What AI means for your staff depends on when you start

Since March, my calendar shows 323 meetings with 224 different people. Most of my work with owners comes back to what AI means for their people.

My answer: your team is going to be affected by AI either way. What you control is whether you get to decide how.

What changed in a year

On OpenAI's GDPval test, experts grade AI work blind against professionals with 14 years' experience on average. In September 2025 the best AI matched or beat the professional 47.6% of the time. By April 2026, OpenAI reported 84.9%.

Those are single tasks with a full brief. Whole jobs are harder: on real freelance projects, the best AI met the client's standard 2.5% of the time in October 2025 and about 16% by July 2026. That gap is why setup matters. AI needs what a new hire needs: examples of good work, the rules, and a person checking the result.

Heavy AI users are hiring

Ramp and Revelio Labs looked at more than 21,000 US companies. Companies that adopted AI grew headcount 10.2% in the two years after, and all of that growth came from the heavy users. At those firms, entry-level headcount rose 12%. Companies that dabbled saw no significant change. To be fair, the companies that adopted were already larger and growing faster, and Ramp calls the results early.

I see something similar up close. A custom home builder I work with now drafts his proposals with Claude, and the writing went from about 40 hours to about 4. He's winning more of the work he bids on and he's busier than he's ever been. For him, the bottleneck moved from paperwork to finding trades to do the builds.

What early and late look like

Picture a team of five in your office, the people handling proposals, scheduling and billing, where AI now does most of the paperwork. That freed-up time has to go into work that matters to the business, or it starts to look like a cost you can cut. If you move early, you have room to work that out. People can take on the work you never had capacity for, spend more time with clients, raise the quality of what you deliver, or get time back in their day where the business can afford it.

If you wait two years, a competitor may already be quoting faster, following up better and pricing lower. You're losing bids. The decision about your people gets made under pressure, and under pressure cutting staff is the quickest way to find margin.

Two papers this year support parts of this. One argues that the faster adoption happens, the more displaced workers leave the workforce for good. The other models how competitive pressure pushes every firm to automate, even when they'd all be better off holding back. The part where laggards end up cutting is still my read more than measured fact. So I don't think waiting protects your people. It moves the decision to a time when you have fewer options.

Why most businesses haven't moved

Setup takes effort: writing down how the work gets done and connecting the systems it runs on. With big general-purpose technologies like computers, that investment came before the gains showed up.

The other reason is that most people are planning around the AI they tried a year ago. In my meetings, I'd estimate most people still quote AI numbers that are six months or more out of date. One number I'd retire is the claim that 95% of AI pilots fail (it counted pilots with no measurable profit impact). That came from an MIT report covered in August 2025, and even then the report blamed how companies set AI up rather than the models themselves.

If you're not sure how far your own firm has come, the four levels of AI adoption gives you a quick way to check.

What to do in the next 30 days

Pick three to five people others already go to for help. They don't need to be senior or technical.
Pick the work that piles up on them, like proposals, follow-ups or reporting.
Give the AI what you'd give a new hire: examples of good work, your rules, and access to the right systems. Use a business account your IT provider has approved, and decide up front what client information stays out.
Note how long that work takes now and how good it is, then check again at day 30. Ask the people doing it whether their week got better.
Decide where the freed-up time goes: new work, quality, client experience, or time back where you can afford it. If the work isn't faster or better at day 30, change the setup or stop. If it is, add the next group.

This is the work I do with owners: set up that first group, coordinate with your IT provider, measure what changed, and get your team to the point where they run it without me. Your leadership stays in charge of the decisions. You can also run it yourself.

If you run a team and you're trying to figure out where the freed-up time should go, our article on where freed time goes walks through it. To check whether there's paid work for those hours on one task, try the hours test. It takes about 10 minutes. Or tell me what your team does and I'll reply with how I'd think about it.