How Intercom rose from the ashes by betting everything on AI | Eoghan McCabe (founder and CEO)

Eoghan McCabe presents Intercom’s AI turn as a response to both external disruption and internal drift. The company had become a late-stage SaaS business with diluted strategy, weaker energy, and low-single-digit revenue growth after McCabe left the CEO role during a prolonged illness. Net new ARR declined for five consecutive quarters and was approaching zero, which would have meant negative growth. He returned to the company, and ChatGPT was announced about a month later. That timing mattered, but McCabe rejects a neat origin story: the transformation was also shaped by frustration with his own compromises, anger about how the company was being run, and a desire to restore momentum.

Intercom was unusually prepared to test the new technology. It already had AI engineers, machine-learning systems for customer-service questions, 30,000 paying customers, hundreds of thousands of active users, millions of end users, and billions of data points. Within six weeks of GPT-3.5’s launch, the company had an early Fin beta. McCabe says the company’s willingness to move so hard was partly bravery, but also necessity: Intercom was already in a difficult position and therefore had less to protect than healthier late-stage companies. The interview describes Intercom’s trajectory and reported metrics, not a controlled comparison showing that every mature SaaS company should make the same bet or that an AI pivot alone caused the outcome.

The apparent contradiction in Intercom’s strategy was its earlier opposition to impersonal support bots. McCabe’s answer is that the relevant comparison is not an ideal human interaction against an AI interaction. It is a continuously available, consistent, fast, and expert agent against a customer waiting two, three, or four days for a poor canned reply. He acknowledges that this is partly a post-hoc interpretation and says AI also forces difficult questions about its effect on humanity. His claim is therefore a position grounded in the company’s experience, not a proof that automated service is always more personal or that customers never need people.

McCabe describes Fin as the clearest evidence that the new direction worked: growth above 300%, first-year ARR rising from $1 million to $12 million, a projection of more than $100 million ARR in under three quarters, leading customer and revenue positions in its category, strong performance benchmarks, direct-competition wins, and a number-one G2 rating. These are substantial claims, but they are the company’s own reported figures and competitive assessments. The conversation gives no methodology for the benchmarking, no denominator for the conversion claims, and no independent verification of the rankings, so they should be read as an account of Intercom’s position rather than a general market benchmark.

The interview’s opening argument is therefore narrower than ‘every company should copy Intercom.’ A mature company with a large installed base, relevant data, existing AI talent, and a business already near negative growth had a particular set of reasons to take a concentrated risk. The early account also leaves important questions open: the episode does not independently verify the reported growth numbers, explain the comparison set behind the rankings, or show how much of the result came from the AI product versus the existing business and broader market conditions.