Why Your Strategy is Dying in a Silo (and How AI Can Resurrect It).

The Illusion of "Data-Driven"

In 2026, we are drowning in information but starving for synthesis.

Most organisations claim to be "data-driven". They have dashboards that glow with real-time metrics and research repositories overflowing with colourful PDFs. Yet despite this abundance of evidence, they continue to build products nobody wants and launch disconnected campaigns that nobody hears or understands.

The problem isn't a lack of insight. It’s an insight-horading problem.

In a typical tech organisation, insights are scattered across different departments and postcodes. Research lives in Product. NPS surveys sit in Marketing. Effort scores are locked in Support. Trend scouting falls under Business Development. Each region conducts its own research. 

They are all looking at the same customer, but they are standing in different rooms with the doors bolted shut.

Strategy, at its heart, is the art of connecting the dots. But if your dots are siloed by department, you aren't strategising; you’re merely making assumptions based on partial views. In the AI Era, "Insights Hooarding" isn't just a nuisance—it is a commercial liability.

The Speed vs. Accuracy Trap

I once worked with a flight compensation scale-up where the "Insights Hoarding" problem nearly cost them their conversion rate.

A well-intentioned Product Manager spent weeks planning and building a "predictive" flight selection feature. The goal was simple: speed up the claim submission journey. To simplify the user flow, he believed he could use a predictive dropdown for airline selection, which would pull journey information from a travel API once he knew the traveller's day, time, and flight company. "Time to Complete" was the north star on his dashboard. He prioritised speed above all else.

If the feature had launched. The "Time to Complete" metric would have looked fantastic, but claim rejection rates would have skyrocketed.

Why? Because the PM failed to consider familiarity.

If he had talked to User Research and the Support team in advance of ideating the feature, he would have discovered that users didn't need a faster dropdown. They needed better preparation. When the feature was tested- not out of choice, but out of intervention- users selected the airline they felt most familiar with, rather than the one they actually flew with. The User Research team knew the solution: users needed a "Before You Begin" checklist to collect all necessary travel documents, supported by an "I'm Ready to Begin" button before entering the flow, rather than relying on memory or creating reasons not to check their travel itinerary and boarding pass. There may have been a bounce at the beginning, but that would be better than a mid-flow drop off because a user cannot answer a question, or an inaccurate submission that delays or inaccurately rejects their claim.

The PM was looking at a single step in a vacuum. Support was looking at the wreckage at the end of the road. User Research was in the middle, but left out of feature planning sessions. Because their data never converged, the company spent hours building a feature that made it easier for customers to fail.

This is the forensic reality of silos: when you prioritise a departmental metric over a cross-functional truth, you optimise for a part and break the whole.

The Obsolescence Trap

The "Insights hoarding" problem is often a symptom of bad architecture. But sometimes, it can be a symptom of something much more human: fear.

In the hyper-growth world of financial services, I witnessed a different kind of silo—one built not on technical barriers but on professional gatekeeping. Researchers were embedded within product teams, yet the insights they generated were treated as guarded secrets.

I watched as research teams held weekly project shares that were strictly "off-limits" to designers, product leads and marketers. The justification? Usually, it was a vague concern about "maintaining the integrity of the methodology." The reality? It was a fear of obsolescence. There was a palpable anxiety that if research insights were democratised—if designers were allowed to peek under the hood—that research would no longer be needed.

The result of this gatekeeping was a series of scattered puzzle pieces that refused to form a cohesive picture.

Marketing was launching high-spend campaigns without knowing which customer pain points were truly intrinsic. Meanwhile, Product was building features without understanding which messaging hierarchies should influence screen copy. Because the researchers refused to share the stage, the organisation was essentially flying blind in two directions. Marketing was selling a dream that the product teams hadn't quite realised, and Product was solving problems that Marketing didn't know how to articulate.

In the AI Era, this hoarding of insight is a death sentence. When you treat information as a finite resource to be protected rather than a multiplier to be shared, you create a bottleneck.

The lesson is simple: An insight is only valuable if it is actionable. If you lock your findings in a room that your designers and marketers can't enter, you haven't "protected" the research; you have simply ensured its irrelevance. Your team doesn't need a gatekeeper; they need a librarian who knows how to turn on the lights.

Institutional Amnesia

Agencies are in the business of selling wisdom. It is their primary currency. Yet, most suffer from a debilitating, expensive condition: Institutional Amnesia.

