TL;DR: AI is great at exactly the work most people are bored doing. Pattern-based, bounded, repeatable, low-stakes, and easy for a human to review afterward. Stop asking it to write your book or replace your customer service team and start using it for the ten categories of work it handles well. Every one of them has the same shape: routine repetition that a human can check at the end. More on how I use AI on a book. Those categories produce the time savings and build the augmented human.
Most of the conversation about AI is about what it can’t do. That’s because most of what’s getting tried right now is the part it can’t do, and people are watching the failures pile up in public.
I get tired of the conversation being all or nothing. AI is a tool, and it’s very good at the dull, repeatable work nobody wants to do and bad at the work that makes a book worth reading. Put it in the first lane and you’ll get hours back every week. Push it into the second and you’ll get prose nobody wants to finish. These are the ten lanes where I think it’s worth using.
1. Summarizing long documents
You have a 200-page report, a recorded meeting, a stack of emails, a research paper, a contract.
You don’t have three hours. AI reads the full document in seconds and gives you a workable summary, the main points, the action items, what the document is really about underneath the formal language. For more, see AI in customer service.
Modern models handle this well enough that the summary is usually accurate on the structure, even when it gets specific facts wrong. It’s a fine starting point as long as you read the original for any claim you intend to act on.
People skip that last part, and it’s the part that burns them. They read the summary, never open the original, and quote the summary in a meeting as if it were the source. I’d tell any client what I tell myself: the summary gets you oriented, and the document is what you act on.
AI is good at the work hiding in a 200-page report, a recorded meeting, a stack of emails.Share on X
This is the single highest-value use of AI for most working professionals, and almost nobody talks about it because it’s boring. For more, see what I actually tell clients about AI.
2. Organizing interview transcripts
I work with clients who have hours of recorded interviews about their life or business.
Before AI, organizing those into a usable structure was a week of work. Now it’s an afternoon. The machine reads every transcript, identifies the recurring themes, finds the moments where the speaker said something they didn’t say elsewhere, and groups everything by topic. The writer still does the writing. The machine just made the raw material navigable.
That’s how I use it every day. Transcripts, session summaries and first-pass outlines get built with AI’s help, and then I do the writing. It saves me hours of sorting, and the sentences the client is paying for are still mine.
3. Generating first-draft outlines
Outlines are easy to start and hard to finish. The hardest part is the blank page. AI handles the blank page well.
You give it a topic, your audience, your three or four core points, and it returns a structured outline with sections, sub-points, and a logical flow. The outline is rarely the one you’d write yourself. It’s almost always good enough to react against, and reacting is faster than starting.
4. Cleaning data and reformatting spreadsheets
Say you have a list of 500 customer records with inconsistent capitalization, mixed date formats, and three different ways of writing the same company name. AI handles this in one prompt. Same for converting a wall of unformatted text into a clean table. Same for finding duplicates in a list that’s typos. The pattern is bounded, the rules are clear, and the output is easy to spot-check.
I worked with a client whose warehouse staff wear AR glasses with AI overlay.
Pickers see a glowing arrow pointing at the exact box they need, can “open” the box virtually to verify the contents, and get real-time correction if they grab the wrong one. Their pick rate doubled. Nobody lost a job. The same crew now does far more interesting work because the system handles the boring search-and-verify part.
That’s the augmented warehouse picker, and it’s the same pattern as the augmented spreadsheet analyst. The boring repeating work disappears. The human now spends their time on the work that pays.
5. Drafting routine communications
Internal status updates. Meeting recap emails. Polite-but-firm follow-ups to the vendor who hasn’t sent the invoice.
Templates for the dozen recurring email situations that drain your week. AI drafts these in seconds, in your voice if you give it samples, and the draft is almost always 80 percent of what you would have written. You edit the last 20 percent, hit send, and reclaim the hour.
This works because the bar is low. Internal communications don’t have to be brilliant. They have to be clear, polite, and arrive. AI clears that bar with no trouble.
I’d draw a hard line here, though. A condolence note, an apology to a customer you let down, a letter to an employee about something that matters to them: none of that is routine, and people can tell when a machine wrote it. Save the machine for the vendor who owes you an invoice.
