← Journal
AI basics

The six things AI is actually good at

Almost every useful AI task falls into one of six types. Here is each one, translated for a small business.

By B2 Builds · 3 min read

Source: OpenAI, “Identifying and scaling AI use cases”

Six glowing sunset-coloured tiles, one for each type of AI work

OpenAI looked at hundreds of real workplace AI use cases and found that nearly all of them fit six patterns. That is a helpful shortcut: instead of asking “what could AI do for us?” you can look at your week and ask which of six jobs you already do by hand.

1. Content

First drafts of emails, posts, product descriptions, proposals and policies. Example: turning call notes into a follow-up email you edit in two minutes instead of writing from scratch.

2. Research

Gathering and summarizing information. Example: comparing three suppliers, or pulling the common complaints out of fifty reviews.

3. Coding

Not only for developers. Example: a spreadsheet formula, a small website fix or a script that renames a folder of files.

4. Data analysis

Making sense of numbers you already have. Example: asking which services brought in the most repeat customers last quarter, using an exported report.

5. Ideas and strategy

A thinking partner. Example: stress-testing a new package price, or brainstorming ten angles for a seasonal campaign before you pick one.

6. Automation

Work that repeats on a trigger. Example: a new inquiry creates a contact, assigns an owner and sends a confirmation without anyone copying it across.

Most weeks contain all six. The trick is noticing which ones you are still doing the slow way.

Where to start

Pick one recurring task in each type and write down how long it takes. The longest, most boring one is usually the best first experiment. Content and research tend to be the quickest wins; automation usually delivers the most lasting time back once tools are connected.

Sources

Explore our services ↗︎

More from the journal ↗︎

Your next step

Put the shift
to work.

Tell us what slows you down. We’ll find the right place to start.

Let’s talk ↗︎