3 November 2026OnDraft

Insight

Put AI on real work now, with a person checking it

Don't wait for AI you can leave alone. Give it a real job, have a person check it, and keep every correction.

On Monday a US robotics company called Tutor Intelligence announced a $75 million funding round led by Andreessen Horowitz. Its robots pick, pack and move goods in warehouses and factories.

What caught my eye was how they get the robots working. Andreessen Horowitz put it like this: "They deploy into real production environments ahead of full autonomy, using human teleoperators ("tutors") to train the robot foundation models and make the fleet more capable over time."

In plain terms, the robots go to work before they can do the whole job alone. A person watches remotely and helps when one gets stuck. Every bit of help becomes training data, so the robots need less help over time.

The company says its robots are now in almost four times as many facilities as a year ago, and nearly a third of customers who got robots in the past year were existing customers adding more.

What this means if you're not buying robots

Most small businesses are doing one of two things with AI. Some let it loose on things it shouldn't touch yet. More are waiting until it's good enough to leave alone.

Tutor's approach sits in between, and I think it's the right one. Give AI a real job now. Have a person check its work. Keep every correction.

Pick a job it can't finish without you

Start with work where the AI does most of it and a person signs it off. A draft purchase order. A list of overdue invoices from Xero. A stock count that needs approving before it changes anything. I've written before about writing down what your AI can touch, and that list is where to start.

The best first jobs are usually the dull ones. More on that in AI on the boring jobs.

Make the check quick

If checking takes longer than doing the job, nobody will keep it up. Ask the AI to show where each answer came from, so a person can check it in seconds. I covered that in make AI show its working.

Keep the corrections

This is the bit most people miss. When someone fixes the same mistake twice, it's a rule the AI should have been given. Write it down and give it to the AI.

After a few weeks you've got two things. An AI that needs less checking, and a written record of how your business actually works. Most firms have never had the second one. It's what we're building into Bud: the AI learns how the business works from the people who run it.

Price it by the job

Tutor's customers pay by the hour or by the pick, with no upfront purchase. That's a useful way to look at any AI tool. What does it cost for each job it does, compared with doing it by hand? More in price AI by the job.

The short version

Don't wait for AI you can leave alone. Put it on a real job with a person checking, and keep what they correct. That's how it gets good enough to trust.

If you want help choosing the first job, give me a shout.

Sources

  • Tutor Intelligence via Business Wire, 2 Nov 2026: businesswire.com/news/home/20261102314852/en/
  • Andreessen Horowitz, Investing in Tutor Intelligence, 2 Nov 2026: a16z.com/announcement/investing-in-tutor-intelligence/
  • The Next Web, 2 Nov 2026: thenextweb.com/news/tutor-intelligence-75m-series-b-a16z-robot-workers

Show us the bottleneck →

We build systems that power growth.

AI agents, integrations and operational software—built around your business.