How to Become an AI-First PM and Designer

Adams uses AI for research, summaries, and feedback on strategy, treating its output as material to evaluate rather than copy. He worries that relying on summaries instead of joining customer interviews can weaken product judgment, which develops through repeated direct contact with customers.

Adams says AI will become a standard part of product work, much as mobile did, though the current labels and hype may fade. He recommends reading widely, trying products hands on, and learning from colleagues. When hiring, he values curiosity, ambition, open mindedness, and a record of making things, even over prior AI experience.

After ChatGPT launched, Intercom changed its roadmap and shipped Fin about four months later. Its earlier Resolution Bot required customers to curate answers manually, while Fin could use a help center or knowledge base with little setup. Adams says Fin initially handled about 25 to 28% of queries out of the box and averaged 45% a year later, with some customers reaching 60 to 70%.

Adams says support teams face growing question volumes, legacy systems, and pressure to control costs. LLMs can answer routine questions quickly and in multiple languages, while more complex troubleshooting still needs people. He describes Intercom's Copilot as support for human agents and expects AI to take on more complex work over time.

Adams says AI teams sometimes need to test what the technology can do before matching it to a known customer problem, because capabilities and models change quickly. Intercom's principles are to start with a real, high energy problem; think big, start small, and ship to learn; and define outcomes before shipping. Teams generally work in small groups, set goals, and use six week cycles.