Human AI
Dr. Ian Malcolm famously said, "Your scientists were so preoccupied with whether or not they could, they didn't stop to think if they should." Lift that line out of a film about rampaging dinosaurs and it lands uncomfortably well on how we're deploying AI right now.
The fact you can automate something doesn't mean you should. I'm broadly optimistic about AI, but the good stuff comes from a partnership with these systems, especially on tasks that lean on human judgement. Not wholesale replacement. Something more like a sensible division of labour where AI amplifies what people already do well.
Take the rush to swap out customer service agents for LLMs. On paper it looks trivial. Everyone's used ChatGPT, so the technical path seems obvious. The hard bit isn't teaching an LLM to sound like a support agent. It's the cage you have to build around it to stop it wandering off somewhere it shouldn't.
That's the real paradox. How do you put meaningful boundaries on one of these things without killing the very thing that made it useful? ChatGPT is powerful precisely because it's general; it'll have a go at almost anything. That same flexibility turns into a liability the moment you need predictable, narrow, well-behaved output from a support agent.
Trying to make an LLM completely replace a human agent misreads what each side is actually good at. When you hire a person, a load of behavioural constraints arrive pre-installed. No manual needs a line saying "don't answer billing queries with a three-page literary essay" or "don't recommend a nice Italian place when someone's software is broken." People carry contextual awareness that falls out naturally from understanding social norms and where the professional lines sit.
A language model has none of that baked in. It's a pattern-matching machine, just as happy generating the next line of a Python script as the next sentence of a refund reply. Anyone sufficiently bored or motivated can nudge it well past where you wanted it to stop.
The application that actually works treats AI as a layer on top of the human, not a stand-in for them. Picture a support call where the person handles the conversation while AI runs alongside: transcribing, pulling up the right documentation through semantic search, flagging patterns from past cases. The human does what humans are best at, talking to another human, while the machine handles the grunt work no individual could keep up with.
There's an ergonomics problem here too, not just a technical one. LLMs with strict guardrails tend to become the thing that traps you in a circular conversation going nowhere. Deploying something this powerful and then frantically boxing it in is like using a nuclear reactor to power your car. The chat bot has quietly become a modern monument to bad design.
When a customer reaches out, they've already hit a wall. They've got a problem they couldn't sort themselves. Bolt on another layer of automation, especially one that misreads context and fires back generic answers, and you've made it worse, not better. Your shiny technical stack counts for nothing if the experience falls apart at the exact moment it matters.
Klarna gave us a tidy case study. After loudly championing AI-driven customer service, they walked it back, admitting customers actually prefer talking to a person. It's a reminder that LLMs aren't the answer to every problem, and that customer psychology has a vote.
The way forward isn't human or machine. AI is brilliant at retrieving information, spotting patterns, and chewing through routine queries at scale. Humans bring emotional read, creative problem solving, and the knack for navigating ambiguity. Stop treating AI purely as a way to cut headcount and start treating it as something that lets your agents do their job better, faster, without stripping out the empathy that makes the interaction worth anything.
None of this is really about customer service bots. The enthusiasm for automation hides a duller truth: tools are still just tools, however clever. Their worth comes from how thoughtfully you fit them in, not from raw capability. The most elegant solution is rarely the most advanced one. It's the one that helps the person on the other end.