Eoghan McCabe on Intercom, AI Adoption and Freedom of Speech

Harry Stebbings asks Intercom co-founder and CEO Eoghan McCabe whether AI threatens a customer-support software company. McCabe says increasingly capable technology could eventually make today’s software obsolete, but believes adoption takes much longer than forecasts suggest. He expects Intercom to have years to build and adapt; that forecast reflects his role at the company.

McCabe says Intercom’s Fin agent handles customer conversations using historical data, multiple language models, prompts and other components. He claims its average resolution rate was approaching 50%, up from the high twenties at launch, and claims a competitor’s rate was lower. The interview does not independently verify the comparison or define comparable customer populations and ticket types.

He says Fin was developed through more than 100 experiments and substantial engineering on top of foundation models. His argument is that domain-specific application companies have time to create value before model providers can replicate their work. He expects increasingly capable models to change that balance over a longer horizon, but the timeline is speculative.

McCabe thinks the speed at which AI transforms the entire economy is overhyped, while the value some current AI products deliver is underappreciated. He cites an investor’s comparison of time to $30 million in annual recurring revenue for AI and earlier SaaS companies, but gives no complete dataset or methodology.

Stebbings warns that some rapid AI revenue may come from pilots and experiments rather than durable subscriptions. McCabe agrees that excitement creates noise and says usage, retention and customer expansion help distinguish lasting value. He predicts many venture investments in AI will underperform the S&P 500; this is a forecast, not a measured result.

Asked about companies building their own software and AI agents, McCabe calls that a poor long-term strategy. He argues specialized systems require extensive engineering and expects even companies experimenting internally to return to purpose-built products. Intercom sells such products, so his view has a direct commercial interest; the interview does not establish that every internal build will fail.

McCabe expects technology costs to fall over time and believes companies may use savings from human support to buy more AI assistance. He predicts that some customers will spend more if service improves, then pay less as capabilities become commoditized. These are economic expectations, not guaranteed savings or performance.

He expects automation to reduce the number of people needed for some tasks over time. At Intercom, he says, Fin had been used internally for nearly two years without layoffs in the support team, but the company stopped expanding that team. He anticipates future growth will be handled by AI and remaining employees will focus on more complex work.

When Stebbings asks about reported customer-service job cuts elsewhere, McCabe says circumstances differ: a business with a narrow set of recurring questions may be easier to automate than one with a broad range. He offers this as a hypothetical explanation, not verified evidence about another company’s staffing decisions or the quality of its service.

He expects foundation-model providers to subsume more application functionality eventually, but argues practical deployment and customer adoption take time. He predicts a large rise in robotics and automation of some physical work. Stebbings asks about a photographer facing lost work; McCabe believes people will adapt and new demand will emerge, but that optimism does not establish outcomes for displaced workers.

Stebbings questions whether AI startups assembled from existing models have defensibility. McCabe says early products can begin as simple tools and that the people building them matter. He argues software advantages are often temporary, although networks, ecosystems and data can create stronger defenses over time.

McCabe sees founder mode as a new label for an old practice: founders trusting their judgment and remaining directly involved. He calls building a company a creative act and favors individual agency over decisions by committee. These are his leadership preferences, not evidence that one management style always succeeds.

He says Intercom has revenue in the hundreds of millions, without giving a precise period or audited figure. Asked why it remains private, he says an IPO offers investors liquidity but public-company compliance and scrutiny can constrain long-term experimentation. His perspective is that of a founder managing a private company.

McCabe says Intercom nearly went public in 2022, prepared an S-1 and abandoned the plan as markets fell. He is glad the company remained private because he believes it allowed major changes. Stebbings counters that investors need liquidity; McCabe expects IPOs, private-market transactions and acquisitions eventually to return, while acknowledging the longer wait for venture investors.

He is unsure whether the venture model still works and suspects abundant early-stage capital will make returns difficult. At Intercom, he says growth slowed partly because the company forced more small customers into sales processes and sought larger upfront contracts. He claims customer counts have since resumed growing after changes; figures are not independently supplied.

