How to Feel Confident Using AI at Work

Someone on your team drops an AI-written summary into the chat and everyone says it's useful. You say it's useful too. Then you go back to your own screen, open the tool your company gave you three months ago, look at the empty box, and close it again.

Confidence with AI doesn't come from understanding AI. It comes from using it on small pieces of work you already know how to do, often enough that opening the tool stops being an event. You build it the same way you built confidence in every other part of your job, by doing the thing imperfectly a few times in a place where the stakes are low.

Let's break this down.

Key takeaways

• You are less behind than it feels. Daily AI use at work is still the exception, not the norm.

• Start where you already know the right answer. If you can judge the output instantly, you learn something. If you can't, you just get nervous.

• The worry that it looks like cheating has something real behind it. People get judged when their work is seen as AI-assisted. The work itself doesn't get marked down.

• Write your own version first, then ask. You keep the thinking and you get the edit.

• Two short sessions this week beat one long Saturday.

Am I behind if I still haven't started?

Probably not as far behind as it feels on a Tuesday afternoon.

Gallup's 2026 tracker of U.S. employees found that 52% of workers use AI on the job and 15% use it daily. Daily use is still a minority habit. The same tracker found that 47% say their organization has integrated AI tools, and only 25% say their organization communicated a clear plan for it. Three in four people are figuring out what to do with AI without being told what the plan is.

Robert Half, in its own 2026 research, reports that 35% of workers feel very confident using AI tools effectively. That is roughly one in three.

There is one careful field measurement worth knowing. Acar, Gai, Tu and Hou, writing in Harvard Business Review, looked at 28,698 engineers at one large technology company and found that 41% had tried the company's own AI coding assistant twelve months after it was rolled out. And these were software engineers at a tech company, with the tool handed to them. Most of them had not touched it a year in.

So when you look around and assume everyone else has this handled, you're guessing, and the available numbers say you're guessing wrong.

One small shift: stop treating "I haven't started" as a verdict on you, and start treating it as a gap in what you were told.

How do I start using AI when I have no idea where to begin?

Here's what your brain is doing with that empty box.

Psychologists call it challenge and threat appraisal, a model Blascovich and Tomaka developed in 1996. You weigh what the task asks against what you already know how to do, and whichever way it tips determines whether it feels like a challenge or a threat. Researchers studying police officers under stress put it this way: "More challenge-like states are experienced when a person evaluates his or her coping resources as meeting or exceeding the demands of the situation or task, while more threat-like states are experienced when a person evaluates the situation or task demands as exceeding his or her coping resources."

The useful part is that both sides of that can change. You can make the task smaller, or you can remind yourself what you already bring to it. Most people go for the second one first, usually by reading more about AI. That's the slower route, and there's always more to read.

Make the task smaller instead. Pick one thing you already do well and already know the right answer to. Draft the status update you've written fifty times. Summarize a document you wrote yourself. Rewrite a paragraph you already like. You'll be able to tell in five seconds whether the output is good, because you are the expert on that particular piece of work.

Nobody builds confidence by starting with the client proposal. The NeuroLeadership Institute made the same argument about speaking up at work in 2019: practicing at the easy end builds the habit that makes the hard end more likely.

One small shift: pick the task where you already know what good looks like, and let the tool try to match you.

Why does using AI make me feel like I'm cheating?

Because there is something real underneath it, and it helps to know how big it is.

In a controlled experiment, raters scored the same piece of work. When the author was described as having used AI, competence ratings dropped by about 9% when compared with authors whose identical work was described as AI-free. Two details matter. The manipulation was perceived use, not verified use, so the ratings shifted based on a description. And the bias was competence-only: raters scored the work and the author separately, and the work itself was not judged worse. Gai, Hou and Tu's own Harvard Business Review write-up notes the penalty was more severe for women and older workers. This is a working paper, posted in May 2025.

So the feeling isn't imaginary. It's just pointed at the wrong thing. Nobody decided your work was weaker. Somebody decided you were, based on one sentence about how you got there.

In 2019 I ran an exercise with title staff, and one line stayed with me: "I'm the go-to person in my office, and I don't know how to use the new system." That is not resistance. That's somebody whose standing at work came from knowing things, sitting in front of something they didn't know, in front of people who come to them for answers.

After some personal trial and error, I found that the best approach is not to hide it or announce it. It's to say what you did with it. "I drafted this with AI, then checked the three figures against the source and rewrote the recommendation." That sentence shows your judgment, which is the part of your work a tool can't do for you, and the part people rarely see.

One small shift: when you share AI-assisted work, name the checking you did, not the tool you used.

Does using AI make me worse at thinking?

I can't point you at settled evidence either way, and I'd be careful with anyone who says they can. The tools are too new, and the studies are too young.

There is something well established close by, though, and it gives you something to do. It's called the generation effect: people remember information they produced themselves better than information they read. Bertsch and colleagues pooled 86 studies and 445 effect sizes in 2007 and found a mean effect of d = .40, building on work by Slamecka and Graf in 1978. That is laboratory memory research, so applying it to your own inbox is inference rather than proof, but the direction is clear enough to act on.

So do it in this order. Write your own version first, roughly, badly, in five minutes. Then ask the tool. Then compare the two and notice where it went somewhere you wouldn't have.

You keep the thinking, because you did the thinking. You also get a much better sense of what the tool is good at, which is the thing you were trying to learn in the first place.

How do I use AI without feeling overwhelmed?

