The greatest danger when using AI in the Foresight process is not introducing it too late but entrusting it with too much. Those who view AI as a specialized employee with clearly defined tasks gain speed, new perspectives, and more room for their own judgement. This enables companies, especially those without a dedicated Foresight department, to access capabilities that were previously available primarily to large organizations. In this way, they can better prepare strategic decisions under uncertain conditions.
Foresight platforms now deliver analyses and results in seconds rather than weeks. This is precisely where danger lies. When a result is produced quickly and sounds plausible, hardly anyone asks what it is based on. We observe this type of blind trust repeatedly among our customers. Speed alone does not necessarily lead to good decisions. What matters is whether the data fits the problem, whether reliable insights can be derived from it and whether a well-founded decision is made in the end.
AI as an agent with a clearly defined task
It is worth thinking of AI less as a generic tool and more as an agent with a clearly defined task. It is particularly well suited to continuously scan for new signals, monitor developments, and assess information against defined criteria. It also provides efficient support when structuring, visualizing, and creating visions of the future. Its strength lies in speed and consistency, not so much in creativity or strategic judgement. This is creating new access to Foresight, particularly for small to medium-sized companies: tasks that previously required dedicated teams of scouts, analysts and editors can now be carried out with AI support.
Recommendations: How companies create value with applying AI in Foresight
Define clear AI tasks:
Use AI selectively for analysis, monitoring and structuring, but not as a substitute for strategic assessment.
Create transparency:
Document sources and results in a traceable manner. Only transparent analyses provide a reliable basis for decisions.
Establish human decision points:
Consciously review and prioritize results at the transitions between scanning, assessment, interpretation and decision-making.
Ensure data sovereignty:
Operate the platform in-house and process sensitive company data securely. This allows you to retain control over data and decisions within the company.
AI accelerates analysis.
The decision remains human.
Quality through contextualization
Generic language models often produce imprecise results when asked Foresight-related questions, because sources, regions and time horizons get blended together. Contextualization provides a solution for this: AI agents are given defined sources and perspectives, such as a regional focus or a scientific viewpoint. A profile called “West”, for example, draws exclusively on media from the US, Canada, the UK, and Europe, while another focuses primarily on patents and academic publications. The key point is that AI agents do not run themselves. Companies must actively maintain them, question their results and continuously adapt them to their own questions.
Transitions between phases keep humans in the loop
Classical Foresight processes generally follow the same pattern: scanning, assessing, interpreting, and deciding. However, the greatest errors occur at the transitions between these phases, for example, when signals are assessed too broadly or personal judgements are interpreted as facts.
Every transition should therefore include a human decision point. AI can work at high speed within an individual phase, but approval to move on to the next phase must still be given by a person.
A company’s own data creates value – and deserves protection
AI generates the greatest value when it compares external perspectives with a company’s own data and projects them into the future. Only then does an observation become a basis for making quicker decisions with greater confidence. What matters is where external and internal perspectives converge, and where they diverge.
For companies to connect their internal data with AI, they must consider the infrastructure trustworthy. For us, this means: the platform can be operated entirely in-house, physically in the company’s own data center and certified according to established security standards – as opposed to a rented service by a major cloud provider. This also applies to smaller companies without large IT departments.
This keeps sensitive data within the company, and leadership teams can trust the assessments because they can understand how they were reached. AI must not be a black box in Foresight: every assessment must be explainable.
AI also has its limits
Despite its strengths in data processing and pattern recognition, AI remains a specialist with a limited mandate. It recognizes patterns in existing information, but not genuine disruptions beyond its database. Where signals are contradictory or strategic priorities need to be set, it provides options but not decisions. Synthesis, assessment and responsibility therefore remain leadership tasks. If an assessment subsequently proves to be wrong, responsibility lies with the person who approved it, not with the agent that produced it.
Conclusion
AI does not make Foresight more successful simply because it processes more information. Its real value lies in better preparing strategic decisions. Companies identify opportunities and risks earlier, reduce the analytical workload, and gain time to develop robust courses of action. Modern Foresight platforms create the necessary link between continuous scanning, AI-supported analysis, and company-specific context. One thing remains crucial, however: responsibility for strategic decisions lies with people.
The author
Thomas Kolonko is the founder and Managing Director of 4strat GmbH. Through the company’s Foresight platform, he supports public- and private-sector clients of all sizes worldwide in translating AI-supported analyses into strategic decisions.
Ad Special ManagerWissen „Strategic Foresight“ at Harvard Business Manager 10/2026