Strategic foresight: getting ready for the agentic era

“Sorry, you have used all of your tokens.” A very 2026 message. Halfway through an idea that was taking shape with the help of an artificial intelligence model, the conversation cuts out. At that point you can choose: pay and the problem goes away, or wait however many hours you have coming. It is no surprise that more and more people pay for a subscription (or several). It does not feel like buying a promise, but the continuity of something already working.

Once you use AI, it is hard not to go back to it. And even so, we are only in the hallway of all this: last year, the largest US technology companies (Amazon, Alphabet, Microsoft and Meta) invested $413 billion in infrastructure. This year the figure is expected to reach $760 billion, and projections point towards approaching a trillion dollars by the end of the decade. To sustain that pace they have issued around $220 billion in bonds in 2026 alone. It is, in all likelihood, the largest investment surge in recent history.

AI is transforming the world; it would be naive to deny it. But while that conversation — the one about investment, infrastructure, “everything is going to change” or the existential risks raised by people like Jacob Coxon — takes up all the room, there is another that interests me a good deal more strategically: what are we humans doing to get ready for the agentic era?

What is foresight design?

A city preparing for autonomous cars, a luxury brand reimagining its online universe through hand-drawn illustration, a teenager farming aura to fit in. Taken separately, these phenomena do not mean much. But understood as part of the same system, connected to each other and read together, they begin to sketch a narrative of what is coming. They are signals: clues in the present that anticipate futures which have not fully arrived yet.

None of them landed all at once. Change announces itself quietly: at the margins, in experiments, or in news stories that read as anecdotes at the time and that, in hindsight, turn out to be key to understanding a trend. This is an idea David Alayón writes about very well in his newsletters, and it explains why the future almost never surprises whoever was watching.

Train the eye, direct the gaze. That is how we come to identify signals and understand how they evolve: first at the margins, then in a niche that adopts them enthusiastically, and finally everywhere, as if they had always been there. Take the rise of remote work in 2020. In Spain it grew massively from 5% to 37% because of the pandemic lockdowns, driving a historic shift.

That structural rhythm is what makes this work possible. We do not guess what will happen, we recognise what stage of maturity it is happening at. To be able to recognise it, you have to know the past and be present in the present: that is what lets you anticipate and shape what is to come.

“History does not repeat itself, but it rhymes.”

Phrase attributed to Mark Twain, American writer.

That exercise of looking at small things systematically is what we call strategic foresight. It is a method for interpreting information rigorously. As my colleague Maria Najarro explains in this post, strategic foresight is not about predicting the future, but about defining strategies and building the capabilities to navigate among multiple possible futures.

Agentic e-commerce: approaching it through foresight

Agentic commerce is the perfect case for understanding the difference between anticipating and reacting, because right now it supports two different readings depending on where you look.

If you look at the infrastructure, the future is already here. In barely twelve months the rails have been laid: the Agentic Commerce Protocol from Stripe and OpenAI, the Universal Commerce Protocol driven by Google alongside Shopify, Walmart and Target, and the infrastructure for agentic payments pushed by Visa, Mastercard and American Express. In fashion, for instance, Gap intends to launch a native checkout inside Gemini. According to the Adobe Digital Insights report, web traffic sent by artificial intelligence tools to US retail sites has grown almost fivefold year on year.

But if you look at user behaviour, the reading is different. In a Gartner survey from May 2026, 31% would accept an AI narrowing their options in everyday categories, but only 11% would let it make the purchase decision. Their analyst Kate Muhl sums it up: consumers “are not looking to outsource their purchase decisions to AI; they want it to help them find better information, compare prices and narrow options, while keeping control of the final decision themselves”. People delegate searching and researching a product, not the act of buying.

That gap between what the industry is building and what people actually do is, to me, the most important data point in the whole phenomenon. And it is precisely the mistake badly done foresight makes: confusing maturity on the supply side with maturity on the demand side, because the former generates far more press coverage than the latter. Vendors selling AI optimisation have an obvious commercial interest in amplifying the story.

What is interesting is the shift in browsing habits it produces, which affects the user experience.

“AI shopping is changing how people discover products, but not consumer trust.”

The funnel breaks and fragments further: AI pushes us from “discover → research → buy” to “research → discover → buy”, says Melissa Minkow, global director of retail strategy and market analysis at CI&T. Delegated transactions remain held back, not by technology, but by who authorises the spending.

That is why, as things stand, treating agentic e-commerce as a trend is less interesting than treating it as an uncertainty. Trends get planned for; uncertainties get staged. The value of asking good questions comes in here too: it is less about “when will the agent buy?” and more about sharpening the question: “what should a brand's relationship look like when discovery happens inside a conversation it does not control?” or “which part of the buying process will users refuse to delegate, and why?”. It is from those uncertainties that we build future scenarios.

How do we work on this at Interactius?

We see foresight design as a direct tool for making better decisions in order to arrive well in the future. Which is why working on it means bringing the business view in from the start.

Our model rests on building a signals radar that runs across a full year: a continuous scan of news, innovations, posts, videos and articles in which we track the environment around the topic under study through six systemic lenses (political, economic, social, technological, environmental and legal), since we start from the premise that no signal changes in isolation. We classify every source by its epistemological level: from mainstream media, which only publish once something has reached critical mass, to cultural periphery (communities, networks, adjacent markets) with less consolidation but far greater anticipatory value.

