In late 2022, facing declining revenue and a failed IPO attempt, SaaS giant Intercom made a radical decision: to bet its entire future on Artificial Intelligence. This bold move, spearheaded by Chief Product Officer Paul Adams, led to the creation of Fin, an AI agent now resolving over a million customer issues weekly with a 65% resolution rate. Adams shares critical insights into Intercom's brutal yet transformative journey, offering invaluable lessons for B2B companies struggling to embrace the AI revolution.

Watch the full discussion with Paul Adams on Intercom's AI transformation.

Paul Adams, a veteran in product leadership, joined Intercom when it was a mere 14-person startup, having previously advised the company. His extensive background includes leadership, product, and UX roles at tech giants like Facebook (Ads, Platform) and Google (Gmail, Docs, YouTube), where he was part of the mobile team during the iPhone's launch. Adams is also the author of the best-selling book Grouped on social software design and co-hosts the Intercom on Product podcast with co-founder Des Traynor.

The arrival of ChatGPT in late 2022 served as a wake-up call for Intercom, which had endured five quarters of declining revenue growth and abandoned an IPO process. Within two weeks of ChatGPT's release, the leadership team made the audacious decision to pivot the entire company towards AI. This high-stakes gamble ultimately birthed Fin, Intercom's highly successful AI agent for customer service.

The Top 5 Takeaways from Intercom’s AI Transformation

1. Embrace Brutal Transformation

Paul Adams emphasizes that transforming a traditional SaaS company into a true AI powerhouse is inherently painful. Intercom's pre-AI struggles, though challenging, provided the necessary pressure. The leadership team swiftly overhauled their strategy and roadmap, making it clear that AI integration was non-negotiable.

“If you’re a SaaS company who thinks you’re an AI company and you’ve not gone through brutal transformation, you’re not there yet.”

He highlights that many companies shy away from the difficult decisions, such as workforce restructuring or completely rebuilding marketing strategies, opting instead for superficial AI feature additions. Adams himself took over two-thirds of Intercom's marketing six months prior and immediately dismantled and rebuilt the entire department from the ground up, believing it was the only way to create an organization fit for the AI age.

2. Push Boundaries: Go Beyond the Obvious

Intercom operates on the principle that true boundaries are only discovered by crossing them. This philosophy permeates their operational changes. For instance, every designer at Intercom now ships code to production, a stark contrast to 18 months ago when none did. This mandate was clear: adapt or find a role elsewhere. Engineering teams are also on a path to double productivity, not through incremental tweaks, but by setting it as a non-negotiable target.

Adams constantly challenges existing structures by asking, “What would a brand new startup incorporated today do here?” This often leads to consolidation rather than specialization, questioning the need for distinct roles like product marketers and content marketers, or separate product managers and designers.

3. Reinventing Software Development for the AI Era

Intercom had long-standing, refined principles for building SaaS products, proudly taught to every new employee. However, with AI, they had to abandon them entirely. The old approach—identify a job, listen to customers, design, build, and ship—relied on stable technology and certain execution, with design being the primary challenge.

The new AI-first paradigm is fundamentally different: "Ask what AI makes possible → Prototype to see if you can build it reliably → Build the UX later → Ship → Learn at scale." Execution is now uncertain, design is cheaper, and the real complexity lies in the AI infrastructure—the RAG (Retrieval Augmented Generation) system, custom models, and empirical evaluation. Adams admits, "This AI layer, our RAG system, has been 3 years in the making by a very talented team. It’s complicated. I do not understand the depths of that RAG system at all." The visible UI has become a minor component, with the invisible AI infrastructure now housing the core product, a complete inversion of traditional SaaS development.

4. The Compounding Power of Incremental AI Improvements

In AI products, especially those chaining multiple AI steps, success rates multiply. A 99% accuracy across 10 steps yields 90% overall, while 95% accuracy drops to 60%. This mathematical reality drives Intercom's obsession with tiny, incremental improvements across every part of Fin's system. They conduct hundreds of experiments, many of which fail, and sometimes an improvement in one area can degrade another.

Their strategy includes building custom models specifically tailored for discrete customer service tasks, rather than relying on general-purpose solutions. Adams stresses that "Each single tiny incremental improvement in each of these steps adds up to the highest performing product, adds up to something people can trust, adds up to something people will replace their humans with." This compounding effect also addresses what co-founder Des Traynor calls the "marketing overhang" problem: many companies can demo an AI product, but few can deliver one that works reliably at scale. Intercom adheres to a strict rule: no product launch unless it's a real demonstration of a solution proven to work at scale, citing examples like Apple Intelligence's delayed delivery.

5. Navigating the Dual Identity: SaaS and AI Businesses

A successful AI transformation often creates a new challenge: effectively running two distinct companies. Intercom now operates its traditional SaaS product alongside Fin, the AI product. The SaaS product exists in an "easy product domain" with predictable metrics, clear differentiators, and customers who articulate their needs clearly. Fin, however, occupies a "new product domain" characterized by chaotic metrics (e.g., "Fin’s grown 300% year-over-year—is that bad?"), customers unsure of their needs, and rapid change making the future unpredictable.

The buyer landscape has also shifted dramatically. Previously, Intercom sold directly to customer service leaders. Now, the buying committee is complex, including the influential but non-decisive customer service leader, a C-level executive overseeing company-wide AI transformation, and an AI-fluent technical evaluator assessing product efficacy. These stakeholders operate in different spheres, attending different events and consuming different information. Paul Adams finds himself engaging with CEOs at private dinners while also attending trade shows for customer service leaders, highlighting the minimal overlap between these distinct buyer groups.

Intercom’s Top 5 Mistakes You’ll Probably Make Too

Mistake #1: Adding AI Features, Not Reimagining Your Product

True AI transformation isn't about simply bolting AI features onto an existing product. Fin is fundamentally distinct from Intercom's traditional SaaS offering, designed and conceived with entirely different principles. Adams notes, "Fin usage is eating Intercom usage at times. They’re totally different products."

Mistake #2: Avoiding 'Self-Harming' Strategic Decisions

Many companies prioritize protecting current revenue, avoiding board disapproval, or upsetting sales leaders. They hesitate to take a short-term revenue hit for long-term strategic gains. However, Adams argues that these "self-harming" decisions are crucial. "If it doesn’t feel really painful, you’re not deep enough," he warns.

Mistake #3: Diluting Vision and Delaying Action

Companies often succumb to procrastination ("Q1 is important, let’s do it in Q2") or scale back their ambitious visions. They focus on easy wins while sidestepping the difficult, transformative work, convincing themselves they've done enough. Large company inertia can set in, and listening too much to skeptical customers who insist on "human service forever" can be detrimental—many of those customers now use Fin.

Mistake #4: Failing to Cultivate a 'Fighting' Culture

Success in AI transformation requires a team willing to "fight" for outcomes, embracing tension and positive disagreement. Adams believes that if Intercom, a company with no inherent special advantages, could achieve this, any SaaS company can, provided they commit to the decision.

Mistake #5: Denying Your Own Mistakes

Intercom's executive team engages in constant, honest, and soul-searching conversations about existential questions that determine their success. Adams stresses the importance of deep, trusting relationships among peers to confront difficult truths. Companies that fail to transform will slowly fade into irrelevance, losing talent to genuine AI innovators, and eventually, ceasing to exist.

Three years into its emergence, AI is no longer a trend but an inevitability. The familiar landscape of B2B SaaS, shaped by the last decade of mobile and social, is irrevocably altered. The critical question for every business now is: Will you change fast enough?

Watch the full discussion with Paul Adams on Intercom's AI transformation.