Yesterday I listened to an interview with Andrey Doronichev . One idea from the conversation stayed with me. Andrey argued that the biggest bottleneck for AI is no longer the technology itself, but the last mile: convincing a 70-year-old dentist that AI belongs in their clinic and can genuinely improve the way they work.
That resonated with me because it’s surprisingly close to what I do every day. At Sully.ai , I work on AI receptionists for healthcare providers, and while building the system is an interesting engineering challenge, getting it to fit naturally into a real clinic is often the harder problem.
The technical side is only one piece of the puzzle. The system has to earn trust, integrate into existing workflows, handle edge cases gracefully, and behave consistently enough that people are comfortable relying on it with real patients. That’s where a lot of the real work happens.
Listening to the interview made me realize that I don’t really think of my job as “building AI” anymore. I think of it as helping bring AI into industries where people have established ways of working and very little tolerance for mistakes. The technology is impressive, but adoption is ultimately a human problem.
It made me curious whether other people working on AI products have experienced the same shift. Has the biggest challenge in your work become the technology itself, or everything that comes after it?