Strategy & Consulting
We interview your team, map where the work piles up, and rank where AI pays and where it doesn't.
Interviews with the people who own the work
One interview per area, about an hour each, with the person who does the job rather than the person who describes it. We ask how the work arrives, where it waits, what gets redone, and what nobody measures. Every "it takes a lot of time" gets turned into hours per week and people involved before we move on. The interviews are recorded, and we repeat each figure out loud, so the numbers come from your team and not from our notes.
- How the work arrives and what it passes through
- Where it waits, and for whom
- What gets redone, and why
- What nobody measures yet
A map of how work runs today
We draw each area's process as it runs, not as the manual says: where a request enters, who touches it, which spreadsheet it lands in, where it sits waiting. Time piles up in the same few places, usually a handoff or an approval, and those places become the candidates. We also write down where the data lives, because a task that runs inside your ERP and a task that runs in email and Excel need different builds, and one of them may need the data fixed first.
Opportunities ranked by ROI
Each opportunity gets a number built from your figures. A is the time it eats, times what those hours cost. B is the revenue leaking today because the work is slow: quotes that go to a competitor, leads that go cold. C is the revenue the freed time could produce, when we can name the activity. A always counts; B and C only when there's a specific leak or a specific new activity, and never both for the same money. Being automatable is not the criterion. Plenty of tasks can be automated and return less than they cost to build. The ranking says which ones pay, in what order.
A roadmap to AI-first, ordered by ROI
One document: the process map per area, each opportunity with its number and its source, what not to touch yet and what blocks it, the sequence by phase, and the open questions marked as questions rather than deduced. The first automation is the one with the best return and the smallest build, so the system earns its keep from the first phase. Sometimes the honest answer is that nothing pays yet, because the data isn't there or the volume doesn't justify a build. We write that down too.
- The process map, area by area
- Every opportunity with its number and where it came from
- What not to do yet, and what blocks it
- The sequence, phase by phase
Phases, not dates.
The order is fixed. The calendar is set once the diagnosis closes.
- Phase 01
Kickoff and scope
We agree which areas enter, who we interview, and which numbers you already track. One session, before any interview.
- Phase 02
Interviews
One hour per area lead, recorded. Every figure gets said out loud so it ends up in the transcript, not in someone's memory.
- Phase 03
Synthesis and ranking
A process map per area, A + B + C per opportunity, a source next to every number, and the low end of every range.
- Phase 04
Roadmap and decision
We present the document to whoever decides. You leave with a sequence and a first step, or with a no and the reasons.
Little, and early.
A short list, all before the first session. None of it takes a day.
- 01The org chart, with who owns each area
- 02One hour with each area lead, recorded
- 03Access to how the work is tracked today: spreadsheets, CRM, ticket tool, inbox
- 04The numbers behind each area: volumes, hours, headcount, what an hour costs
- 05What you've tried with AI before, and what happened
- 06The person who decides, in the room for the final session
What people ask about this.
Then you get the map, the numbers and the reasons in writing, and nothing gets built. The most common no is a data problem: the information lives in loose spreadsheets or a half-filled ERP, and a model on top would return the same noise your team already rejects. In that case the roadmap says what to fix first, and it's usually not AI.
An audit reports the state of what exists. This ends in a decision: a ranked sequence with a first step, or a reasoned not yet. It also doesn't start from the assumption that AI is the answer. Architecture & Governance does, because by then the diagnosis is done.
No. If something ends up running by the end, it wasn't this service. The build is the next step, quoted on its own once the roadmap says what pays and in what order.
Three parts, all with your figures. A, time cost: hours per day, times people, times 260 working days, times what a loaded hour costs. B, lost revenue: volume affected, times the share lost to the delay, times its value, times 12. C, upside: the extra revenue-producing activity the freed time allows. A always applies; B and C only when there's a specific leak or a specific activity, and never both for the same dollars. We take the low end of every range, and every number carries its source: who said it, in which interview, or which system it came from.
Ready to become AI-first?
One discovery call. We tell you where to start, what it takes and what it's worth. If the answer is no, we say so.