AI for Your Second Act
You don't need to become "an AI person" for your next chapter. You need to aim a tool like Claude at the thirty years of scar tissue, wins, and pattern recognition you already own. The expertise stays yours; the grunt work goes to the machine.
Key Takeaways
- The core idea: Your decades of pattern recognition become the input, while AI handles the production layer that used to require a team or drain your weekends.
- Why it matters: If a competent associate could bill for the task, delegate the first pass to AI and reserve your time for the judgment clients actually pay for.
- How it works: Load a Claude Project with your best past work and frameworks once, then use it to draft every deliverable in your voice without starting from scratch.
- What to do: The expertise that took twenty years to build can be packaged into a sellable product in a single weekend when AI collapses the production bottleneck.
Source: The Leveraged Years Briefing. Permalink
Using AI for a second act means taking what already earns you $600 an hour and letting a system like Claude handle the parts a solid associate could do. You're not auditioning against 25-year-olds who code. Your edge is judgment, and AI is the production layer that lets one seasoned operator ship what used to take a small team. The real job is to feed it what you know, not to reinvent yourself from scratch.
I ignored this until early 2025. I'd skim the headlines, poke at a chatbot, then go back to my old stack. I assumed AI was for people starting at zero, and my problem was the opposite, too much accumulated context to throw away. I was wrong. What changed my mind was watching a former CFO friend turn a single bespoke client memo into a recurring, productized service over one weekend, using nothing but Claude and his own archive. That's the move this piece is about.
Why your experience is the asset, not the obstacle
The standard AI story is written for beginners: learn the tool, build a skill, start a thing. For someone 45 to 62, that framing is backwards. You don't lack a skill. You carry thirty years of pattern recognition no model has and no junior can fake. The 25-year-old's problem is "I don't know anything yet, what do I do with this tool?" Yours is "I know a lot, how do I get it out of my head and into something people can buy without burning out?" That's a far better problem, and AI is finally good at it.
Stop picturing AI as a competitor and start treating it like a roomful of fast junior staff who don't sleep. The bottleneck was never your expertise. It was the production layer, the writing, the formatting, the research, the fifteenth retelling of the same explanation, that turned one good idea into three lost days. That's exactly the layer Claude collapses. A diligence framework you carry in your head becomes a documented process in an afternoon. You still make the calls; you just stop hand-building every artifact around them.
Which second-act paths does AI actually accelerate?
Some moves get a real lift from AI. Others barely change. Here's where I see the difference.
| Second-act path | What actually changes | Realistic difference |
|---|---|---|
| Solo consulting / advisory | Drafts proposals, research, and deliverables from your frameworks and call notes | An 8-10 hour proposal drops to 2-3; one person looks like a three-person shop |
| Productized expertise (course, playbook) | Turns scattered SOPs, decks, and war stories into a structured, sellable product | A 4-6 week "someday" project becomes a focused weekend of editing and decisions |
| Fractional executive work | Preps board decks, redlines drafts, and writes memos between engagements | Room for 3-4 clients where prep used to cap you at 1-2 |
| Writing / authority building | Turns raw notes and scratch thoughts into essays in your tone, for editing | A weekly piece becomes a 90-minute edit, not a lost Saturday |
The "could I bill for this?" test
Here's the one filter that's held up. If a competent associate could do the task and you'd bill a client for it, AI gets the first pass. If the client is paying for your judgment specifically, you keep it and use Claude only as a sparring partner.
Drafting a market scan or a first-cut model? Associate-grade work, hand it over. Deciding whether the client should actually enter that market or restructure the deal? That's the $600-an-hour call they hired you for, and it stays with you. The test keeps you out of two ditches: clinging to grunt work you should delegate, and outsourcing the one thing that makes you hard to replace.
How to start this week, with work already on your desk
Don't take a course first. Don't watch tutorials. You need one real deliverable and about ninety focused minutes, because the fastest way to learn this is on something you already owe someone.
- Set up a context Project in Claude. Create a Project named for your practice and load it with your bio, two or three of your strongest past deliverables, and a framework you lean on. I loaded mine with two decks and four old proposals; it took 45 minutes. Every output since has sounded like me, not a generic chatbot. This is the highest-return hour you'll spend.
- Pick one live deliverable you owe this week, a proposal, a board memo, an analysis, and have Claude produce a first draft from your material. Time it. Most people clock a 60-to-75% cut on the first try.
- Find the one repeatable thing you do for clients and ask Claude to help turn it into a template or a productized offer. This is where a job quietly becomes an asset.
- Edit hard. The first draft is raw material, not the deliverable. Your judgment about what to keep and what to kill is the value you add, and the reason it still sounds like you.
A labeled scenario: the operations VP who built a product in a weekend
A composite from the kind of case I see weekly. A 54-year-old former operations VP, recently out of a manufacturing role, has a supply-chain risk assessment she's refined across a dozen turnarounds. It lives in her head and a few old decks. Over one weekend she loads the decks and her notes into a Claude Project and works through it: structure the methodology, draft a 30-page playbook, build a client-facing diagnostic, write the sales page.
It wasn't frictionless. Her first pass at the diagnostic was generic, because she'd fed Claude conclusions without the messy judgment behind them. She went back, added the three war stories that actually shaped the method, and the second pass was sharp. By Monday she had a $7,500 productized assessment she can sell on repeat. The expertise took twenty years; the packaging took two days, because the production layer that once needed a writer, a designer, and a month was a tool and a weekend. She didn't learn a new field. She finally made the one she already owned sellable.
Is it too late to start with AI in your 50s?
No, and the premise is backwards. The people struggling most with AI right now are often the youngest, because the binding constraint isn't tool fluency, it's having something worth saying to point the tool at. You spent decades building exactly that. Learning Claude well enough to be dangerous takes a few focused hours. Building the expertise it amplifies took you a career. You're not behind; you're holding the part that can't be downloaded.
So what do you do Monday morning?
Pick the one deliverable on your desk you dread most, the proposal, the board memo, the analysis you keep pushing to next week. Open Claude, give it your context and your raw material, and let it take the first pass. Then edit it like the expert you are. You'll have your real answer about AI and your second act before lunch, and it'll be a better answer than any tutorial could give you. The next chapter doesn't start when you finish learning the tool. It starts the first time you put what you already know to work through it.
Frequently Asked Questions
How do I know which parts of my work to hand to AI without losing what makes me valuable?
Apply the associate test: if you would bill a client for the task but a competent junior could execute it, give AI the first draft. Reserve your time for the judgment calls and strategic decisions that justify your rate.
I have decades of expertise scattered across old files and memory. Where do I actually start?
Create one Claude Project, load it with your bio and two to four strong past deliverables, then use it to draft one real thing you owe this week. Time the difference and refine from there.
Is it worth learning this if I'm already established and my current approach works?
The question is whether you want to stay at your current client capacity or expand it without hiring. AI lets one person deliver what used to require a small team, turning expertise into repeatable products instead of one-off hours.