The Obsolescence Trap: Why Long-Term Plans Kill Innovation

A Czech bank is approving an AI project right now. It will be finished in 2027. Sounds like a reasonable timeline — two years of development, proper stage-gate process, milestones, review boards.

The problem: nobody in that room knows what AI will look like in two years. Nobody knew in 2022 either, when large language models suddenly managed to generate code, analyze documents, and hold conversations at a level experts had predicted wouldn't arrive until the end of the decade. Or in 2024, when agentic models started autonomously browsing the web, making purchases in apps, filling out forms, and managing entire workflows without human intervention. Every six to nine months, a new breakthrough rewrites the rules. That project from the approval meeting will ship into a world we can't describe today.

And yet it's being decided as if the future were predictable. One budget. One timeline. One scenario.

This isn't a failure of any individual manager. It's a systemic problem with a name: the Obsolescence Trap. R&D outputs become morally or technologically obsolete before they reach the market. And in the age of AI, the trap is closing faster than ever.

Billions in the Bin — and Nobody's to Blame

Equinor, the Norwegian energy giant, cancelled its hydrogen infrastructure program in 2024. Billions of euros. The planning team in 2021 took EU political proclamations about a hydrogen future, ran a linear extrapolation, and built a business case on top of it. Nobody systematically verified whether real demand matched political promises. Nobody tracked the signals that customers weren't willing to pay a premium for blue hydrogen. When the regulatory environment and project economics shifted, there was no mechanism for a pivot. All that remained was a write-off and a strategic restart.

JP Morgan, HSBC, and other major banks jumped on the Metaverse in 2022–2023. Virtual branches, avatars for financial advice — a strategic bet on the future of banking. PKO BP in Poland invested in a presence on Decentraland, Komerční banka in the Czech Republic experimented with similar concepts. Consumer adoption never materialized. The technology was immature, the user experience dismal. Most projects were quietly shelved. What was missing was rigorous technology foresight that could distinguish a speculative bubble from a genuine shift in behavior.

Banks across CEE invested heavily in PSD2 compliance and open banking platforms between 2018 and 2020. The PSD3 draft now changes requirements so significantly that an estimated 60% of these platforms are incompatible with the new regulation. The development cycle outlasted the regulatory horizon. Companies built technically flawless solutions for a world that no longer exists.

The common denominator across all three cases? Linear planning based on a single scenario of the future. Ignoring weak signals during development. No mechanism to redirect a project when its assumptions change. And sunk cost bias — the more money and time poured into a project, the harder it is to stop, even when everyone in the meeting suspects it won't work.

The Stage-Gate Model and Its Blind Spot

Stage-Gate was born in the 1980s for the automotive and pharmaceutical industries. Discover, Scoping, Business Case, Development, Testing, Launch. At each "gate," the decision is made — GO or KILL. Elegant. Proven. It worked for decades.

It worked in an environment where technology changed slowly, regulation was stable, and competition was predictable. A three-to-five-year development cycle wasn't a problem because the world at the end looked more or less like the world at the beginning.

Today? The average AI product development in banking takes 18–24 months. A new major release of a foundation model arrives every six to nine months. Regulation mutates in real time — AI Act, PSD3, NIS2, ESG reporting. A product that clears every gate can be obsolete on launch day.

And it's not just AI. Telecom companies invested billions in 5G networks but, with R&D spending around 1% of revenue, failed to develop the services and applications that would monetize the infrastructure. They built the pipes, but others — Google, AWS, tech hyperscalers — create the content and value inside them. Energy companies plan decarbonization on 10–15 year horizons while the EU regulatory environment shifts every two to three years. Pharma develops drugs over ten to twelve years while AI reshapes therapeutic standards in months.

The data from practice confirms this harshly: up to 80% of digital transformation and AI projects in enterprises fail to deliver a tangible impact on profit. In banking and telecom, projects chronically stall at the Proof of Concept stage — a phenomenon known as "PoC purgatory." Teams systematically underestimate development time (by an average of 40%), underestimate costs (by 30%), and ignore external risk factors.

Why? Because Stage-Gate doesn't address the question "what if the world changes." It only addresses "did we hit our internal milestones." And in doing so, it creates an illusion of control while the real risk — the obsolescence of assumptions — quietly grows in the background.

The Specific Problem of CEE

Central and Eastern Europe adds its own layer of dysfunction.

EU funds require fixed milestones, fixed budgets, fixed outputs. Companies therefore design projects backward — from grant criteria, not from market fit. The incentive isn't to create an innovation that succeeds in the market. The incentive is to meet reporting requirements and receive the next tranche. The result? Innovation centers that are built, equipped, and empty.

The data reveals a structural issue: the share of R&D financed by the business sector in CEE averages 43.5%. The EU average is 57.7%. The region's innovation agenda remains disproportionately dependent on public funds, which systematically favors projects with clean paperwork over projects with real market impact.

