Claude AI at Work: Two Habits That Save You Real Hours
The most common thing I hear in class is some version of this: “I tried the AI, it was impressive for about ten minutes, and then I went back to doing it the old way.” That is almost always a setup problem, not a tool problem. Claude AI at work only starts paying off when you stop treating it like a search box you visit once and start treating it like a coworker who already knows your job.
I have watched people retype the same background paragraph at the top of every single chat — who the client is, what the project is, what the format has to be, who reads it. Twenty minutes a day, gone, before any actual work happens. And then they wonder why the output still reads like a press release written by nobody in particular.
Here are two things you can set up in about ten minutes each. Neither one requires you to attend anything.
Claude AI at Work Starts With a Project, Not a Chat
In the left-hand sidebar at claude.ai there is an item called Projects. Click it, then click New project. Name it after something real that you actually do — “Monthly Board Reporting”, “Anderson Account”, “HR Policy Rewrites”. Not “Test”.
Two things go inside it. The first is project knowledge: upload the background documents you would otherwise paste in every time. Last quarter’s report. The template your boss likes. The org chart. The signed contract. The second is custom instructions — a short block of standing rules. Here is roughly what I put in mine, and you can copy the shape of it:
I am the operations manager for a 120-person distribution company. Everything you write for me goes to executives who are not technical. Use short paragraphs. No bulleted list longer than five items. Never use the words robust, seamless, or leverage. If a number is not in the files I have given you, write [NOT IN SOURCE] instead of estimating it.
From then on, every new conversation you start inside that project inherits both the files and the rules automatically. You open a chat and say “draft the September summary” and it already knows what that means. That is the whole point: you stop paying the context tax at the start of every conversation.
The gotcha: project knowledge fills up, and Claude shows you a percentage as you add files. More importantly, it goes stale. If you loaded the Q2 report in June and never took it out, Claude will cheerfully quote Q2 numbers inside a Q4 document and it will look completely confident doing it. Open the project once a month and pull out anything that is no longer current.
Teach It Your Voice in About Fifteen Minutes
AI writing sounds like AI writing because nobody ever told it what you sound like. That is fixable, and it is a one-time job.
Find three pieces of writing that are genuinely yours: a real email you sent last week, a memo or a proposal, and something a bit more casual. Paste all three into one Claude conversation and type this:
Here are three things I wrote. Analyze them and write me a style guide describing how I write: typical sentence length, how formal I am, how I open and close, specific words and phrases I actually use, and things I never do. Be specific and quote examples from the samples. Do not flatter me.
What comes back is a page of observations, and some of it will be uncomfortably accurate. Mine told me I open without pleasantries, average about fourteen words a sentence, lean on “here is the thing” as a transition, and never use exclamation points. All true.
Now take that style guide and paste it into the custom instructions of the project you built in the first tip. Every draft from that point on starts from your voice instead of from nobody’s. The manual alternative is rewriting every generated draft from the ground up, which is why most people give up on this after a week.
The gotcha: feed it typical writing, not your best writing ever. If you hand it the one beautifully polished piece you slaved over, you get a style guide for a person you are not, and every draft comes back too formal to send. And read before you hit send — a style guide fixes the voice, it does not verify the facts.
Plenty more on the calendar
If you want to see all of this done live, the related session is Claude AI for Modern Professionals: Smarter Work in 2026, running Tuesday, September 1. We go through Projects and project knowledge, custom instructions, the voice exercise above, and working with documents and spreadsheets — on screen, with questions welcome as we go.
The full current schedule is on the front page of PCWebinars.com, with new dates going up as they are confirmed. Register for this one, or have a look through the others and find the one that fits what you are stuck on this month.
About your trainer
I’m Tom Fragale. I have been a computer trainer and consultant for more than 30 years, and I have taught well over 30,000 people how to get more out of the software they already own — Excel, Access, Power BI, Word, Outlook, and now the AI tools that sit on top of all of it: Claude, ChatGPT, and Microsoft Copilot. My background is applied, not theoretical. Alongside the teaching I build real databases and reporting systems for clients in Excel, Access, and Power BI, so the examples I use in class are the problems I ran into last week, not textbook exercises. I also publish free tutorials on YouTube and longer courses online.
Everything — the live schedule, the recordings, and the custom training options — is at PCWebinars.com.