DE
EN
Let me start with the admission, because anything else would be dishonest: I have produced innovation theater myself. Three days from my consulting years, twelve people from four departments, four future images, a poster everyone was proud of. It hung in the meeting room for two years. Nobody ever did anything with it.
The uncomfortable part is not the poster. It is that I considered that workshop a success: everyone satisfied, nobody causing trouble. Today I know that was the problem.
Since moving from consulting into a corporate group I carry one specific worry: not that a project fails, but that after three years I am the woman who runs workshops and nothing else, and that somebody says “innovation theater” in the hallway and means me.
When even the regulator reschedules its own ten-year plan
Consider something planned more carefully than any strategy offsite. The EU’s carbon border adjustment mechanism had a timetable: reporting obligations until the end of 2025, certificates from January 1, 2026, and free allocation under emissions trading melting away — 97.5% left in 2026, 14% in 2033, nothing from 2034 on, spelled out year by year in the legal text.
In July 2026 the Commission proposed stretching that path: 15% of the allocation already phased out returns from 2028, the exit drags on until 2038. A proposal, not law in force — but a ten-year plan with impact assessment, consultation and legal scrutiny is to be pushed back four years after three. If it gets adjusted where the rules themselves are written, on what grounds do I expect my poster to hold after three days?
The graveyard of promised dates
All of it is on the record. In 2021 ArcelorMittal announced up to 3.5 million tonnes of steel at two German sites moving to hydrogen-based direct reduction by 2030. In November 2024 it put the decision on hold: green hydrogen maturing slowly, weaknesses in the border adjustment mechanism, customers barely willing to pay a premium. By June 2025 it could not proceed at all: the EUR 1.3 billion in federal funding was tied to construction starting that month.
That same June, SSAB postponed its new mill in Luleå by twelve months — not steel technology, but grid reinforcements not delivered as planned. Stegra in northern Sweden closed a EUR 1.4 billion financing round in June 2026 and has its timeline under review. None of these is a steel problem: they are boundary conditions treated as given in 2021, because they were not the interesting variable.
Projects that hold their dates are cut differently. voestalpine names the first half of 2027 and about 30% less CO2 by 2029 against 2019, at an investment of EUR 1.5 billion — a first stage built around an electric arc furnace. Date only a stage that does not hang on a hydrogen market, and there is nothing to take back when the market fails to appear.
The pressure to plan big is real: iron and steel directly accounted for 2.6 gigatonnes of CO2 in 2019, about 7% of global energy-related emissions. But which ten-year commitment holds?
My blind spot, uncomfortably specific
Years of pushing an AI portfolio through a group where every division holds its own numbers to be right sharpen the view of two things.
First: I plan too optimistically. Not occasionally but systematically, and I have known it since my consulting years. On every rollout since, I was wrong about the time required, always in the same direction. Knowing it has not cured me — it only lets me explain afterwards why this time was different.
Second, and more expensive: I most enjoy talking to people like me. Innovation people sit with innovation people, share a vocabulary and the same assumptions about what is obvious — and take the agreement that comes of it for a result. It is not one. It is a side effect of the seating plan.
If all four future images rested on the same two assumptions — digitalization prevails, regulation tightens — I did not have four images. I had one in four colors. The purchasing manager who would have said “price decides here in the end, and that will not change in ten years” was not on my list: he does not belong to the innovation community.
The number I used myself
For years I claimed in presentations that 60% to 90% of all strategies fail in implementation. The figure lands well. It justifies budget.
It is not substantiated. The trail leads to a Harvard Business School working paper from 2005: “various sources” have found failure rates of between 60 and 90 percent — and “various sources” is the whole citation. No footnote, no author, no methodology. Two researchers checked it in 2015: the true rate remains undetermined, the estimates in the literature outdated, fragmentary, fragile or simply absent. I carried that number for years without checking it, because it justified my work — the mechanism I spot instantly in others.
What does exist is less spectacular. A survey of nearly 8,000 managers in more than 250 companies found that only 55% of middle managers could name even one of their company’s top five priorities. Not all five. One. And a study with 400 managers found in 2025 that foresight measurably improves strategic conversations — while the expected effect of those conversations on decision-making could not be confirmed.
Foresight produces conversations. Whether they become decisions is the open spot — and open spots close with practice, not creativity.
Four practices that make the difference
Futures studies has had antidotes for decades, unspectacular ones — probably why I drop them when the afternoon gets tight.
Ask expectation, wish and fear separately. People answer differently depending on whether they expect a future, want it or dread it. Ask in one lump and the worry vanishes into the average. The triad goes back to Robert Textor’s handbook on ethnographic futures research; it costs three questions instead of one.
Force the contradiction. Jim Dator coined the line that any useful idea about the futures should appear ridiculous at first. As a rule: every voice must name something that sounds absurd today. Nobody does it voluntarily, least of all in front of a division head. As a mandatory field, everybody does.
The scene instead of the driver list. In 2019 Alessandro Fergnani described a method that brings future images to life through narrated figures: a Tuesday morning in the target year, three sentences, one person doing something. Not a creativity technique but a precision instrument — you can contradict a scene. You cannot contradict “rising regulatory density.”
Work backwards. Here my poster failed: images, then stumbling forward. In 1990 John Robinson described the reverse route under the lovely title “a recipe for people who hate to predict,” and in 2006 Quist and Vergragt extended it for stakeholder work: set the end state, work backwards, name the stages until you reach next quarter — with Roy Amara’s correction built in, short run overestimated, long run underestimated. Then you land on something that can start Monday morning.
What this has to do with synthetic panels
I am professionally biased here, so I will say it openly. For this column I tried a Future Round at Radical Personas: five to sixteen voices answer one question about the future separately, and out of that come contrasting future images, a map, and a roadmap back to next quarter.
What interested me was less the output than what the process would not let me skip. The three questions could not be collapsed into one. The uncomfortable possibility was a mandatory field, not a suggestion. And I could not stop at the images: nothing moved on without working backwards. Precisely the four places where I cut corners once the afternoon gets tight.
It also reaches my second blind spot: a panel can hold voices I cannot get into a real room — the purchasing manager for whom price decides, the plant operator who lives through a conversion in shift work.
Two limitations belong here, or I would be the person I criticized above. These practices were developed for human respondents and tried out on humans; applying them to AI-modeled personas is an adaptation, not a validated use case of the research I cite — the methodological literature does not carry that step. And what comes out is not a survey: modeled assumptions, not a forecast, no substitute for a conversation with real customers.
If you are curious: the Future Round is a small, guided beta whose reports end with first steps today.
How I recognize a good foresight workshop now
Not by the poster. By three things: two camps that could not convince each other, both left in the minutes; one statement uncomfortable enough to question an assumption a running project rests on; and something that can start next Monday without a budget request.
If all three are missing, you had a very pleasant day. Just not foresight.
Martin Kocijaz 
Franz Hinterberger