6. Translating between technical and plain language
Maybe you have a technical document and an executive who needs to understand it.
Or you have an executive summary and an engineer who needs the details. AI translates between registers with high accuracy. Same content, different vocabulary, different sentence structure. This is a hugely useful application for anyone who works across departments or audiences.
The caveat is that the translation can flatten nuance, especially in the technical-to-plain direction, so for anything stake-bearing you have a real expert sanity-check the output. For routine cross-team communication, the machine does the work fine.
7. Generating test cases and edge-case lists
Building a new product, process, or system? Ask AI to generate everything that could go wrong. The output is a checklist of failure modes and edge cases, some obvious, some you wouldn’t have thought of. You won’t use all of them. You’ll catch issues you would have missed. This is one of the few places where the AI’s tendency to be exhaustive becomes an asset. You want every angle, even the unlikely ones, because the unlikely one is the one that finds your customer in six months.
8. Code completion and review
Developers are getting measurable productivity gains from AI code assistants. The pattern that works is the assistant suggesting the next line or block, the developer accepting or rejecting, and the developer staying in charge of the architecture and the hard parts. The assistant handles the boilerplate. The developer handles the thinking.
I got tired of the booking services I’d been using, because none of them did what I wanted. So I told Claude I needed a plugin that did this, and this, and this, and gave it a screenshot. About ten minutes later I had the plugin. It sets up Riverside automatically. I didn’t have to debug anything. I said you forgot this, I want that, change this, and it did.
The pattern that doesn’t work is letting the AI write the architecture, the security-sensitive code, or anything that touches user data without a human reviewing it.
I think that’s reckless. When that code fails, nobody on the team understands it well enough to fix it fast, and the customers whose data sat behind it pay for the gap. Keep the assistant on work a person can check line by line.
9. Research synthesis
From the podcast
Diana Lee points to something the tool offers that a human advisor structurally cannot, on Leaders and Their Stories:
Turning to AI who doesn’t judge you, who doesn’t make any judgments and being able to be so transparent with what you’re struggling with and asking it to help you bridge that gap.For a senior person expected to have the answers, that absence of judgment is the feature that makes the honest question possible.
You’ve got ten sources on the same topic and you need to understand what they all say.
AI reads them, pulls out the consensus, the disagreements, the gaps, and gives you a synthesis with citations back to each source. You verify the citations match the source claims. That catches the small percentage of misrepresentations. Then you walk into your next meeting with a working understanding of a field you couldn’t have absorbed in a day on your own.
For a writer working on a book, this is the difference between a research phase that takes six months and one that takes six weeks. A human still has to write the book. The reading and sorting is now offloaded.
10. Brainstorming when you’re stuck
Sometimes you have a problem, you’ve stared at it for an hour, and you’re not making progress. AI is excellent at generating possibilities you haven’t considered. Most of them will be wrong or impractical. One or two will be the thing you needed to hear to break the logjam.
This works because the AI has read more about your problem than you have, even if it understands less. It pattern-matches against everything everyone has ever written about adjacent problems. The cost of getting twenty bad ideas is low. The value of one good one is high. Net positive.
What do AI’s best uses have in common?
Every category on this list has the same shape: repetitive work with a bounded pattern and output a human can review before it leaves the building. AI saves time on work shaped like that, and that’s how the augmented human gets built.
The 71-year-old client I work with on a memoir uses AI exactly this way. He runs his manuscript through it to find weak chapters and inconsistencies. When he can’t remember if a detail belongs in chapter three or chapter seven, he uses it to organize his thoughts. He also has it generate notes for me about what he wants me to focus on next. He doesn’t let it write a word of his book.
The machine handles the routine, and the human handles everything it can’t. The Birth of the Augmented Human covers the broader shape of what that looks like across every profession. If you’re not getting this kind of use out of AI, you’re either using it wrong or using it for the part of your work where it doesn’t belong. The companion piece, Six Places AI Will Break Your Work, walks through where the other line falls.
I think most of the disappointment with AI comes from pointing it at the wrong job. People ask it for the book, the strategy or the heartfelt apology, get something flat back, and decide AI is hype. If you can’t afford help, writing with AI is fine, and it beats leaving the book unwritten. Just know which parts of the work need you, and put the machine on everything else.