I have seen this pattern play out across the world’s most reputable agencies, who complete hundreds of projects a year. They perform deep-dive research into everything from commerce infrastructures to the psychographics of healthcare consumers. But the moment the final invoice is paid, that collective intelligence is buried in a digital graveyard. It is filed away in a folder named "Final_v2_USETHIS_FINAL," and the team—the actual holders of the context—moves on to the next assignment.

The cost of this amnesia is never more apparent than during a high-stakes pitch.

A request comes in for a specific case study—perhaps a platform transformation from two years ago. The Slack messages start flying. "Does anyone know where that deck is?" "I think Sarah had it, but she left the organisation six months ago." "Is it on SharePoint somewhere?"

It is a frantic, manual search for a needle in a haystack of PDFs and incomplete business cases. "Pitch-Panic" quickly escalates because of missing client information and inaccurate impact projections. Agencies claim to be the vanguard of efficiency and innovation, but they are often rife with structural waste. You are effectively paying your most expensive talent to re-learn what the organisation already knows.

Today, efficiency is no longer about how many hours you bill; it’s about your Time to Insight. The strategic formula for 2026 and beyond is simple: Efficiency = (Collective Knowledge / Time to Retrieve).

If your "Time to Retrieve" involves calling around and digging through dead folders, your efficiency is zero. This is where RAG (Retrieval-Augmented Generation) models has great potential to move from a "tech trend" to a survival requirement.

True "Pro" status in the next five years will be defined by the ability to move from a file-folder culture to a "Living Brain." It’s the shift from hoarding documents to chunking, tagging, and distributing intelligence so that a new hire can access twenty years of organisational wisdom in twenty seconds. The challenge shifts from “Do we have insights?” to “What tag should I use to find them?”

The Resolution: Building the Converged Insights Ecosystem

So, how do we fix the Insights Hoarding Problem? Well, we don’t just buy better software; we redesign the organisational map. To move from "file-folder" amnesia to a "Living Brain" ecosystem, you must build through three distinct phases.

💠 Phase 1: The Unified Taxonomy (Standardising the Language)

A RAG model is only as useful as its index. If Marketing calls a user friction "Brand Detraction" while Support calls it a "Level 2 Ticket," your AI will never connect the dots. Extensive alignment is needed to build a global taxonomy for every business unit and communicate it so that everyone is forced to tag deliverables in the same way and upload projects to the same location. This is the lesson we steal from the Goliaths: the gold is in the metadata.

💠 Phase 2: The RAG Architecture (The Engine)

Once the language is unified, we move the data from static graveyards (SharePoint, Google Drive, Confluence) into a Retrieval-Augmented Generation (RAG) architecture. This isn't just "Search 2.0." It is a system that "chunks" supports logs, and pitch decks into a vector database. It allows any employee to ask: "What do we know about X customer friction for X industry in France?" and receive a synthesised answer that draws from a variety of research papers, real-time industry user cases, and a market audit from last week.

💠 Phase 3: The "Truth Bomb" Workflow (The Circulation)

The final hurdle is cultural. Insights are like fresh produce; they have a shelf life. A converged ecosystem doesn’t wait for someone to ask a question; it pushes the truth to the people who need it. This involves "Truth Bomb" automation—setting up triggers that automatically push critical qualitative feedback into dedicated Slack channels or Marketing brainstorming boards. You want to move the insight from a destination to a constant background hum.

The Insights Strategy Boardroom Mandate

In the boardroom of 2026, the most valuable asset won’t be the size of your database, but the speed of your synthesis.

Strategy is not a static document that sits on a shelf (or a server) gathering digital dust. Strategy is an organisational behaviour. It is the ability to pivot when a new insight reveals itself. If your researchers are still gatekeeping, if your PMs are still dashboard-blind, and if your agencies are still charging you a premium to re-learn something you already paid for, you don't have a tech problem. You have a leadership problem.

The mandate for the C-suite is clear: While building an AI that can find your data is necessary, it is useless if that data is worthless. Focus first on building the organisation that makes your data worth finding. Break the silos, unify the taxonomy, and turn on the sun. This foundational work is what truly allows an efficient AI-powered system to succeed.

Your customer doesn't see your departments; they only see the friction. It’s time your organisation saw the same and built a culture and data foundation that AI can actually leverage.