It has now become obvious that AI will find great product-market fit in places where repetition is the norm, and what better place than customer support? It is a universe that relies heavily on canned responses, knowledge bases, resolution paths & adherence to policies.
So, a safe bet you can make today is labeling customer support as the next AI victim (as some like to call it). But I don’t believe there are going to be any victims when you look at the long-term view of how AI will shape the support experience
Phase I – Efficiency gains
AI at the most basic level will prove to be incredibly powerful when working with huge bodies of text (a.k.a. knowledge bases). People are going to be bewildered by a model doing the work that content and enablement teams used to do. There will be no need for long, tiring reviews of resolution paths; an LLM can handle that, and basic text editing will become 10x easier once you get a hold of this technology.
Phase II – The AI bot (we are here)
As of the day of writing, there isn’t an exemplary AI chatbot that can claim to have completely solved the "AI as a support agent" problem, but we’re getting there. Rudimentary RAG provided a solid first leap, and now skill-based support agents are proving to be better than expected. Given the right tool access and correct prompting, an AI model based purely on skills can now be 10x more helpful than its predecessor.
Phase III – Proactive support
As we get better at building AI support agents driven by specific skills, we will shift our attention toward finding patterns, identifying broken customer journeys where AI can step in, surfacing directly inside the product the moment it detects friction. The customer will no longer need to go out and look for help; instead, the product itself becomes the support surface. It’s just a matter of time until we figure out how to embed these models seamlessly across the entire customer journey.
Phase IV – Personalized support
This is where I envision the support experience heading: an ecosystem tailored so closely to its user that the product knows you inside out. It won't just anticipate a generic problem; it will anticipate your problem. Because it understands your specific configuration, your billing history, and the exact button you clicked ten seconds ago, the solution is already waiting for you. It will be an experience so seamless that you’ll feel the product knows you perfectly, understanding possible friction points in your journey and proactively unblocking you from them.
To go from Phase I to Phase IV is a big leap, but it's one I’m excited to build and push for. I truly believe AI will get rid of our tedious problems and effectively enable us to enjoy a seamless product experience, without ever having to think about support.