To group hundreds of signals week by week, we lean on AI. That in turn carries a real risk: cognitive offloading. When the output arrives tidy and wearing the appearance of analysis, the temptation not to review it is enormous. But those hours of friction with the information and the material are necessary: they are exactly what builds judgement. Otherwise we are feeding homogeneous thinking in a discipline whose entire value depends on the opposite.

That is why our method is built from phases designed by humans for humans, with contact points between teams chosen deliberately to look after judgement in agentic times:

1. Setting up the radar (purpose framing). We understand the context, the business vision and the strategy in play. We sit down with the team involved to focus the topic under study, the audience, the core markets, the inspirational markets and the time horizon.
2. Continuous scanning (scanning the horizon). We switch on the radar and start collecting signals systemically. The radar does not stop while the project lives.
3. Analysis (sense making). Each week we group the signals coming off the radar, cross-reference between lenses and assess how far the detected phenomena have matured — that is, whether a phenomenon is a weak signal, an emerging one or a trend. For this we rely on a matrix and on criteria based on the methodology proposed by the French economist and foresight expert Michel Godet and the futurist Elina Hiltunen. All of it is shared quarterly in deliverables designed to be easy to consume, so the team involved stays current with everything the radar is picking up.
4. From trend to action. Over the course of the year, the accumulated material is translated into three or four scenarios with their leading indicators and triggering events, and into concrete strategic initiatives for each scenario, through a backcasting session: a strategic planning method that starts from a desired future and works backwards to define the steps and actions needed in the present to reach it.

How do we get ready for the agentic era?

We have moved from VUCA times (volatility, uncertainty, complexity and ambiguity) to the BANI era: brittleness, anxiety, non-linearity and incomprehensibility. All of which demands rather more than simply adapting.

And yet only 30% of organisations use foresight tools, even though 53% acknowledge their value according to McKinsey & Company. The gap is not defined by scepticism but by calendars: Fast Company explains in this article that 52% of leaders describe their work as chaotic and fragmented. Thinking three years ahead demands an attention that busyness destroys. And here lies the trap of the moment: as AI accelerates operational work, we risk filling the time we gain with more tasks that carry no strategic judgement.

Getting ready is not about having an opinion on AI. It is about having a method: a system flexible enough to force you to look at your environment continuously, a calendar that protects the time to interpret it, and the discipline not to accept the first tidy answer a machine hands back.

That is what we humans can do to get ready for the agentic era. Train the gaze, follow a method and design strategies to arrive well in the future, whatever it turns out to be.

And what, exactly, is a signal of change?

“The future is already here. It is just not evenly distributed.”

As William Gibson used to say. Signals of change are precisely that: the first indications of a future still invisible to most.

The economist Pippa Malmgren, author of Signals (2016), illustrates it well with her concept of shrinkflation: when a toilet roll suddenly has fewer sheets but costs the same, it is not a whimsical marketing decision, it is a signal of deep economic strain and supply chains under pressure.

Several definitions strike me as especially illuminating. “A signal is a specific example of the future in the present. I like to say a signal is a clue that things could soon be different,” says the futurist Marina Gorbis. Ozcan Saritas and Jack E. Smith describe it as “the first indications of possible but unconfirmed changes that may later become more significant indicators of critical forces for development, threats, business and innovation”. And there is a third, more operational way of putting it: a signal of change is a measurable, traceable, tangible fact, occurring in the present and pointing to something that could change in the future. A CEO's decision, a product launch, the adoption of a technology, a new patent. They happen infrequently, they are almost by definition “first times”, but they are already here, waiting for someone to connect them.

The Claude ad “There's hope in hard questions” is a good example of where this is heading. It makes an honest statement about the cultural shift we are living through and answers it from collaboration rather than substitution between people and technology. It can stir feeling or it can stir confusion. “Probably both at once”, depending on who you ask: a little Black Mirror.

But what is interesting is less the ad itself than what we do with it: how very human it is to rush to read the comments before forming our own opinion, before even letting the feeling of the video settle. Some time from now, once AI is completely normalised, we will probably go back to that same video repository — one we still call, with an almost archaeological fondness, YouTube — to look for who we were, so we can marvel at who we are now. We cannot help it. Remembering is very human: letting nostalgia in, feeling Sundays.

Entering this agentic era, I keep returning to the question that drives me as a researcher: what makes us humans so human? My bet is on the madness. The madness of buying toilet paper by the armful at the start of a global pandemic. The contagion and the grandeur of travelling somewhere remote just to see a total eclipse, or of teenagers meeting up to do aura farming for no reason other than seeking social connection. We are social creatures — a bit foolishly social, perhaps, but social — and that clumsiness is, for now, the one thing no machine quite knows how to fake.

Which is why it is no coincidence that what appeals right now is imperfection. In a world flooded with AI-generated content, brands are returning to old-school craft, to hand-drawn cartoonish illustration. Slower, more deliberate strokes that signal authenticity precisely because they cost more. Imperfection has become the last available brand asset. At the same time, and apparently unrelated, you can now get into an autonomous car in San Francisco without thinking twice — and soon, we are told, in Madrid too.

Each of these signals converges on a future, near or distant, and we will be able to make sense of them. We will understand why those signals were there, and how they affected not only the way we see the world but the way we buy. Strategic foresight can start detecting the behaviour of the future buyer in your digital commerce.

What if we teamed up to anticipate change?