Raffi Tchakmakjian: In the AI world, you can’t build just one “fortress” — you need a whole defense system

In the era of the Fourth Industrial Revolution, survival depends not on a single innovation but on a comprehensive strategy. This was the key message shared by Raffi Tchakmakjian — a CEO with three successful exits in SaaS and AI — during his presentation at DigiConf 2025 in Armenia.

Tchakmakjianbegan by explaining the concept of a “defensible moat” — a competitive advantage that is difficult to copy and grows stronger over time. Drawing on his experience advising startups in advanced technology sectors, including AI and quantum computing, he shared practical insights on how companies can protect themselves and thrive in today’s fast-moving AI landscape.

The Fragility of the AI Market: “Every Two or Three Months, There’s a New Threat”

According to Tchakmakjian, differentiation in AI is fleeting. Someone might invent a new feature — such as video generation — and within a week or two, competitors will have it as well. He pointed out the breakneck speed of innovation and the constantly shifting rules of the game. In the past, startups gained an edge over slow-moving giants like Microsoft; now, he said, “every two or three months brings new existential threats.”

He gave an example from one of the companies where he serves on the board: “On Monday, as I boarded a flight from Montreal, the CEO called me in a panic — OpenAI had just launched Agent Kit, an automation tool. And there we were again — spending the whole week trying to figure out how to respond.”

Even product–market fit, which Tchakmakjiandescribed as his least favorite part of the business, has become unstable. “You find the perfect balance of product, pricing, and marketing — and then someone pulls the rug out from under you with a new feature. It’s an endless treadmill.”

The Five “Defensive Moats” for Survival — From Speed to Taste

Tchakmakjian outlined five key “moats” that help AI companies stand out. Technology alone, he said, no longer guarantees success; what matters is how effectively a company uses it to grow and retain customers.

1. Speed and Momentum.
   Speed, he emphasized, is the most fundamental moat — but it’s not just about computational power. The true differentiator is the speed at which a company learns, experiments, and improves its systems. The faster an organization tests and iterates its models, the stronger its products become. Just as importantly, fast deployment and customer acquisition create a flywheel effect, where growth reinforces growth. He cited AI-native companies that outpace massive ERP systems like SAP by offering automation in a week instead of two years. “Speed isn’t just a strategy — it’s a real moat in the AI world,” he said.
2. Memory Personalization.
   Another vital moat is the intelligent use of memory. If a platform remembers user preferences, workflows, and organizational context, it becomes a “home” for the user — switching to a competitor becomes painful because the new system must be retrained from scratch. Tchakmakjianrecommended investing in context engineering rather than limiting efforts to prompt engineering, allowing AI to learn from interaction history and become more useful over time.
3. Workflows as a System of Record.
   He identified workflows as another key moat. When AI systems create new workflows, they generate behavioral data — the “DNA” of a company or user. By owning and analyzing these data, a company effectively becomes the system of record, making smarter decisions based on insights users themselves may not even realize. He cited examples such as Notion, which became sticky through innovative workflows and data, and Figma, which conquered the design market through collaboration that generated unique datasets. “Whoever controls the workflows and data controls the user — and that’s what makes a product truly indispensable,” he said.
4. Distribution.
   Tchakmakjianstressed that because AI innovation cycles are short, no company can rely solely on technology. Growth comes from partnerships, integration into customer workflows, and strong go-to-market strategies. “In this world, you need a strong bias toward distribution to build your moat,” he explained.
5. Judgment and Taste.
   Finally, he highlighted human judgment and taste as differentiators in a crowded AI market. AI produces a flood of content, much of it low quality, but with human curation and thoughtful design, products can become genuinely appealing. He cited Webflow, which combines engineering precision with smart partnerships, and Tenweb, which releases localized products tailored to user preferences. “Teach AI to have taste — and you’ll win the user’s heart,” he said.

The Future of Defensive Moats: A Composite and Dynamic Approach

Tchakmakjianalso discussed two emerging types of moats gaining traction.

Private Reproducibility — the assurance of accuracy and consistency, especially valuable in fields like law and medicine where mistakes are costly.

Technology Agnosticism — focusing not on the underlying models but on workflows and user experience. “It doesn’t matter which LLM you use — private, cloud-based, or open-source,” he explained. “What matters is owning the workflow. Technologies are interchangeable.”

His conclusion was clear: successful companies build composite moats — combinations of mechanisms that reinforce each other and create self-sustaining growth loops. “It’s a dynamic world,” he said. “Constantly reassess, adapt, or rebuild your moats.”

He added that these moats, like the AI landscape itself, now change every two to three months — a reminder that in the new industrial era, defense is no longer a structure but a living system.