There is a growing gap between what the market is saying about AI and what we are actually hearing from customers. The media, VCs, AI labs, and influencers have all talked about AI replacing people, issuing trusted software, and token-maxxing as the outcomes to be pursued. But leaders who run real businesses are increasingly asking the right questions. How do I make my people better with AI? What programs can I trust? How can I measure the ROI of this spend? We hear these questions every day.
After three and a half years of building, deploying, and watching many of our growing customers put AI to work, the AI ideas we’re most confident about at HubSpot are the things almost no one else is saying out loud.
Here are six of them.
AI work is not AI results.
The industry has a confusing proposition for progress. Writing emails, producing summaries, doing research. These are tasks that AI has made very easy. They’re useful skills, and we’re posting them at HubSpot. But the work is the input, not the result. A work without results is theater.
Companies that win with AI are those that work backwards from a business problem, not forwards from a demo model. For example, customers using a Customer Agent respond to tickets 25% faster, while those using a Prospecting Agent generate 76% more leads.
That’s why we moved Customer Agent and Test Agent to result-based pricing in April. AI results are what matter. And that’s what growing businesses help deliver. We put our values where our vision is.
AI is needed. It’s not enough.
Generating code is certainly easier now. Anyone can build a prototype in a weekend, but it breaks and falls apart when actually used. Downgrading to production codes doesn’t raise the ceiling on ship value because the things that drive a growing business are hard, not easy.
You still need to have clean data, not another silo. You still need to integrate with tens of applications. You still need a comprehensive view of the customer across all marketing, sales, and service, powered by context.
The factory will sell you a model or single purpose agents. But it won’t sell you a system in between: data cleanliness, workflow design, change management. That is left up to the customer. And when cut-off point agents pile up, the job gets tougher.
The future belongs to companies that build AI in a parallel system, where data, workflows, agents, and people share context. That’s what we’re building at HubSpot. AI is a new layer, not a replacement for the foundation.
AI needs to be built for the Future 5000, not just the Fortune 500.
Today’s AI guide is written by a company that can’t make it work. According to their disclosure, frontier labs are spending billions of dollars on distributed front-end engineers to make AI work within large corporations.
That model works if you are a large business. It doesn’t apply to the millions of growing businesses that will fuel the next decade of growth. A small or mid-sized company can’t find front-end developers, rebuild its data pipeline, or build a core platform to make it all work.
So when the consensus is “AI is for everyone,” it’s worth asking who it’s really working for today. Basically, it’s the customers who can already make it work, rather than the armies of engineers and developers behind it. That is not democracy.
We prepare results per token, not tokens per transaction.
There are business model conflicts in the AI industry that customers have yet to fully recognize. Marketers who benefit most from the use of AI are not motivated to make AI cheaper or more efficient. They are encouraged to keep the meter running. So customers are asked to pay for the job and are told they are buying change.
The honest version of AI economics is the opposite: be specific about the outcome the customer is trying to drive, and then find a cost-effective way to drive it. That’s the customer’s job. It should also belong to the seller. Right now, it isn’t.
Token-maxxing is a dealer’s game. Outcome-maxxing is for the customer. Retailers who align with the customer will win. Dealers who match the meter may not.
AI should empower humans. It cannot be changed.
The loudest AI narrative is autonomy: agents are replacing humans, the head count is falling, the future is less populated. That narrative was created for Wall Street, not Main Street. We reject that draft.
We build for the person who does the work, not the person who is taken out of the budget. A lawyer closes many deals. An advertiser sends multiple campaigns. A service person who solves complex problems. Owner who runs most of the business himself. The job of AI is to make it more powerful, not to make it disappear.
Yes, we send independent agents. But independence is a skill, not a mandate. Customers decide where to delegate, where to keep people in the workflow, and where AI suggests. Our defaults are designed to help the user, not cut the org chart.
We believe in human truth and AI efficiency. Things that AI cannot replace – trust, judgement, taste, relationships will gain more importance as things that AI can do become ubiquitous. Companies that bet on humans will lose the customer, the employee, and ultimately society, when 57% now think that the risks of AI outweigh its benefits.
Trust is more than a privacy policy.
Every AI vendor wants to be trusted. But many describe it as a security posture: we will not train ourselves with your data, we comply with SOC 2, we offer business SSO. Those things matter. They are also table poles. None of them is a separate claim. That’s what you promise.
This proves otherwise. Real trust is a complete business model: how you choose a model and handle the costs, reliability, and management of your agents. That’s what customers are asking for. Can I rely on model selection? Can I trust the cost? Can I trust you to be honest? Can I trust you to rule?
Privacy answers what we won’t do. Trust responds to our desires. Most of the industry is still answering the first question. The second is what customers need.
All this includes you
The AI consensus holds for a long time as no one in the room has to answer you. Cut the figure. Remove the old stack. Keep the meter running. Trust us.
Growing businesses can’t afford to waste time sorting through what’s hype versus what’s real. They don’t have forward-deployed developers to throw at deployments. They can’t absorb the labor-charged pricing model and call it change. They can’t build on a stack that treats people as different.
They need AI built on a foundation that works for them, designed to empower and not exhaust their people, and delivered by a vendor whose business model is aligned with theirs, not against it.
That’s what we’re building at HubSpot.
