STKR News0 of 3 free this monthAI is becoming a commodity. The real value for founders isn't the model itself, but how deeply they can wedge their software into the daily grind of their customers' operations.
Originally on Crunchbase News →Adrian Boysel
Contributor
Oct 9, 2026
5 min read
Photo illustration / STKR News
We have reached the point in the AI cycle where the novelty is officially dead. If your entire pitch is that you have a faster LLM or a cleaner interface for a chatbot, you do not have a business; you have a feature that will be swallowed by a platform or undercut by an open-source model within six months. The technical barrier to entry for AI is dropping to near zero. When everyone has access to the same intelligence, the intelligence itself ceases to be a competitive advantage.
I talk to founders every week who are obsessed with their tech stack. They want to show me how they optimized their inference costs or how they fine-tuned a specific model for a niche use case. That is all fine, but it is not a moat. A moat is something that makes it painful for a customer to leave you. In the current landscape, the only durable moat left is the workflow.
For the last two years, the industry has been focused on the power of the model. We saw massive rounds for companies that were essentially just wrappers around GPT-4. These companies grew fast because they were first to market, but they are now seeing churn rates that would make a traditional SaaS founder lose sleep. Why? Because they were built on top of shifting sand. As soon as a bigger player integrated that same AI capability into an existing tool, the standalone startup lost its reason to exist.
If you are building in AI today, you have to assume that the underlying technology will become a commodity. You have to assume that Google, Microsoft, and Amazon will offer whatever clever trick your model does for free, or at least bundled into a subscription the customer already pays for. To survive this, you have to move past the output and start looking at the process.
A workflow moat is created when your tool is no longer just a place to get an answer, but the place where the work actually happens. It is the difference between a tool that writes an email and a tool that manages the entire sales sequence, updates the CRM, and triggers the next follow-up based on real-world signals.
When you embed your product into a customer's essential daily tasks, you are creating a form of operational debt. If the customer wants to switch to a competitor, they aren't just switching models; they are breaking their internal systems. They have to retrain staff, rebuild integrations, and risk data loss. That friction is what protects your margins.
Founders often treat integrations as a checkbox for the sales team. This is a mistake. Integrations are the anchors that keep your product from drifting away. The more data sources you pull from and the more downstream systems you feed, the more indispensable you become. If you are building for the enterprise, you should be thinking about how to become the connective tissue between their legacy databases and their modern front-end tools.
Trusted relationships in this space are not built over steak dinners; they are built through reliable data pipes. If a company trusts your AI to handle a sensitive workflow without human supervision, you have achieved a level of stickiness that no raw performance benchmark can touch.
How do you know if you actually have a moat? You need to look at measurable dependence. If your AI tool went offline for 24 hours, would the customer's business stop? If the answer is no, you are a luxury. If the answer is yes, you are a utility.
Investors and acquirers are starting to catch on to this. The days of being valued on a multiple of your API usage are over. They are looking for retention metrics that prove the software is deeply rooted. They want to see that users are spending hours inside the platform, not just popping in to generate a single asset and leaving. They want to see that your tool has become the 'system of record' for whatever problem you are solving.
The danger for many founders is building what I call a 'thin layer' application. This is a tool that sits on top of a workflow but doesn't actually own it. For example, a browser extension that summarizes meetings is a thin layer. It is useful, but it is easily replaced by the platform itself (like Zoom or Teams) or by a competitor with a lower price point. To move from a thin layer to a thick moat, you have to own the data that the meeting generates and use it to drive the next step in the business process.
If you are currently building an AI-first company, you need to pivot your focus from 'what can this AI do' to 'where does this AI live'. Here are the three areas where you should be spending your time:
I have seen plenty of 'technological breakthroughs' fizzle out because the founders didn't understand the boredom of business. Most businesses don't want the most advanced AI in the world; they want the tool that makes their lives easier and their employees more productive. They want reliability and integration. They want to know that if they invest the time to set up a workflow, it will work every time.
Stop trying to win the arms race of model parameters. You will lose that fight to the big tech giants who have more compute and more data than you ever will. Instead, win the war of utility. Find a boring, repetitive, essential process that a company does every day, and own it so completely that they can't imagine doing it any other way.
The value of AI is not in the generation of content, but in the elimination of friction within a business process.
In the end, the winners of the AI era won't be the ones with the smartest machines. They will be the ones who managed to weave their software into the fabric of how companies operate. If you can't be the brain, be the nervous system. The brain is replaceable; the nervous system is not.
Read the original at Crunchbase News →
A six-minute morning dispatch on the markets and the technology shaping them.
Free. No spam. Unsubscribe anytime.
Earn $STKR for published stories on markets, protocols, and culture.
Comments
24 reader responses
