Skip to content
Agentic·Foresight

Concept · Method

Agentic Foresight

Agentic Foresight is foresight work in which specialised AI agents search for signals, build scenarios and test strategies against them. Humans ask the questions, judge and decide.

A project by Prof. Dr. Ulf Pillkahn (FOM University of Applied Sciences, TU Munich)

01 / The question

Where do agents fit into scenario work?

Scenario work has followed the same textbook sequence for decades: scan the environment, turn observations into drivers, develop scenarios from them and measure your own strategy against them. Every step ties up expert time, and much of that time goes into research, sorting, assessing and writing.

That work can be broken into tasks and handed to specialised agents. The result is three levers along the sequence. Each has its own agent and a clear boundary: what the agent prepares, a human decides on.

Lever 01

Scout

Signals and drivers

Scout searches a defined search field, records signals together with their evidence, groups them into themes and proposes drivers. Humans set the question and the search field, and they choose the drivers.

How agents find signals →

Lever 02

Outlook

Scenarios by the book

Outlook takes the drivers through an impact matrix and a Wilson matrix to the key factors, sets up the scenario framework and writes out the scenarios. Humans confirm key factors, framework and plausibility.

How agents build scenarios →

Lever 03

Tunnel

Strategic fit in the wind tunnel

Tunnel tests your strategy against every scenario, shows where it holds and where it breaks, and derives actions and early-warning indicators. Judgement, priorities and the decision stay with humans.

How agents test strategies →

02 / In context

Classic, continuous, agentic

Agentic Foresight does not replace the familiar forms; it shifts the work. The difference lies in who does the research, who builds the scenarios and what humans spend their time on.

FeatureClassic foresightContinuous foresightAgentic Foresight
Rhythma one-off project with a start and an endongoing scanning; scenarios are revised at intervalsongoing; signals and scenarios are updated as soon as the situation changes
Who researchesproject team: desk research, interviews, workshopsmonitoring team with databases and alertsScout agent, with evidence; humans set the search field and make the selection
Who builds scenariosexperts in workshopsexperts, updated when neededOutlook agent, following the textbook sequence; humans confirm framework and plausibility
Role of humansdo almost everything themselvesobserve, interpret, reportask, judge, decide
Outputstudy or reportradar, trend report, newslettercontinuously updated scenarios with evidence, a strategy check, early-warning indicators

03 / Division of labour

Machines design futures, humans decide

The division of labour behind this is called Flipped Foresight. It reverses the usual use of expert time: the scalable part, producing material and variants, goes to the machine. Scarce human attention goes where it cannot be replaced: judgement, ownership and decision-making.

Agents take on

Searching, recording, grouping, rating against fixed criteria, writing out scenarios, checking the strategy against each scenario, keeping track of the evidence.

Humans decide

Which question matters, which drivers are selected, whether a scenario is plausible, what it means for the strategy and what follows from it.

The method is described in the working paper on Flipped Foresight on SSRN. For an overview of its phases and principles, see the Flipped Foresight method page.

04 / Limits

What agents cannot do

It helps to think of the future in three bands. Two of them can be calculated quite well, the third cannot.

Stable

What barely moves can be extrapolated. Here agents are fast and reliable.

Change

What is already moving leaves signals. Agents can find them, group them and play them through in scenarios.

New

What does not exist yet leaves hardly any trace in data. Recognising it and taking it seriously remains a matter of human judgement.

05 / Lab

Tested in our own agent cluster

These methods are not drawn up on a whiteboard. They are tested in a cluster of AI agents that runs around the clock: research, scenarios and publishing happen there every day, with all the errors that real sources and real data volumes bring.

A look inside the agent cluster and its logbook (in German)

06 / For organisations

Agents for your foresight work

Would you like to use Scout, Outlook or Tunnel in your own organisation? Operation Zukunft designs, builds and runs AI agents for innovation and foresight, on your own or on hosted infrastructure.

Build agents for your foresight work →
Portrait of Prof. Dr. Ulf Pillkahn

Who is behind it

Prof. Dr. Ulf Pillkahn

Prof. Dr. Ulf Pillkahn (FOM University of Applied Sciences, TU Munich) develops AI-supported methods for corporate foresight and scenario development (Flipped Foresight) with Operation Zukunft and builds AI agents for trend and technology monitoring.