BBG Owner's Manual

■ · People & Tech

Technology and AI: tools do the routine, people do the judgment

Software and AI can take over a lot of repetitive work in a small business. But a computer is only as good as the person using it. The goal is a business where technology handles the routine and people spend their time on what machines can't do.

4 min read

What to hand off

Much of the repetitive work in a small business can now be handled by software, and some of it by AI tools that you set up and train on your own processes. The best candidates are tasks that follow the same steps every time and are easy to check.

  • Data entry and sorting: pulling numbers from receipts and invoices, categorizing transactions.
  • Scheduling and reminders: booking, follow-ups, payment reminders.
  • First drafts: emails, proposals, reports and social posts that a person then reviews.
  • Answering common questions: hours, pricing, how-to questions, with a person handling anything unusual.
  • Summaries: meeting notes, long documents, client histories.

How to "train" AI for your business

For most small businesses, training AI doesn't mean building your own model. It means giving an AI tool the right instructions and examples so it does a task your way.

  1. Write down the process first. If you can't explain the task step by step, AI can't do it either. This is where your written procedures pay off.
  2. Give it examples of good work. Past emails, finished reports or correct entries show it what "right" looks like.
  3. Start with one small task. Test it on real work, compare with what a person would do, and fix the instructions.
  4. Keep a person reviewing the output, especially anything that goes to a client, involves money, or touches private information.
  5. Protect client data. Check what each tool does with the information you put in before you use it for anything confidential.

■ The human edge

You still need people. Better ones, doing better work.

A computer is only as good as the person using it. Software follows instructions, and AI can produce confident answers that are wrong. Someone has to set it up, check its work, catch the mistakes and take responsibility for the result. The goal of technology isn't a business with no people. It's a business where people stop doing work a machine can do, and spend their time on what machines can't.

Hand to technology

Repetitive, rule-based work

Examples
Data entry, scheduling, reminders, sorting, first drafts, routine answers.
Why
Same steps every time, easy to check, low judgment.
Still needs
A person to set it up and spot-check it.
Keep with people

Judgment, trust and relationships

Examples
Client relationships, hard conversations, final decisions, ethics, unusual problems, selling.
Why
Clients trust people, not software. Responsibility can't be handed to a tool.
Leverage
Let technology do the prep so people spend more time here.
The research

What studies show about AI at work

These are real studies with real numbers, from specific settings. They won't transfer exactly to your business, but they show where the gains tend to be and where the risks are.

AI helps most on routine work

In a study of 5,179 customer support agents, an AI assistant raised issues resolved per hour by 14% on average and by 34% for newer, less-skilled agents, with little effect on the most experienced.

Brynjolfsson, Li & Raymond, NBER Working Paper 31161
Writing tasks get faster

In an experiment with 453 college-educated professionals, ChatGPT cut the time to complete writing tasks by about 40%, and evaluators rated the quality 18% higher.

Noy & Zhang, 2023, Science
People still have to check

In a study of 758 BCG consultants, those using AI finished suitable tasks about 25% faster with higher quality. But on a task outside what the AI handled well, they were about 19 percentage points less likely to get the right answer than consultants without it.

Dell'Acqua et al., 2023, Harvard Business School Working Paper 24-013

The pattern across all three: AI helps most on routine, well-defined work and with less experienced staff, and it can make people worse on work it doesn't handle well. That's why a person always checks.

Sources

Before you buy any tool

  • Run the Dollar Test. Hours saved only count if they go to paying work.
  • Start with one task, not a whole system. Prove it on a small scale, the way the pivot timeline does in days 31–60.
  • Pick who checks it. Every automated task has a named person who reviews the output.
  • Set a tripwire. If it isn't saving the time you expected by day 60, drop it.
  • Plan the training. A tool nobody knows how to use saves nothing.

When technology changes which roles you need, see the Role Review on the Pivot page.