Using AI too little, trusting it too much: what the GoTo study shows
Employees underuse AI and trust it too much, GoTo's 2026 study finds. Why more pressure won't fix that, and why checking has to be part of the workflow itself.
Many companies make two AI mistakes at once: they use it too little, and where they do use it, they trust it too much. The "Pulse of Work 2026" study by GoTo and Workplace Intelligence shows both in the same set of numbers. 43% of respondents used AI output even though they suspected it was wrong. At the same time, 80% say they don't use AI to its full potential. More pressure fixes neither problem. The workflow does: whoever passes on an AI result has checked it, and whatever a system can check on its own, it checks first.
Chapter 1 of 6
The study: 2,500 respondents in ten countries.
GoTo published the study with Workplace Intelligence in May 2026. Between November 2025 and January 2026, they surveyed about 2,500 employees and IT leaders in ten countries: the US, Canada, the UK, Ireland, Germany, Austria, Switzerland, India, Mexico and Brazil.
| Finding | Share |
|---|---|
| used AI output they suspected was wrong | 43% |
| say they don't use AI to its full potential | 80% |
| feel pressure to use AI to be more productive | 60% |
| IT leaders whose company has an AI policy | 44% |
| say checking AI output takes more effort than checking work done by people | 77% |
Source: GoTo and Workplace Intelligence, "The Pulse of Work in 2026". The numbers show a direction. Only your own usage data shows what it looks like in your company.
Chapter 2 of 6
Too little and too much aren't a contradiction.
The same workforce uses AI too little and trusts it too much.
At first glance the numbers don't fit together. People who use AI too little should be careful with it. In fact, the same workforce uses AI too little and trusts it too much. Often nobody touches the tool for routine work, because nobody has shown what it's good for. Where it does get used, there's no rule about who checks the result.
These are the two mistakes I see most often in companies, and they usually share a cause: AI gets handed out as a tool while the workflow stays the way it was. If you steer by usage numbers alone, you only see half the problem. For each workflow, settle what AI should be used for and who checks the result.
Chapter 3 of 6
Pressure creates usage, just not good usage.
60% feel pressure to use more AI. Pressure raises the number of attempts. It doesn't raise the quality of the results. In my experience it lowers it: people who have to deliver faster check less. It fits that only 44% of IT leaders report an AI policy in their own company.
So if you ask your people to use more AI, you also have to say who checks a result before it moves on. For me that's always the person who passes it on. They own it, whether they wrote it or a model did. What checking looks like in practice is in Quality control for AI output. Pair every push for more AI with a clear rule for checking.
Chapter 4 of 6
Checking can't depend on one person's patience.
77% say checking AI output takes more effort than checking work done by people. That matches practice. A model writes fluently even when it's wrong, and skimming won't catch the error. When checking depends only on one person's attention, it's the first thing to slip under time pressure.
That's why checking belongs in the workflow. Whatever can be checked automatically, the system checks first, such as format, required fields or a comparison with data you already have. At SUMAX, every AI call runs through our own layer between our applications and the models. A spot like that lets you build a check once instead of rebuilding it in every tool. A named person decides on the rest. The machine check belongs before the human one, so the person can focus on expert judgment.
Chapter 5 of 6
Where the study falls short.
GoTo sells IT and communication software itself. The survey ended in January 2026, and everything is self-reported. Suspecting an error doesn't mean there was one. The direction still matches what I see in companies. Use the study as a reason to write down your own checking rule.
Chapter 6 of 6