We Are Running Companies on Chat Windows and Calling It a Revolution

Analysis by Iaroslav Belkin, founder of Belkin Marketing, who looks at the challenges of running companies these days. Sources: Microsoft Work Trend Index 2026, Accenture 2026 enterprise AI data, ManpowerGroup 202,6 Global Talent Barometer, HBR/BetterUp Labs/Stanford workslop research, Faros AI Productivity Paradox Report 2025, Product Talk context rot analysis, Markswebb hybrid UI research 2026.

Here is the thought experiment. You hire the most capable consultant you have ever worked with. Reads everything. Thinks fast. Excellent synthesis. The catch: complete amnesia every time they leave the room. Every morning, you re-explain the company, the decisions already made, the things you tried that did not work. Every document you built together vanishes when you close the window.

Running Companies on Chat Windows

That is the state of AI in most organisations in 2026.

Only 32% of organisations have achieved sustained, enterprise-wide AI impact despite 86% of C-suite leaders increasing AI investment, per Accenture’s 2026 enterprise data. The gap between “we use AI” and “AI is making our business measurably better” is the largest it has ever been. In theory, every organisation is now “AI-first.” In practice, its memory of the organisation is a markdown file someone pastes into a chat window and hopes is still current.

The dominant AI interface — the chat window — was designed for a single user having a single conversation at a single moment in time. It was not designed for teams, persistent institutional knowledge, audit trails, or scale. We took the most powerful cognitive tool in human history and gave it the interface of a 2009 messenger application.

The failure modes are structural, not accidental.

Product Talk’s February 2026 analysis of context rot documents that AI systems do not merely lose context between sessions — they degrade within sessions as conversations grow longer, favouring tokens at the beginning and end of input while ignoring the middle. The longer the conversation, the worse the model’s attention to the context you painstakingly accumulated. Users learn to start fresh constantly, which means losing everything built in the previous session.

The workaround that has emerged: the markdown file. A growing document containing everything the AI needs to know about your company, pasted at the top of every new conversation. It works until it contains contradictions nobody noticed, outdated decisions sitting next to current priorities, and no version history a non-technical person can operate. A markdown file is not a knowledge management system. It is the absence of one.

HBR published research in September 2025 on what BetterUp Labs and Stanford called “workslop” — AI-generated output that looks complete but requires significant rework, averaging nearly two hours of rework per instance across 41% of workers surveyed. The reason it spreads: there is no record of how the output was produced. Recipients cannot assess whether it came from a thoughtful process or from someone who typed “write me a report” and sent back the first result.

ManpowerGroup’s 2026 Global Talent Barometer, across 14,000 workers in 19 countries, found regular AI use increased 13% in 2025 while confidence in the technology’s utility fell 18%. People are using it more. They trust it less. Individual wins are not compounding into organisational capability — because the interface was never built for organisational use.

On 14 June 2026, Satya Nadella named the strategic stakes in language a board understands. The model you choose is not your competitive advantage. What compounds instead is the learning loop — your workflows, institutional knowledge, and accumulated judgement, captured as traces that make your AI measurably better over time. He called it “the new IP of the firm.” Own that loop and you can swap the underlying model without losing the expertise built on top of it. Fail to, and the value of everything your people know flows to whoever owns the model.

Big tech has noticed. Four days after Nadella spoke, Perplexity launched Brain, a memory system that builds a living context graph of a user’s work and improves overnight, explicitly designed to remember what worked and what failed rather than just preferences. It is the clearest signal yet that the platforms understand exactly where the value sits. But the fundamental problem survives the upgrade. That memory still lives in their cloud, tied to their model. Your knowledge becomes portable everywhere except out — locked in someone else’s infrastructure, locked to a single AI. Open, model-agnostic standards are starting to change that: engrams, for example, store what your AI learns as a portable, inspectable format that lives with you rather than the vendor, and travels across models. The question is no longer whether AI should remember. It is who owns the memory.

“Every conversation with a chatbot is your knowledge leaving the building and landing in someone else’s cloud, one chat at a time. The model gets smarter. You don’t. And a privacy policy won’t save you because policies change overnight. Architecture doesn’t. The only real moat is intelligence you own: memory that lives where you do, that no vendor can read, revoke, or hold hostage. Own your intelligence, or rent it forever.” (Gregor Žavcer, Swarm Foundation, Plur)

Now look at the failure modes through that lens. Every one is a hole in the learning loop. The amnesiac interface throws away the trace before it can compound. The markdown file is a learning loop maintained by hand, badly. Copy-paste-and-pray leaves no trace to learn from. Fragmented memory across tools means the loop never compounds.

The chat window is not merely inconvenient UX. It is how organisations are prevented from accumulating the one asset that becomes a moat.

Two years into the most consequential technology shift of our careers, most organisations have no learning loop to show for it. The revolution was real. The interface was not built for it.

Iaros Belkin is a founder of Belkin Marketing, a boutique agency serving as Strategic Advisor to Deep Tech, Web3 and AI Founders. Two decades of experience navigating high-stakes global markets and orchestrating everything from grants and key partnerships to VVIP events elevated experience.

See more breaking stories here.