Claude at Work

Nobody uses the AI you bought? Here's the fix

Over half of workers skip their company's AI tools. The teams that keep them train one person on one job. The data and the 30-minute fix.

Dominik · Published · Updated

TL;DR

More than half of workers skipped their company's AI tools in the last month. The fix is not another tool. Train one person on one weekly job, 30 minutes, with a 10-second check on every output. In the LSE and Protiviti study, trained employees used AI 93% of the time and saved 11 hours a week, against 57% and 5 hours untrained.

Why is nobody using the AI tools we bought?

Nobody uses the AI tools you bought because you bought a tool but never assigned a job. This is normal, not a failure: a 2026 survey found more than half of workers skipped their company’s AI tools in the last month and worked by hand. Tools spread; tasks do not. Adoption follows a job, not a login.

Is it normal that our team stopped using AI?

Yes, it is normal that your team stopped using AI. A 2026 Fortune-reported survey found 54% of workers bypassed their company’s AI tools in the past 30 days and 33% never used them at all. US Census data showed large-firm adoption actually fell from a 14% peak to 12% in late 2025, the first measured decline. Abandonment is the pattern, not the exception.

What makes a team actually keep using AI?

Training makes a team keep using AI, and it costs far less than the tools. In the LSE × Protiviti study (October 2025), employees who got even a little AI training used it 93% of the time, versus 57% for the untrained, and saved 11 hours a week against 5. Same tools, double the return. The gap is teaching, not technology.

How do I get my team to use the AI we already pay for?

Get your team to use AI by giving one person one job, not a tool tour. Pick the single weekly task they dislike most. Spend 30 minutes showing them how the AI you already pay for does exactly that, then add a 10-second check: is the name right, is the date right? A repeatable job with a quick check beats another subscription. If you want help picking the task, the free AI-visibility audit is a place to start.

Should we buy more AI tools to fix low usage?

No, buying more AI tools rarely fixes low usage. The median small business already runs about five AI tools (Business.com, 2026), and more logins do not create more usage. Each new tool adds another tab nobody opens. The fix is narrower, not wider: one trained person, one weekly job, one quick check before anything is sent.

What should AI not do in my business?

AI should not make your judgment calls. Keep it away from pricing and hiring, where a wrong answer is expensive and hard to catch. The same goes for anything with legal weight. Point it at the boring, repeating work instead: drafting follow-up emails or turning messy notes into a clean summary. Review every output briefly. Treat it like a fast intern, not an oracle.

Sources: LSE Inclusion Initiative × Protiviti, Bridging the Generational AI Gap (28 Oct 2025); Fortune, AI backlash survey report (9 Apr 2026); US Census Bureau, Business Trends and Outlook Survey (2025–26); Business.com 2026 SMB AI Outlook

FAQ

Why did our team abandon the AI tools we paid for?

Most teams abandon AI because the tool arrived without a job. A 2026 survey found 54% of workers bypassed company AI tools in a month and 33% never used them. Usage follows a specific, repeated task, not a subscription, so a tool with no assigned job quietly goes unused.

Does AI training really change usage that much?

Yes. In the LSE × Protiviti study (October 2025), trained employees used AI 93% of the time versus 57% untrained, and saved 11 hours a week against 5. Training here is not a course. Thirty minutes showing one person one job is enough to move usage and time saved.

How many AI tools does a small business actually need?

Most small businesses already have enough. The median SMB runs about five AI tools (Business.com, 2026), so low usage is rarely a tool-count problem. One tool doing one weekly job well, with a quick accuracy check, returns more than a sixth subscription nobody opens.

What is the fastest way to get value from AI we already own?

The fastest way is to pick one weekly task someone dislikes, show one person how your existing AI does it, and add a 10-second check for the name and date before sending. This turns a dormant tool into a repeatable habit without new spend or a big rollout.