Knowing How to Use AI Isn’t the Same as Using It

Everyone is trying to get value from AI, but most companies don’t feel they’re achieving the impact they wanted. It isn’t the models. Today’s AI systems are remarkably smart, and so are the people trying to use them. And yet the promised transformation still isn’t showing up. There’s even a term for it: capability overhang, the gap between how much the models can do and how little we get out of them.

So how should companies address the gap? It’s easy to assume that people are the limiting factor and that the solution is to improve their AI skills. Companies put their energy into rolling out tools, running fluency courses, and standing up enablement programs. The logic follows that once people know how to use AI, they will use it more, and the value will come.

So companies roll out their AI upskilling initiatives, and people’s understanding of AI improves. And yet, usage numbers remain low. According to PwC, only 14% of workers use AI tools every day. So if the tools aren’t the problem and people’s skills aren’t the problem, why aren’t people using AI more?

From our perspective, it all comes down to the interface. Right now, using AI means remembering to stop and go do something outside your workflow, which relies on an individual’s self-motivation. It’s a bit like a dentist asking you to floss every day. The problem isn’t that you don’t think flossing matters or that you don’t know how to floss. It’s that it’s kind of inconvenient, easy to forget, and not part of what you’re already doing. You’re not the problem, and neither is the floss. It’s the routine.

AI skills are not the same as AI habits

The fix is a different interface, one that helps you build a habit. A habit needs two things to form: It has to be easy and worth it. That combination is what creates behavior that runs on autopilot, and it has nothing to do with how good you are at prompting. You build an AI habit by putting the tool directly in the flow of work, so reaching for it never feels like a detour.

We’ve seen the effectiveness of this interface firsthand. For years, Grammarly has quietly shown up where you’re already doing your work and has been used enormously. It now runs more than 100 billion LLM calls a week, thousands of interactions per user every day, far more AI than most people realize. They get that help within the flow of their writing, rather than copying and pasting their text into another tool. There are other ways to get writing help, so why do they keep reaching for Grammarly? Because it’s seamless, right where they’re working. That’s what makes it a habit: It comes to you, and it only asks you to say yes.

We extended what we learned from Grammarly into Superhuman Go. You use Go the way you use Grammarly. You type, and the agents underline suggestions directly in your flow of work. At first, they’re small, like an incorrect stat caught before you hit send. But those little inline saves add up, and as you learn to trust them, you start handing off bigger jobs, like checking a claim against company data or pulling together a report. At some point, using the tool becomes a habit without you ever deciding to build one.

Bring the smart friend to your desk

Here’s how I like to describe the difference between the two interfaces and their relationship to building habits. Until now, AI has been like a really smart friend sitting at a coffee shop down the street. If you want their help, you have to pack up everything you’re working on, walk to the coffee shop, show them what’s in your notebook, and ask. They’ll give you great advice. But you have to remember they’re there, re-explain what you’re working on, then carry their advice back to your desk.

Go is that same friend at the coffee shop, but now they’re sitting beside your desk, glancing at your screen, and tapping you on the shoulder when they can help. You don’t have to explain anything or even remember they’re there. They just whisper the help in your ear. You may know perfectly well how to walk down to the coffee shop, but it doesn’t mean you’ll build a habit out of it. Yet the person sitting next to you? Working with them becomes a habit before you make a conscious choice.

And staying put has additional benefits. Knowledge workers are interrupted every two minutes during core work hours, roughly 275 times a day, and every interruption pulls them out of their work. When the help comes to you instead, you stay in flow, keep your momentum, and do your best work without a tool distracting you. Right now, getting value from AI feels like both a burden and a distraction. It should feel like neither.

What a habit-forming tool looks like

An interface built to help you form a habit looks something like this.

1. It should be proactive. You shouldn’t have to remember it’s there, stop what you are working on, and go seek it out.
2. It should help you everywhere you work. When AI follows you across your inbox, a doc, a Slack thread, or your CRM, using it no longer requires you to take a detour.
3. It should feel like collaborating with a teammate. Imagine writing a blog post with someone, but the only way to collaborate is back-and-forth over Slack. It’s a much better experience to leave comments and suggested edits right in the doc, where you can review, accept, or dismiss them.

Companies are trying to instill AI skills, but skills don’t add up to a habit. We think the problem is the interface, and that when it’s right, habits form. That’s what Grammarly taught us, and it’s why we built Go. When getting value from AI stops being a burden and starts showing up in your work, it ceases to be a reward for the motivated few and becomes the default outcome for everyone.