If you're braced for AI to mean more work rather than less, you are reading your situation correctly.

Aruna Ranganathan and Xingqi Maggie Ye, writing in Harvard Business Review in February 2026, argue that AI doesn't reduce work; it intensifies it. In their research, employees "worked at a faster pace, took on a broader scope of tasks, and extended work into more hours of the day." Nobody tells you that in the rollout email. I can personally vouch for this assessment. I start out setting up my website, and two hours later I'm in blue ocean strategy mode talking with my AI.

You can't do much about the pace. You can do something about the scope. Pick one kind of task. Not a category like "writing." One repeated thing: the weekly summary, the meeting notes, the first pass at a client reply.

Then give it time, not intensity. The NeuroLeadership Institute's AGES Model™ names four conditions for learning that lasts: attention, generation, emotion, and spacing. Its guidance on spacing is that several short sessions with real time in between beat one long session, and that coming back to something after a gap is what makes it stick. NLI doesn't put a number on the gap. The cadence I'd pick is two short sessions on the same task this week, rather than a Saturday spent learning AI in general.

One small shift: pick one repeated task and stay on it, instead of trying five tools once.

What AI skills do I need to stay relevant at work?

It isn't prompt tricks. Those change every few months, and they were never the hard part.

IBM's Institute for Business Value, in its own 2026 study of 2,000 chief executives across 33 geographies, found that 83% say what decides whether AI works is people picking it up, not the technology itself. That matches the Gallup numbers above, where roughly twice as many employees say their organization has put AI tools in as say they were given a clear plan for them. The technology is not the bottleneck.

Three things hold their value:

• Describing your own work precisely. Most poor AI output is a poor brief. The skill of saying what you want, in what form, and for whom is a work skill that also happens to be an AI skill.

• Telling good output from plausible output in your own field. Nobody can hand you this one. It comes from the years you spent learning what a real answer looks like, and AI makes that worth more, not less.

• Being willing to say how you got there. When people say what they tried, nobody else has to start from scratch.

If you are still hesitating, access my free AI Confidence course at Feed Your Brain | Learning.

The belief underneath all of this

For most people, the sentence running in the background isn't about AI. It's "I should already know this by now."

That belief is worth addressing directly, because it's what keeps the tool closed. Here's The Shift Method applied to it.

• Identify the belief. Say it out loud in plain words. "I should already know how to use this."

• Question the evidence. Who taught you? When? What did your employer provide, and when? Only one in four people say their organization communicated a clear plan.

• Reframe the narrative. "I'm behind on a tool that keeps evolving, and nobody showed me." That version is true, and it is survivable.

• Master the art of vision. Picture the specific moment. A Thursday where you paste in your own draft, get three versions back, pick the good one, fix two words, and send it. The more detail you can picture, the better it holds.

• Take action. Choose the one task before you close this tab.

If you lead people, the same is true of anyone on your team who hasn't started. There's more on that in AI for Executives. On why sending them to training rarely fixes it, see Why Training Doesn't Change Behavior.

Frequently asked questions

How long does it take to feel confident using AI?

Less time than most people expect, because confidence here tracks repetitions rather than months. What moves it is opening the tool on work you already know, where you can judge the answer yourself. What doesn't move it is reading more about AI first. If you use one tool on one familiar task a couple of times a week, you'll know within a few weeks whether the hesitation is easing.

Is it cheating to use AI at work?

Not unless your employer has a policy that says so, and many people have never been told what theirs is. Check yours, because some roles have real restrictions around client data and regulated records. The research does show that people get judged on perceived AI use even when the work is identical. That's a reason to be deliberate about how you describe what you did, not a reason to hide it.

Should I tell my manager I used AI?

In most workplaces, yes, and the framing matters more than the disclosure. Saying "I used AI" invites a judgment about you. Saying "I drafted this with AI and then verified the numbers and rewrote the recommendation" describes your judgment, which is what your manager is evaluating anyway. If your team has no shared norm yet, someone has to go first.

What if my company hasn't given me an AI tool?

You are in the majority. Fewer than half of employees say their organization has integrated AI tools at all, and only about a quarter say they were given a clear plan. Ask two questions rather than waiting: what are we allowed to use, and what data must never go into it. Those two answers unblock most people. If nobody owns the answer, that is worth knowing, and worth naming to your manager.

Do I need to learn prompt engineering?

No. Prompting techniques change faster than you can learn them, and most of what works is describing your task clearly, giving an example of what good looks like, and saying what you want back. If you can brief a capable new colleague, you can brief a model. The rest is practice on your own work.

Where to start this week

Pick one task you already do well. Use the tool on it twice before Friday. Write your own version first each time, then compare.

If you want a structured way through it, FYB | AI Confidence walks you through exactly this, built for people who are good at their job and quietly stuck. It's free.

About this article. Tiffany Whitney is a NeuroLeadership-trained brain-based coach and the founder of Feed Your Brain. She spent 18 years as a Senior Director of Learning & Development inside a Fortune 500 company, where she led national rollouts reaching thousands of employees and customers, and she is the Amazon bestselling co-author of Strength with Kathy Ireland. Every statistic here was checked against its primary source, and the population and conditions behind each figure are stated in the text. Where a figure comes from an organization's internal data rather than independent research, that is noted. This article is general professional guidance and is not medical, psychological or clinical advice.

The AGES Model™ is a trademark of the NeuroLeadership Institute.