Stop Asking What AI Can Do, Start Asking What Work Needs to Be Done

This week I want to tackle a question I hear constantly from business owners, and I think it's actually the wrong question to be asking: "what can AI do in my business?" It feels reasonable given how fast AI capability is moving, but starting there makes AI the center of the conversation when it should be the work. Your business exists because of outcomes that need to be produced, not because of a tool that needs a job to do. This is the first part of what I call the AI Labor Architecture framework in my new book, The AI Workforce, and it's the starting point for every other decision that follows. If you try nothing else this week, pick one part of your business and ask what actually needs to happen there, not what an AI agent could do, and not even who's currently doing it. Episode Timeline & Highlights \[0:00] – The common question business owners keep asking, and why it's the wrong one \[0:57] – Why AI isn't the center of your business, the work is \[1:28] – Reversing the question: from "what can AI do" to "what work actually needs to be done" \[2:00] – Introducing the SaaS onboarding example \[2:33] – What a CEO sees when a capable new AI agent looks able to handle onboarding \[3:39] – Why "could she do onboarding" is still the wrong question, even if the answer is yes \[4:12] – Breaking onboarding down piece by piece: information gathering, account setup, and interpretation \[4:41] – What actually happens after setup: helping the customer understand what's been done \[5:10] – The follow-up work: noticing when a customer gets stuck or disappears \[5:49] – Why "onboarding" was always a bundle of very different kinds of work sitting inside one role \[6:34] – A break to highlight smrtPhone, the show's sponsor, and its 5,000 free calling minutes offer \[7:07] – Introducing AI labor as a way to unbundle execution from any one person \[7:37] – What this actually looks like: an AI agent gathering information and following up consistently \[8:06] – When a person still needs to be brought in, and why they'd have full context already assembled \[8:47] – Why capability doesn't tell you how your business should be designed \[9:18] – The real questions to start with: where judgment, consistency, context, and relationship matter most \[9:57] – Why so much of the current AI conversation gets the order backwards \[10:24] – The hiring analogy: choosing a job for someone because of their resume, not their fit \[11:31] – Why businesses have always been built around people, and why that shaped how we see "the business" itself \[12:07] – Separating three things that used to travel together: the work, the responsibility, and the execution \[13:14] – Why the order matters: starting with the work, not the AI agent \[13:56] – Introducing work architecture as the first part of the AI Labor Architecture framework \[14:29] – Why capability comes later, after you understand your business and its outcomes \[15:04] – Where to sign up to be notified when the book, The AI Workforce, launches \[15:32] – The one exercise to try this week: pick one part of your business and start with the work 5 Key Takeaways 1. Start With the Work, Not the Capability — Asking "what can AI do" makes the tool the center of the conversation. Asking "what work needs to be done" keeps the business's actual outcomes at the center, which leads to much better decisions about who or what should execute that work. 2. Job Titles Hide Bundles of Very Different Work — A role like "onboarding specialist" looks like one job, but it's usually a bundle of distinct types of work, information gathering, interpretation, administration, education, follow-up, and exception handling, that only ended up together because one person had to do all of it. 3. AI Breaks Apart Three Things That Used to Travel Together — Historically, hiring someone into a role answered three questions at once: what work exists, who's responsible for it, and how it gets executed. AI labor lets you separate those three and assign execution more intelligently. 4. Capability Doesn't Tell You How to Design Your Business — Just because an AI agent can technically perform a task doesn't mean it should, and just because a person can perform something doesn't mean it's still the best use of their time. 5. Judgment and Exceptions Still Belong With People — When you break work down clearly, predictable and consistent pieces are well suited to AI, while genuine complexity, judgment calls, and relationship-dependent moments are where a person should still be brought in, ideally with full context already gathered for them. Links & Resources - smrtPhone: Listeners get 5,000 free calling minutes: https://www.smrtphone.io - The AI Workforce: https://thefutureworkforce.ai - That Real Estate Tech Guy: https://thatrealestatetechguy.com Thanks for tuning in to this one. If this got you rethinking how you look at a role in your own business, try the exercise: pick one part of it this week and ask what actually needs to happen, not what an AI agent could do. Head over to thatrealestatetechguy.com for all the episodes and some great discounts on the tech we talk about. More high-signal conversations coming next.