Un plan que se puede defender en comité, no una lista de ideas.
Partimos de sus procesos y de sus datos reales para producir una hoja de ruta priorizada: por dónde empezar, por qué, cuánto cuesta, cuánto aporta y qué hay que rechazar.
- Priorizado por esfuerzo/retorno
- Con cifras y calendario
- Lo que no hay que hacer
Why AI roadmaps end up in a drawer
Almost never through lack of ambition. Most often through excess of it.
Twenty use cases, no priority
The workshop produced a long and exciting list. Nothing says where to start, so nobody starts, and the document ages.
No figures behind the promises
Gains are stated in round percentages, with no basis of calculation. The board can neither approve nor arbitrate, and the subject is deferred.
The real state of the data is ignored
The plan assumes clean, accessible data. It is neither, and the first real project — the one not in the plan — takes six months.
How we build the plan
Two to three weeks, with your teams. We look at your real data, not a comfortable sample.
Mapping processes and data
Where the work happens, what hurts, what data actually exists and in what state. This step almost always surfaces workarounds and unofficial files nobody mentions in meetings — and those determine feasibility.
Use-case inventory
Everything that could be done, including ideas coming from the teams rather than from management. No selection at this stage: selecting without a complete inventory is a bias, not a method.
Costing and prioritisation
Each case is assessed on effort, expected return, data dependency, regulatory risk and integration complexity. The resulting ranking is arguable — that is the point: it gives the arbitration a frame.
Target architecture
What must exist technically for the priority cases to be possible: document layer, APIs, rights management, infrastructure. Prerequisites appear explicitly instead of being discovered during the first project.
Roadmap and budget
A twelve-month sequence, with the deliverable, budget, timeline and success criterion for each step. Plus the list of what we recommend not doing, with reasons.
What you receive
A working document usable at board level, not an inspiration deck.
- A map of the processes concerned and the real state of your data
- A complete inventory of identified use cases, including those from the field
- The effort-to-return ranking, with the costing method made explicit
- The target architecture and the technical prerequisites of each priority case
- A twelve-month roadmap: deliverables, budgets, timelines, success criteria
- A reasoned list of what not to do, and why
- Regulatory risk analysis per use case
Three starting points
The board asks for an AI strategy. Every department has ideas, none is costed, and the arbitration goes in circles.
Cross-functional inventory, consistent costing of every case, reasoned prioritisation and a twelve-month roadmap.
The board arbitrates on comparable figures. The first project starts with a written budget and success criterion.
Business units launch experiments on their own. IT finds out about the tools afterwards.
Census of actual uses, shared target architecture, pooled technical foundation and a governance framework.
Initiatives converge on a common foundation instead of multiplying isolated components to maintain.
An AI project failed last year. Management wants to know whether the subject still deserves investment.
Analysis of the previous failure, re-examination of use cases against real data, conservative costing and a narrowed perimeter.
A documented answer, including the possibility that the answer is to wait — which is also a result.
The frame of the engagement
Scoping is billed, and deducted if you continue with us. That is what guarantees it is done seriously.
- of work with your teams, on your real data
- 2-3 weeks
of work with your teams, on your real data
- roadmap horizon, with a budget per step
- 12 months
roadmap horizon, with a budget per step
- of cases costed with the method made explicit, including those we set aside
- 100 %
of cases costed with the method made explicit, including those we set aside
Estos órdenes de magnitud proceden de lo que medimos en el piloto sobre alcances comparables. Sobre su corpus se miden antes de la industrialización, no se prometen antes.
Preguntas frecuentes
The diagnostic takes thirty minutes and identifies the first use case. The transformation plan takes two to three weeks, examines your real data and covers all use cases over twelve months.
Yes, as a fixed price, and it is deducted if you continue with us. Free scoping would be rushed scoping.
No. The plan is yours, including if you have someone else execute it. That is a fair test of its quality, in fact.
The people who do the work, not only those who describe it. That is where the workarounds live, and therefore the real feasibility.
We say so and argue it. It has happened, and it is more useful than a plan that costs you a year.
Yes, each use case is analysed for regulatory risk and the obligations that follow. The final legal qualification rests with your counsel.
Lo que esta solución no hace
We sometimes conclude that nothing should be launched this year, and we say so. An honest plan can recommend starting by cleaning a document estate, or replacing a tool before adding anything to it. We do not produce plans that assume data you do not have, and we do not quote gains we would not know how to measure at pilot.
Where should you start?
The thirty-minute diagnostic already answers part of that question, for free. The full plan is built on your data.