Stand by the sea for a moment and look straight ahead. How far can you see? It feels endless. In fact: barely five kilometres. Then the curvature of the earth closes in, and everything beyond it simply does not exist for your eye. However good your eyes are, however expensive your binoculars were: the horizon is a hard physical limit on your vision.
Seafaring never made a drama out of this; it made a principle of it. No navigator ever tried to fix a route over thousands of kilometres in one stroke and then sail it blind. You sail towards the horizon, gain new sight, correct your course and then take on the next horizon. Stage by stage.
And now the cut to the meeting room: in data and AI strategy projects I regularly encounter the opposite of this principle. A five-year roadmap, a Gantt chart running to the target picture “data-driven company”, every initiative tagged with a quarter and a budget. It looks assured. But it claims a range of vision nobody has, because in a data and AI journey, realistically, three to six months are reliably plannable. Everything after that is cartography of waters no one has seen yet.
Why the five-year chart capsizes
A data and AI journey is a multi-year process that changes three dimensions at once: business, people and technology. And that is precisely why master plans fail reliably on two assumptions that almost never hold in practice.
Assumption one: all stakeholders commit to the whole journey up front. They do not. A CFO signs for what they can see, not for a target picture five years out. And frankly: rightly so.
Assumption two: a central data and AI team sails to the target picture alone. That does not work either. Value is created where business units use data and AI in daily operations. And they only come aboard if they see results along the way, not just at the end of the crossing.
The consequence is not giving up on planning, but a different structure: a horizon plan. Behind it sits the flywheel principle: the first success in production generates the momentum, meaning the motivation and the budget, for the next step. You plan sharply to the next horizon and directionally beyond it.
Four horizons, three tracks
The horizon plan we work with has four stages, and on every stage three tracks run in parallel: business, people, technology. The basic rule: business leads, people and technology enable. All three develop in sync, but only ever as far as the business actually needs. You do not build a container ship for a coastal run.
Horizon 1 — strategic foundation, roughly three months. This is where the course is set: define the value ambition, prioritise the use case portfolio, secure executive sponsorship. On the people track a cross-functional team takes shape, along with a first operating model and a governance framework; the data and AI culture gets measured honestly instead of talked up. On the technology track: assess the data landscape and AI readiness, define the target architecture, identify quick-win enablers. No big ship is being built yet, but the chart, the crew and the route are in place.
Horizon 2 — validated impact, another three months or so. Now you prove the vessel floats: one or two proofs of value go into production and are measured against hard figures. Deliberately proof of value, not proof of concept, because “is it technically possible?” is rarely the interesting question. On the technology track an architecture MVP emerges, on the people track agile delivery is tried out at small scale, and the business sharpens the funding and scaling case from the results.
Horizon 3 — industrialise core value, around twelve months. The validated core use cases get scaled, data- and AI-driven decisions move into the processes. The platform is expanded to scale, data models and pipelines are standardised, monitoring and data quality move in. On the people track: establish data product ownership, institutionalise governance and master data management, systematically upskill the organisation. The expedition becomes scheduled service.
Horizon 4 — continuous innovation, no end date. The portfolio grows, now including AI use cases with a higher degree of autonomy and use cases that change the business model itself. The culture carries, the platform is continuously modernised. Not a finish-line photo but a state. From here you see horizons that were invisible at the start.
Why horizons work where roadmaps fail
At this point the question reliably comes: is this not just a roadmap with prettier names? No. The difference lies in three properties a classic roadmap does not have.
Every horizon carries a burden of proof. Horizon 1 proves: we know where to, why, and who stands behind it. Horizon 2 proves: measurable value is created here, and we can fundamentally do this ourselves. Horizon 3 proves: it works in daily operations, without someone performing heroics every week. Horizon 4 proves continuously: we get better on our own. You earn the next horizon rather than inheriting it from a Gantt chart.
Investment follows value. Management does not have to release millions for horizon 3 today, only fund the next stage, based on what the last one demonstrated. Risk decreases with every horizon instead of piling up at the end. That is exactly what makes the horizon plan signable for a sceptical CFO.
Business, people and technology stay in time. All three tracks develop together in every horizon. Nobody races ahead, nobody lags behind. That is what separates a journey from an IT project with a change chapter bolted on.
In my experience a data and AI journey rarely dies from lack of money anyway. It dies from lack of visible progress — and horizons are built precisely against that.
And now the asterisks
First: three plus three plus twelve months are experience values, not a law of nature. A regulated corporation often needs longer for horizon 1; a focused mid-sized company is faster. The sequence and the burden of proof are fixed, the duration is not.
Second, there are two classic shipwrecks. One: eternal horizon 1. The strategy becomes a 200-page document, every department still wants to be heard, and eventually the journey sits in the archive as a slide deck. Nobody ever sets sail. The other: the PoV carousel. Pilot follows pilot, each a success in itself, but none is ever industrialised, because the uncomfortable build-out of platform, governance and ownership in horizon 3 keeps getting postponed.
And third: the most tempting shortcut is skipping horizon 2 and building the big platform straight away. That is the most expensive known method of discovering that nobody uses it. Equally predictable is the failure of anyone who cuts one of the three tracks: without people, adoption problems; without technology, delivery problems; without business leadership, a technically elegant journey with no destination — and no ROI.
Back on the coast
The beautiful thing about the horizon is this: it is not the limit of your journey. It is only the limit of your sight. Anyone who has sailed the first stage sees waters that were simply not visible from the harbour: new use cases, new shortcuts, sometimes shallows that would never have appeared in the master plan.
The first step is pleasantly unspectacular: determine the course. That does not take a year-long project. Often a structured workshop day is enough, where business and leadership perspectives jointly map and prioritise the potential. Afterwards you have what every good journey begins with: a chart, a crew and a first horizon.
Nobody gets more sight than that. Nobody needs more. Plan what you can see. Sail there. Then look again.