Stop Hiring Roles. Start Mapping Workflows.
The shift that changed everything for me: I stopped thinking “I need to hire an editor” and started thinking “what does an editor actually do?” The answer was 16 discrete activities across 4 workflows. Once I could see those activities clearly, I could ask a better question: which of these can an agent handle?
This isn’t about replacing people. It’s about seeing your business clearly for the first time. Most of us have never actually mapped what happens in our organizations at the action level. We describe work in role-shaped chunks because that’s how we’ve always organized it. “Marketing handles that.” “Sales does this.” But marketing doesn’t do one thing. Marketing does 47 things, and some of them have nothing to do with each other.
When you think in roles, you hire a person and hand them a vague mandate. When you think in workflows, you can be precise about what actually needs to happen, in what order, with what inputs and outputs. That precision is the difference between AI that works and AI that generates expensive noise.
Here’s the exercise I recommend to everyone I work with: draw your entire business as one linear workflow, from first customer touchpoint to final delivery. Every single action. Not departments. Not roles. Actions. If you can’t draw it, you don’t fully understand your business. And if you don’t understand it, neither will any AI you deploy.
The decomposition process looks like this:
- Start with the 5 to 7 major stages of your business (lead generation, qualification, onboarding, delivery, follow-up, whatever yours are).
- Under each stage, list the 6 to 7 actual actions that happen.
- Under each action, list the sub-actions.
- Keep going until you can’t reduce further.
- Those irreducible actions are what you evaluate for automation.
What you’ll discover is that most roles are actually bundles of wildly different cognitive tasks. Some of those tasks require deep human judgment (negotiating a contract, making a creative call, navigating a sensitive conversation). Others are pure pattern execution (reformatting a document, pulling data into a template, sending a follow-up email three days after a meeting). The first category stays human. The second category is where agents thrive.
This is what meta-work-is-now-work looks like in practice. You’re not doing the editing. You’re designing the system that does the editing. That’s the player-to-coach transition applied to every function of your business. And once you’ve mapped everything at the action level, you can start asking the really interesting questions: Which actions are bottlenecks? Which ones are error-prone? Which ones does everyone hate doing? Those are your highest-leverage automation targets.
There’s something clarifying about this discipline. When you force yourself to specify exactly what “good” looks like in observable terms, you discover how much you’ve been operating on vibes and assumptions. It’s humbling. And it’s the same discipline that prayer requires: being specific about what you’re actually asking for, not just vaguely wishing things were better. The people I know who pray with real precision tend to be the same people who can write clear workflows. Both practices demand that you stop hiding behind abstraction and say what you actually mean.
Key Takeaway
Roles are how we’ve always described work, but workflows are how work actually happens, and seeing the difference is the prerequisite for any meaningful AI deployment.
References
- Gerber, Michael E. The E-Myth Revisited. The original argument for systematizing business operations at the process level rather than the role level.