I'm offering you four specific ways to get more out of AI: better prompting, improving AI memory, starting a business using AI and running a business using AI (tonight).
Congrats! You started a business using AI. Now, you've got to run it. AI can help with that, too.
- The old rule: After launch, the hiring surge and spiral begins. Every hire slows the business down before it runs.
- The new rule: The next generation of companies will be designed before they're staffed. You can use AI agents to execute a lot of the work. You supervise outcomes, not big teams of people, until business is rolling in.
Why it matters: This could be the real jobs story of the decade — perhaps even bigger than "AI takes your job." The same technology that threatens millions of existing roles can create a wave of small, profitable, lower-headcount companies that couldn't have existed five years ago.
- Our guess is both happen at once: an explosion of new startups alongside the destruction of millions of existing white-collar jobs.
- The loss will likely be more acute than the gain — at least in the short run. But if startups truly surge and operate at lower costs and higher margins, this would be a huge win.
Remember the three buckets. Every business, regardless of model, breaks into them. AI will soon handle all three better, faster and cheaper than a generalist team.
- These buckets go beyond the research, design and analysis that AI does quite well already.
The front office handles external engagement.
- The picture: 6:47 a.m. Monday. The AI agent has already pulled the weekend's inbound leads, enriched each from LinkedIn and their company site and drafted personalized follow-ups in your voice. By the time you open your laptop, three are flagged as worth a personal call.
- Your job: Review the 10 emails the agent almost sent to your top accounts. 15 minutes, not a headcount.
The back office manages internal friction.
- The picture: A client signs. The agent triggers the onboarding packet, generates the first invoice, books the kickoff call and adds the project to your management to-do list. If something stalls, it pings Slack. Month-end books close themselves, with a memo flagging the three anomalies you actually need to look at.
- Your job: Design the workflow once, then only touch the exceptions.
The intelligence layer is where decisions are made.
- The picture: Sales data and customer feedback flow into one place. Every Monday, you review real-time dashboards and you get a one-page memo: "Two power users went cold last week. Three accounts spiked. Here's what I'd test."
- Your job: Decide if the pattern the AI spotted actually matters to where you're taking the business.
Everything described here can be done with agents that most people can utilize with a small amount of training and draft off Claude, ChatGPT and Gemini, as well as much cheaper-to-run open-source AI tools.
What doesn't change: You don't lose value when machines do the work. You migrate it. You stop being a "manager of do-ers" and become an "architect of systems." The humans who win excel at the four things machines can't touch:
- Judgment: Knowing what's worth building, selling or shutting down.
- Relationships: The human-to-human trust that customers won't give a bot.
- Synthesis: Spotting the edge cases and knowing which are signals and which are noise.
- Taste: Separating "good enough to ship" from "this will embarrass us."
If you consider yourself a systems architect and think we missed something, shoot Jim a note: finishline@axios.com.
📈 If you're a CEO or on a CEO's team: Ask to join Jim's new weekly Axios C-Suite newsletter.
