Consultant Itay Sagie: Customer Workflows Are Becoming the Moat for AI Companies

An AI company's durable advantage comes from embedding its product in the customer's workflows. That is the thesis technology consultant Itay Sagie puts forward in a column for Crunchbase News. The text contains no figures; it is a framework for decisions, not a study. But the framework is a sound one: in a market where model capabilities are converging quickly, the question "what protects you?" needs a different answer.

Why a model alone isn't enough

Almost anyone can now get strong language models through an API. A startup that looked like a technology leader yesterday finds itself facing a competitor tomorrow that runs on the same underlying engine and has a similar interface. Good answer quality no longer sets a product apart from its neighbor.

Sagie suggests looking elsewhere: where the product sits in the customer's working day. If it sits at a point that requests, contracts, tickets or code pass through, replacing it becomes a project with costs, risks and staff retraining. If it lives in a separate tab, the customer leaves with one click.

What he advises founders and investors

The author recommends that founders prioritize measurable customer dependence. The key word is "measurable": enthusiastic reviews and pilots do not count as dependence. You need to know what exactly would break for the customer if the product were switched off, and how often it is used.

For investors and company acquirers, Sagie proposes assessing three things: integrations, trusted relationships with customers and access to workflows. In his logic, these strengthen retention and growth. For M&A deals this is a practical filter: buyers purchase not only code but also a place in the customer's processes.

The author does not offer a single set of metrics, so below is our own editorial list of signals worth checking:

• the share of daily active users among those who paid for licenses;
• the number of customer systems connected to the product (CRM, ERP, trackers, data warehouses);
• the share of critical operations that run through the product;
• the time and cost of migrating to a competitor, as estimated by the customers themselves;
• the trend in contract expansion within companies that are already connected;
• the depth of accumulated context: data history, settings and rules that a newcomer does not have.

Where the idea is weak

Embeddedness works only as long as the customer gets value. If a product holds a customer only through the difficulty of leaving, that is deferred churn: sooner or later a migration tool or a stronger competitor will appear. Trust is also an asset here, not a formality: a company that gets access to sensitive processes and data is responsible for their security.

There is also a structural risk. Deep integration often relies on other companies' platforms. The owner of a CRM or cloud environment can change API terms, raise prices or release its own AI feature. A startup whose access to workflows depends entirely on a single partner is building a moat on someone else's land.

Practical takeaways

A founder should decide in advance which of the customer's working operations he wants to occupy, and track dependence as regularly as revenue. It helps to ask: what would the customer lose if the product were switched off tomorrow? If the answer is "nothing significant," it is time to change the roadmap.

During due diligence, an investor should request usage data by integration, not just revenue and license counts. An acquirer should check who owns access to the workflows: the company itself or an intermediary platform.

And all parties should remember that Sagie's argument remains a consultant's thesis. Its value has to be tested against a specific company's data: retention, contract expansion and the real cost of leaving.