To rebuild the company, McCabe says he refocused it on customer service, promoted hard work and resilience, and set a 52-week reinvention project with weekly communication. He says employees who did not fit the revised values left and aligned employees were promoted or received raises. This is his account of the restructuring.

Stebbings questions whether “hard work” is a meaningful value when few people explicitly oppose it. McCabe argues some views of work-life balance are less suited to exceptionally difficult goals and says he favors very hard work. He acknowledges people may choose different lives, but his claim that greater balance is less productive is an opinion, not a universal finding.

He describes setting policies directly rather than seeking consensus, including a two-day office mandate, and says he would choose five days if starting again. He believes decisiveness is necessary for his company; the conversation does not independently assess the effects on employees.

In the 2024 US election context, McCabe says he speaks publicly because he is concerned about individual liberty and peace. He claims Western governments threaten speech, technology and peace, and argues Donald Trump and JD Vance would better protect those values. These are his political views and predictions, not established facts or an endorsement by the interviewer.

Stebbings asks about concerns over Trump and democracy. McCabe says he supports and donates according to his view of liberty and believes many people in technology privately agree with him. His characterization of those people is anecdotal and cannot establish the views of the industry as a whole.

McCabe says Intercom does not discuss politics at work, but he would not tell employees what to say in their personal lives and would defend them against attacks even when he disagrees. He says one person left in connection with his public positions and reports no customer or board problems; those are his own account of the effects.

Reflecting on diversity initiatives, McCabe says he opposes what he considers discrimination in hiring and recounts hearing recruiters emphasize a candidate’s minority status. He presents this as a personal experience and criticizes diversity programs broadly; the anecdote does not establish how such programs operate generally.

McCabe argues that censorship by institutions and technology companies is dangerous. Stebbings distinguishes workplace rules from public expression and asks how society should handle false or harmful content. Their disagreement centers on who may decide what is true and what protections are appropriate.

Stebbings challenges an example of an allegedly false social-media post and asks whether labeling something a joke should protect damaging misinformation. McCabe favors public criticism and reputational consequences instead of a central authority deciding truth. His assertion about COVID-era treatments and vaccines is unsupported in the interview; it should not be treated as medical evidence or advice.

McCabe says speech that directs people to attack a named person at an address can cross into organizing harm. Stebbings raises a hypothetical staged video that could incite hatred; McCabe says the creator should be criticized and shunned, but resists giving authorities broad power to decide what is permissible. They do not resolve the boundary between expression and harm.

He argues individual freedom carries risks and says governments should protect citizens rather than suppress expression. He attributes different attitudes toward freedom in the US and Europe partly to history; that is his interpretation. Stebbings describes feeling constrained in what he can say, and McCabe sees that as evidence of a cultural problem.

Discussing his return as CEO, McCabe says illness, slower growth and negative press had left him physically and psychologically exhausted. The experience forced him to abandon an idealized self-image and return with greater authenticity. He says he now feels more comfortable being direct and accepting that some people will disagree.

He describes being bullied at school and still feeling a need to prove himself. He believes that insecurity can contribute to entrepreneurial drive, while recognizing it as something he once felt ashamed of. His claim about founders’ motivations is personal observation, not a diagnosis or general rule.

McCabe describes the San Francisco Freedom Club as a way to combine parties and social connection with his advocacy for liberty. He reports a waiting list of 3,700 and says 500 people attended an event to which 600 were invited. Those numbers are his account and do not establish broader public support for its political values.

In the quick-fire round he warns that nuclear war is possible and cites a scenario in which billions could die. The number is a claim attributed to a scenario he references, not a prediction or a verified estimate of a specific conflict. Asked about the Middle East, he says he does not know what will happen.

He says he has changed his mind about language models quickly replacing software, now expecting a slower transition. He praises investors he knows, including Ilia Fushman and Mamoon Hamid, noting that Hamid invested in Intercom’s Series A. His final answer on leadership is that being more authentically oneself can make a leader more effective.