Most large R&D operations in CEE also report to foreign headquarters. Komerční banka to Société Générale. Škoda Auto to the Volkswagen Group. The decision horizon is set in Paris or Wolfsburg. Local teams are executors, not strategists. The lag between a local insight and HQ approval? Six to twelve months. In the age of AI, that's an eternity.

And yet the Czech Republic is no technological laggard. 48% of Czech firms use generative AI — above the EU average of 37%. Czech companies plan to prioritize innovation at double the European average rate over the next three years. The problem isn't the people or the level of technological sophistication. The problem is a decision-making system designed for a stable world.

What Works Differently

The problem isn't long-term thinking. The problem is a long-term fixed plan. The distinction is crucial.

An adaptive approach to R&D means having a clear vision of where you want to be in ten years — but short-term flexibility in how you get there. Not locking into one roadmap, but responding to what's happening outside.

In practice: every R&D project should be tested against multiple future scenarios, not just one. Take the two key uncertainties of your project and combine their extreme states. You get four distinct futures — four worlds with their own internal logic. A project that survives in only one of them is a gamble. A project that works in three out of four is robust.

This isn't an academic exercise. Both BCG and Deloitte document the shift toward systematic scenario planning as a standard component of innovation portfolio management. Projects are classified into three categories: "robust bets" — they work across all scenarios, like process digitization. "Contingent bets" — they work only in some, like small modular reactors under a high carbon price. And "hedges" — projects that protect against the worst-case scenario.

Siemens applied this in practice. Through systematic identification of weak signals in urbanization and demographics, it redirected R&D toward Smart Infrastructure long before the competition. In 2023, that division became the largest contributor to revenue growth. The company didn't wait for the trend to become obvious to everyone — it caught the signals while others were still ignoring them.

The key mechanism is continuous monitoring. The traditional approach works with quarterly strategy reviews — data from the last quarter, decisions implemented in the next. Total lag: six to nine months. In an adaptive model, you track key dimensions continuously. Technological breakthroughs, regulatory changes, competitor moves, weak risk signals.

Picture it in practice. Your company is developing an AI product for banking. You have four scenarios based on two uncertainties: how strict EU AI regulation will be and how fast the technology advances. In January, the "permissive regulation" scenario sits at 60%. In February, a restrictive AI Act draft leaks. The probability drops to 40%. You have a predefined trigger: if it crosses the threshold, you pivot the architecture. You don't find out six months later at a board meeting. You find out in six days.

This isn't a minor process improvement. It's a different operating system. Stage-Gate says: if we hit our internal milestones, we move forward. The adaptive model says: if scenario X crosses probability Y, we change course. Decisions are driven by changes in the outside world, not the internal roadmap.

And the results speak clearly. Companies that systematically work with multiple scenarios achieve twice the growth rate of their industry average. In pharma, companies with adaptive portfolios save an average of $1.2 billion per project stopped in time — because they kill it before pouring in even more. It's not about being smarter. It's about being faster to admit that assumptions have changed.

Trends, Uncertainties, and Where Real Decisions Are Made

Every company's strategic plan rests on assumptions. Most never name them explicitly. And that's precisely where the risk lives.

Some assumptions are trends — they have a clear direction. Aging populations, digitization, urbanization. These won't surprise us. We can prepare for them linearly.

But alongside trends, there are uncertainties — factors whose future trajectory is critical for the company, but nobody knows which way they'll go. How will AI regulation evolve in Europe? Will a global tech player enter our market? Will customer behavior shift faster than we can respond?

These uncertainties form the axes around which scenarios are built. When you combine the extreme states of the two most critical uncertainties in your industry, you get four distinct futures. None of them is a prediction. Each is a consistent possibility you can prepare for.

And then come the questions with real strategic impact: What capabilities do we need to succeed in each of these worlds? Are there "no-regrets moves" — actions that make sense regardless of which scenario materializes? And are there moves we prepare in the drawer, ready to deploy if the world heads in a particular direction?

For that, you need indicators — measurable signals that continuously show which scenario reality is approaching. Not once a year at a board meeting. Continuously. Strategy then stops being a document pulled from a drawer at the annual conference and becomes a living system that evolves alongside the world around it.

The Future Is Not Fate

The future is not fate. Nor is it an equation you can solve with enough data. It's a space we navigate through today's decisions.

A company working with a single plan is playing roulette. A company working with multiple scenarios holds a portfolio of options. It doesn't need to guess which future will materialize. It only needs to ensure that none of them catches it unprepared.

The world won't slow down. AI will keep compressing innovation cycles. Regulation will keep mutating. New players will keep entering markets at speeds we're not accustomed to. The question isn't whether something will surprise you. The question is whether you'll be ready.

The first step is simple: look at your strategic plan and ask — how many of its assumptions still hold today?