Methods · 09/10
Three Horizons
A foresight framework that treats present, transition and future as three overlapping horizons, giving transformation a shared, discussable structure.
Purpose
Three Horizons organises strategic conversations about change along three curves. Horizon 1 is the dominant system of the present—the current business that delivers results today but is gradually losing its fit with a changing environment. Horizon 3 is the envisioned future, whose seeds are already visible in the present. Horizon 2 is the zone of transition: the innovations and ventures that mediate between the two. The framework's central insight is that these perspectives do not follow one another in sequence; they coexist and compete for attention, resources and legitimacy.
Rooted in the growth model published by Baghai, Coley and White at McKinsey in 1999 and reworked as a foresight practice by Bill Sharpe and the International Futures Forum, the framework now serves two main purposes: balancing a portfolio across core business, growth options and long-term renewal, and giving groups with different time horizons—operators, entrepreneurs, visionaries—a common language in which their conflicts become productive rather than destructive.
Procedure in five steps
- Describe Horizon 1
Capture the current state of the dominant system: business model, core capabilities, the logic that makes it successful. Equally important are the symptoms of declining fit—eroding margins, drifting customers, regulatory pressure. These symptoms establish why change is necessary at all.
- Envision Horizon 3
Articulate the desired future: which values, business logics and structures should carry the organisation in ten to fifteen years? Then search the present for seeds of that future—niche players, pilot projects, social pioneers. This grounds the vision in observable reality.
- Map Horizon 2
Collect the innovations, experiments and ventures that mediate between present and future. Distinguish rigorously: H2-plus activities pave the way towards the Horizon 3 future, while H2-minus activities merely extend the life of the old system. This distinction is the analytical core of the model.
- Analyse the tensions
Make the conflicts between horizons explicit: where does day-to-day business drain resources from renewal? Where do Horizon 1 metrics and incentive systems block Horizon 2 options? These tensions are where the real strategic decisions lie.
- Set portfolio and monitoring
Assign initiatives to horizons, allocate resource shares deliberately and define indicators that signal transitions. Revisit the horizon picture regularly—it is a snapshot, not a roadmap.
Strengths and limits
- The three-curve picture is immediately intelligible and works as an entry point even for groups without foresight experience.
- The framework connects the operational present with a normative future instead of discussing them separately.
- The H2-plus/H2-minus distinction exposes pseudo-innovation that merely preserves the status quo.
- It gives the different voices in an organisation—maintainers, builders, visionaries—each a legitimate place.
- The visual suggests a temporal sequence, although all three horizons exist simultaneously and must be managed in parallel.
- The framework generates no content of its own; the quality of Horizon 3 images and Horizon 2 options depends entirely on the analysis that feeds it.
- Horizon 3 invites wishful thinking unless it is tested against alternative courses the future might take.
- It offers no quantitative criteria for resource allocation or transition timing.
With AI and agents
The most laborious part of a Three Horizons exercise is filling the horizons with substance, and this is precisely where language models and agents change the work. For Horizon 1, an LLM condenses internal sources—annual reports, KPI reviews, customer feedback—into a defensible diagnosis of declining fit. For Horizons 2 and 3, agents take over continuous scanning: they monitor journals, patent data, start-up databases and funding programmes for seeds of the future and for emerging transition options. The workshop event whose output was stale after six months becomes a continuously updated horizon picture.
The model's most delicate analytical step can also be machine-supported: classifying initiatives as H2-plus or H2-minus. A language model that has been given the explicit criteria of the Horizon 3 vision can test each initiative against them and formulate a reasoned counter-position—exposing the rationalisations by which self-preservation gets relabelled as innovation. For monitoring, agents connect the defined transition indicators to live data sources and raise alerts when thresholds are crossed, instead of waiting for the next strategy off-site.
Human judgement remains irreplaceable at two points. Horizon 3 is normative: no model can decide which future is worth wanting, because that is a question of values and responsibility, not of patterns in training data. And the productive negotiation between horizon perspectives—the actual heart of the method—depends on people with real stakes having to listen to one another. AI supplies the material and the counter-check; agreeing on where the organisation wants to go remains the work of leadership.
Relation to scenarios
Three Horizons and scenario analysis complement each other directly. Scenarios provide alternative environmental futures against which a Horizon 3 image can be stress-tested—instead of a single preferred future, the result is a target picture that holds up across several plausible worlds. Conversely, Three Horizons gives scenario work a frame for action: it translates the question of which futures are possible into the question of which transitions to prepare today. In practice, a useful sequence is to develop scenarios first and then test the horizon picture and the H2 options for robustness across them.
How ready is your organization for AI and agents? — Take the AI Readiness Check