your real dataWeekly review with your team
- GoMove to production with the agreed controls.
- AdjustChange what didn’t work and test again.
- StopA valid outcome: you learn it before a full rollout.
During the pilot we need
- A process owner and reviewers
- Test access to the systems involved
- A weekly feedback session
At the gate you have
- A working agent team and command centre
- Connection to one or two of your systems
- Evaluation against the agreed measures
- A roll-out plan and cost model
Success measures
How we’ll know it worked.
Before the pilot starts, we agree what success means and how it is measured, and both sides sign it. The goalposts don’t move.
- The baseline is recorded during the sprint.
- The same kinds of cases are measured every week.
- Results are read out at the decision gate, good news or not.
Pilot scorecard
Agreed and signed before the pilot starts
- Workflow
- The use case you choose in the sprint
- Reviews
- Weekly, with your reviewers
| ✓ | Measure | How it’s measured | Baseline |
|---|---|---|---|
| Time per case, before and after | Timed on the same case types | Set in the sprint | |
| Share of outputs approved without edits | Counted from reviewer decisions | Set in the sprint | |
| Cost per completed task | Model, hosting and review time | Set in the sprint | |
| Cases escalated to a person, and why | Every escalation logged with its reason | Set in the sprint |