Services / Intelligence · CAP.007
An AI plan with numbers, not vibes.
A short engagement that maps your workflows against what current AI can reliably do, and returns a prioritized, costed roadmap with success metrics your CFO will accept.
01 · When you need this
If any of these sound familiar, you’re in the right place.
- The board keeps asking about AI and you need an answer that isn't a shrug or a moonshot.
- You tried a chatbot pilot; it embarrassed itself and the budget.
- Competitors claim AI features and you can't tell marketing from substance.
- Your team drowns in a repetitive task that feels automatable; you want to know if it actually is.
02 · What we deliver
Exactly what you walk away with.
No vague 'solutions'. These are the concrete deliverables: each one is a line in the proposal and a checkbox at handover.
DEL.01
Workflow audit
Your processes ranked by automation potential, with the 'looks automatable, isn't' traps flagged.
DEL.02
Feasibility verdicts
For each candidate: what today's models genuinely deliver, error rates included.
DEL.03
Costed roadmap
Build cost, running cost, and expected saving per use case. Sequenced.
DEL.04
Success metrics
What you'll measure so an AI project can fail honestly instead of limping forever.
DEL.05
Risk & governance notes
Data privacy, failure modes, and what a human must still check.
03 · How it runs
Phase by phase, with nothing hidden.
Every phase states what happens and what you see in your project portal while it does.
Immersion
2-3 days
We watch the actual work, where the hours actually go, not where everyone assumes.
In your portal: Observation notes.
Feasibility testing
3-5 days
Quick technical probes with your real data samples, evidence, not benchmarks from a blog.
In your portal: Probe results, including the failures.
Roadmap
2-3 days
The costed, sequenced plan with metrics, and the board-ready summary.
In your portal: Final report and walkthrough.
04 · Standards
Non-negotiables, in writing.
'No' is a valid finding
If AI won't beat a checklist or a €100/month tool for your case, the report says so, in bold.
Tested on your data
Feasibility claims come from probes on your actual documents and tickets, not vendor demos.
Running costs included
Model costs at your real volumes, the number most AI pitches forget.
Failure modes named
Every recommendation states what wrong looks like and who catches it.
05 · Stack
Tools chosen for your handover, not our comfort.
Model-agnostic by design: the roadmap prices at least two model options per use case, because vendor lock-in is a cost too.
06 · Questions
Asked often, answered straight.
Isn't this just riding the AI hype?
Is our data safe during the probes?
Do we have to build with you afterwards?
Ready to talk about ai strategy?
Describe where you are. We’ll respond within one business day with honest next steps, even if the honest next step isn’t us.
Start the conversation