Every morning, Kolin Koehl wakes up to a list of to-dos, email drafts, and calendar items that AI has already prepared for him. It's not laziness - it's strategy. The HubSpot GM for Service Hub has a clear philosophy: reclaim time from the mundane, and spend it on the things that actually compound. Here’s how he sees the AI revolution.
Mindstream: If you had to explain your AI philosophy in a single sentence, what would it be? How has it evolved over time?
Kolin Koehl: Intelligence for the masses. AI is a once-in-a-lifetime opportunity to distribute cheap intelligence to every person on the planet; uplifting their careers, their families, and their ability to live more meaningful lives.
Early on, I'll admit I was more worried about a singularity-type event. Over time, I've come to see this as profoundly non-zero-sum. The pie is growing exponentially for everyone.
How do you personally decide what to automate and what not to automate?
Trial and error, constantly. Every morning, I wake up to a list of to-dos, social posts, email drafts, and calendar items that AI has helped me prepare. I try to automate nearly everything, and when the tools can't handle it yet, I simply try again on the next model release.
The goal is to reclaim time from the mundane. It means I can accomplish significantly more, and it means I get more time for the things that actually matter: coaching my team, thinking long-term, and being with my family.
Which human skill do you think is becoming more valuable in the AI era?
Empathy and taste. Empathy is the ability to sit, listen, and actually comprehend someone's problem; not just the one they're articulating, but the one two or three layers deeper. That kind of foresight is still deeply human. And taste; really knowing what good looks like and setting the bar is incredibly hard for AI today. Both of these are compounding in value.
Are you worried we’re in an AI bubble at all?
No. This isn't the dark fiber scenario of the dot-com bust. Every GPU that gets bought goes online, and GPUs plugged in ten years ago are still running at full throttle. Consumption and demand are real and growing.
Since December, when we saw a step-change in the cognitive capabilities of these models, the trajectory has only steepened. It's hard to call something a bubble when the usage is this concrete.
As an AI agent expert, which agent use case has blown your mind the most?
OpenClaw. The idea of a persistent, locally-run personal assistant that's always on is a genuine paradigm shift.
What makes it remarkable isn't just that it can control your computer, it's the architecture. The agent teaches itself new skills recursively. Ask it to turn on a light in your house, and it'll scan the network, find the device, and if there's no API available, it will actually write the tool itself to control it, then use that tool going forward. The possibility space is infinite.
And the heartbeat functionality is what really changes the modality: every 30 minutes, the agent wakes up, looks at your world, and decides if there's something it should do for you. No prompt required. That shift from reactive chatbot to proactive assistant is the real unlock, and I think it's how we'll all interact with AI within a few years.
What’s one small decision you’ve made that had an unexpectedly large impact?
Conventional wisdom was to optimize for cost; pick the cheapest model that's "good enough." We made the opposite bet: optimize for intelligence and customer value.
Give every user the best possible experience, solve their actual problem, and trust that the models will get better and cheaper over time. That one decision shaped everything downstream. The team stopped debating cost tradeoffs and just focused on customer value.
It's paid off in spades; every new model generation has come in smarter, faster, and cheaper, which means our margins improve automatically while the customer experience keeps getting better. The lesson is simple: when you're riding an exponential curve, optimize for the thing that compounds.
When you think 10 years ahead, what can you imagine people doing with AI agents that they aren’t doing today?
Robotics and physical AI. When intelligence can inhabit a physical form, it doesn't replace human works; it expands what's possible. Think about lithium mining, deep-sea infrastructure, disaster recovery; environments that are genuinely dangerous or inaccessible for people.
AI-powered robots and drones will let us do work we literally couldn't do before, or could only do at great human cost. That creates abundance: lower costs of goods, safer supply chains, new industries. And just like every other wave of automation, the human roles around that work will grow, not shrink.
Which recent AI breakthrough made you rethink something fundamental?
Andrej Karpathy's auto-research concept. The ability for an AI agent to take a prompt; a system prompt, an ad, whatever, and recursively iterate on it, improving itself through hundreds of cycles, is a genuine breakthrough.
Marketers are already using it to evolve Facebook ad copy daily with measurable results. And when you point it at an agent's system prompt, you wake up the next morning with a meaningfully better system, sometimes with massive performance jumps. Recursive self-improvement at the application layer is here.
What’s your AI hot take?
This is the worst AI will ever be. It only gets better from here. This is Moore's Law in action.
Finish this sentence: In five years, AI will make people feel ___ about their work. Why?
Creative.
Right now, the ability to build software, design products, and ship ideas is gated behind years of specialized training. AI is removing that gate.
In five years, a marketer will build their own tools. A support rep will automate their own workflows. A founder with no technical co-founder will ship a real product. The feeling people will have isn't just productivity; it's the feeling of being a creator when they never had permission to be one before.
Intelligence is becoming cheap, and when intelligence is cheap, the number of people who can turn an idea into reality goes from millions to billions.
Kolin Koehl is the General Manager of Service Hub at HubSpot!