The Brochureware Moment

In 1998, a retailer built an impressive website that couldn't take an order. In 2025, companies are deploying AI with the same structural relationship to their business. The transition is faster this time, and less forgiving.

In 1998, a mid-sized British retailer with two hundred stores decided it needed to be on the internet. The board approved a budget of £400,000. A project manager was hired. An agency was contracted. A web designer was flown in from London. The website launched eight months later, on time and under budget, with a homepage featuring the company logo, a store locator, and a section called “Our Heritage” that contained three paragraphs about the founder’s commitment to quality.

The CEO presented the website to the board. The board was impressed. The website was modern, professional, and reassuringly expensive. The press release described it as “a bold step into digital retail.” An analyst note mentioned the company positively in a roundup of retailers “embracing the internet.”

The website could not process an order. It could not check inventory. It could not tell a customer whether a product was in stock at her nearest store. It had no connection to any operational system inside the company: not the point-of-sale system, not the warehouse management system, not the supplier ordering system, not the customer database. It was a brochure. A very expensive, very modern-looking brochure, bolted on top of a business that had not changed in any way.

Nobody at the company considered this a problem. The website was the internet strategy. The internet strategy had been executed. The board could report to shareholders that the company was digitally engaged. The analyst could report to investors that the company was keeping pace with the sector. The project manager could report that the project was delivered. Everyone had produced the artefact they were measured on, and none of those artefacts required the website to do anything.

Eighteen months later, a competitor launched an online ordering system. The system was ugly. The design was functional rather than polished. The homepage did not feature the founder’s heritage or a high-resolution store photograph. The system connected to the warehouse, checked inventory in real time, processed payments, and shipped orders to the customer’s door. Within two years, the competitor had captured a significant share of the market in the categories where online ordering mattered. The brochureware company’s response was to redesign its website. The redesigned website still could not process an order.

The story feels distant. 1998 was a long time ago, and the lesson seems obvious in retrospect: of course a website that cannot take orders is not an e-commerce strategy. Of course bolting a brochure on top of an unchanged business is not a transformation. The companies that won, Amazon, Google, and the early online banks, did not have better brochures. They built structure: systems where the technology connected to the actual operations, organisations where the people who understood the technology had the authority to change the business.

In 2024 and 2025, software-dependent corporates across Europe and North America are deploying AI tools with the same structural relationship to their business that the 1998 brochure bore to the retailer’s operations. Copilot generates code. Chatbots handle customer queries. AI assistants produce strategy documents, status reports, and board packs. The tools are modern, impressive, and reassuringly expensive.

The AI cannot reconcile the strategy document against the code, because nobody has described the relationship between them in a form a machine can read. The AI cannot tell the CEO whether the architecture model matches what is actually deployed, because the architecture model is a slide deck and the deployment manifests are managed by a team the architects have never met. The AI cannot trace a budget decision to its operational consequences, because the financial model prices departments while value is created at process boundaries, and nobody has mapped the processes.

The AI is bolted on. It makes the existing dysfunction faster, more polished, and more confident. Strategy decks that previously took weeks to produce are now generated in hours. Status reports that previously required a programme manager to assemble are now generated automatically from Jira data that has no verified relationship to what is actually happening in the system. Board packs that previously contained carefully curated narrative now contain AI-curated narrative, which is the same fiction produced at lower cost and higher volume.

The organisations deploying AI this way are not doing something wrong in the narrow sense. They are doing exactly what the brochureware retailers did in 1998: they are adopting the technology of the moment, using it to produce the artefacts their governance requires, and calling the result transformation. The technology is real. The adoption is real. The connection to the actual business is absent, and nobody whose job depends on the current structure has an incentive to point this out.

The cost of being wrong

In 1999, the cost of being wrong was manageable. The brochureware retailer lost market share gradually. The board had time to notice, time to course-correct, time to hire a CTO who understood that the website needed to connect to the warehouse. The internet changed distribution, and distribution changes unfold over years. Some of the brochureware companies eventually caught up. Many did not, but the ones that failed had time to fail gracefully.

The AI transition is faster and less forgiving, for a reason that is structural rather than technological.

The internet changed what an organisation could sell and to whom. AI changes what an organisation can know about itself. An organisation that uses AI as a reader, comparing strategy against code, process definitions against system behaviour, budget decisions against operational consequences, accumulates structural knowledge at a rate that is qualitatively different from an organisation governing through quarterly narrative. The reader-path organisation detects a problem in minutes. The narrative-path organisation detects it in a quarterly review, if the problem survives the filtering that occurs between the system and the slide deck.

The gap between a reader-path organisation and a narrative-path organisation compounds at the rate of every reconciliation cycle the reader runs and the narrator does not. The reader-path competitor accumulates structural knowledge weekly. The narrative-path organisation updates its understanding quarterly, if the signal survives the filtering chain at all. That gap widens with every cycle. That does not mean every reader-path firm will win or every narrative-governed firm will lose. It means the organisation that learns faster from its own operations acquires an advantage that its competitor cannot close with presentation alone.

In 1999, the brochureware company could redesign its website and catch up. In 2025, the narrative-path organisation cannot catch up by deploying more AI, because more AI on top of an unchanged structure produces more narrative, not more knowledge. Catching up requires structural change, and by the time that need is obvious, the competitor that started earlier is already learning faster.

← The Separation The Strategy Deck That Ate the Company →