Before you pay anyone to build you an AI agent, read this. A simple way to tell whether AI is the right tool, or whether you are about to over-engineer the wrong problem.
A former client texted me last week:
“I don’t have time to figure this AI stuff out. How much for you to just do it for me?”A former client, last week
I started putting together the proposal. Then I really thought about what he was asking me to build, and the whole thing fell apart.
He had been reading Dan Martell’s Buy Back Your Time, and the logic is clean: figure out what your hour is worth, then offload the cheap work first so your time goes to the expensive work. He runs a real estate private equity firm, owns a couple of other businesses, and wants to drop from 40 hours a week to 15. The cheapest, most maddening work on his plate is email. So the ask was simple: build me an AI agent that runs my inbox and gets me out of it.
Reasonable. Tidy. The more I sat with it, the less sure I got.
Here is the part nobody selling AI says out loud. I have worked in AI for three years, and I am a little embarrassed to admit I still cannot hand off my own inbox in a way I trust. I have one agent that flags my VIP clients. Beyond that, I have not cracked it, and honestly it has not been a priority.
To truly hand your email to an agent, you have to document every scenario it could face. Every sender, every exception, every “actually, in this case, do that.” It is tedious, and it is close to endless.
Halfway through the proposal, an uncomfortable thought showed up. It might cost him less to hire a sharp human to run his email than to pay me to build an agent that does it worse. Not a fun thought to have when someone just offered to pay you to build it. So I stopped and asked a better question: is something broken about what I am about to sell him?
That reminded me of a rule I picked up from Jake Van Cleif, a former Marine and software engineer. Not a guru. A well-paid expert doing the work inside enterprise-level organizations.
His rule is 60/30/10. When you put AI to work in a business, the AI itself is about 10 percent of the solution. The other 90 percent is structure and orchestration: the systems, the handoffs, the clear definitions, the process wrapped around the model. The shiny part is the smallest part.
Sit with that. If you are about to spend real money “adding AI,” you are spending it on the 10 percent and quietly hoping the 90 percent sorts itself out. It will not. That is how people torch a budget and end up with an expensive tool that still needs babysitting.
My client wanted to automate the lowest layer of his work because it annoys him the most. But the work in a business is not all one thing. At the bottom is execution, the checklist tasks. At the top is strategy, the thinking that decides where the whole business goes. AI fits those layers very differently.
|
High-Leverage Work
Strategy, the thinking that decides where the whole business goes. |
Looks like
|
|
Low-Leverage Work
Execution, the checklist tasks. |
Looks like
|
A mentor of mine, Geoff Woods, teaches leaders to use AI as a strategic thinking partner at the top of that stack. I have done it, and it is genuinely powerful. Automating email is the opposite end of the ladder.
So I started debating with myself. Was my client about to pour money into over-engineering the bottom of his work, when the thing that would actually give him his life back lives at the top? Something about “let’s just automate the stuff that sucks” does not sit right with me. What if the stuff that sucks does not have to exist in the first place?
He did not have an email problem. He had a “how is my time and my business actually built” problem. The inbox was just the symptom screaming the loudest.
So I am not sending him a proposal to build an email robot.
I am going to help him rethink the work. Maybe that is an AI-enabled assistant who handles the inbox and a dozen things a rigid agent never could. Maybe it is changing what he touches at all in a given week. The goal was never “automate email.” The goal was 15 hours instead of 40, with the time he buys back going into the investor calls that grow the firm. Automating his inbox barely moves that. Rethinking his work moves all of it.
That is worth more than the price of any tool. Before you pay to make AI do a task, stop and ask two things:
AI isn’t always the answer. Sometimes the answer is to step back and ask a sharper question. And the people who do that are about to quietly run circles around everyone still bolting robots onto broken work.
Natalee
You can, but it is rarely the best place to start. Email sits at the bottom of your work, the low-value layer. The bigger win is usually to rethink how your week is built, not to bolt a robot onto the part that annoys you most.
Ask two questions before you spend a dollar. First, is AI even the right tool here? Second, is this task the real problem, or just the symptom? If the task should not exist in the first place, automating it only makes a broken process faster.
When you put AI to work in a business, the model itself does about 10 percent of the job. The other 90 percent is structure and orchestration: the workflows, the handoffs, the clear rules, and the documentation around it. Spend on the whole system, not just the shiny part.
Yes, when it fits a real problem and you build the structure around it. The money gets wasted when someone buys the tool and hopes the process sorts itself out. Start with the problem, then decide what to build.
Training your team on AI, getting them up to speed, and putting AI to work inside your business is the work I do. If building agents like these, or installing a full AI operating system, feels like the next move, the next step is a conversation.
Natalee Champlin, The Affluent Affect®