AI Didn’t Make Me Dumber. It Made Me the Strategist.

A colleague told me last week that AI is making us all dumber. I disagree, and instead of arguing I want to show you a workflow. It took me a few hours to build and it took real strategy to design. The demo is a click away. The decisions are the point.

“It’s making us all dumber.”

That was a colleague, last week, about AI. She wasn’t being lazy about it. She has watched people paste a question into a chat box, paste the answer into an email, and call it thinking. I’ve watched that too. She isn’t wrong that it happens.

But I don’t think the tool decides that. I think you do. And the best way I know to make that case is to walk you through something I built last month, because it is the clearest example I have of what AI actually changed in my work. It didn’t replace the thinking. It moved the thinking to where it belongs.

A process that was failing quietly

I run operations for a company with a handful of contractors. Every month they submit invoices. Every month, some of those invoices didn’t match the hours logged in our time keeping system. Not because anyone was careless. Because life gets busy, and double-checking your own timesheet against your own invoice is the kind of task that slides when it’s Friday and the invoice is due.

So the numbers came to me. I checked them against the timekeeping system myself. I chased the ones that were off. I searched email and Slack for the ones that had been sent to the wrong place. Then I approved them and forwarded them to accounting.

Nothing about this was a crisis. That’s what made it dangerous. It was failing in the way that creates more work rather than the way that gets noticed. Every month, a few hours of a COO’s time went into reconciling numbers that a system could have reconciled in seconds.

What I built

The fix is a Slack bot. A contractor types one command. The bot pulls their hours directly from the timekeeping system, so the number is the number. One click turns those hours into a formatted, correctly numbered PDF invoice. The contractor reviews it and submits it. It lands in a channel where I see it, open it, give it a second set of eyes for compliance, and click once more to send it to accounting. Accounting pays on the usual cadence.

No spreadsheets. No exports. Nobody logs in to Harvest. Nobody searches email for an invoice. Fifteen people useing it equals two days of my time and no errors.

You can click through the whole flow on my Case Studies page: https://www.valorierobles.com/case-studies

Claude wrote most of the code. I want to be clear about that, because it matters for the argument. I am not a developer. I described what I needed, it built, I tested, it fixed. That part of the work went from impossible for me to a few hours. That is the “it does the research and puts it in a nice packet” part, and it is real.

But look at what it could not do.

The six decisions AI could not make

Before a single line of code existed, I had to sit with the workflow and answer questions that have nothing to do with software.

1. Who needs to approve what?

The contractor approves their own invoice before it goes anywhere. I approve it before accounting sees it. Accounting doesn’t approve anything; they pay. That is three different roles with three different kinds of authority, and the bot had to be built around them, not the other way round.

2. Who keeps control at each step?

The contractor can’t send an invoice to accounting directly. I can’t edit their hours. Accounting can’t send anything back up the chain. Each party owns exactly one part and nothing else. That is a design decision, and it is the same decision I would make with paper.

3. Who owns the bot?

Somebody has to. If it breaks on the first of the month, whose phone rings? Mine. I built it, I own it, and I am the one accountable when it is wrong. A bot without an owner is a liability waiting for a quiet Tuesday.

4. Where does governance live?

Every invoice passes through one channel where I see it first. Not because I don’t trust the bot. Because a human check before money moves is a control, and controls don’t get removed because the process got faster. If anything, faster processes need better controls.

5. Who is accountable when it is wrong?

The bot pulls the number from Harvest. If Harvest is wrong, the contractor is accountable for their timesheet. If the invoice is wrong in a way Harvest wouldn’t catch, I am accountable, because I approved it. The tool doesn’t hold accountability. People do, and you have to decide which people before you automate anything.

6. Does everyone keep their ability to approve as it moves along?

Yes, and this was non-negotiable. Automation that removes a person’s approval right in the name of speed is not efficiency. It is a governance gap with a nice interface. Every party in this workflow can still stop it at their step.

Those six answers are the workflow. The code is just what carries them.

So did it make me dumber?

No. It made me think more strategically in a part of my job where I hadn’t needed to.

Here is what I mean. I have always been able to look at a process and see the paths to efficiency. That is the job. What I couldn’t always do was build the path, because building it manually meant wrangling the paper, the people, the systems and the exceptions until the effort outweighed the gain. So a lot of good paths stayed on the whiteboard.

Now the building is cheap. Which means the only thing that matters is whether the design is right. Whether the roles are right. Whether the controls are right. Whether the accountability is right. The strategy moved from a nice-to-have on top of the work to the entire work.

Same techniques. Same instincts. Same effort I have always put into making an operation run cleaner. One more tool in the arsenal. The tool could not have designed this workflow without me, because the workflow is a set of decisions about people, and it doesn’t know my people.

What this means if you run a company

If you are worried AI is making your team dumber, I would look at whether anyone in your company is making the six decisions above before they automate something. If nobody is, the tool isn’t the problem. The gap is the problem, and it existed before the tool arrived.

If you have a process that is failing quietly, the kind that costs hours instead of headlines, that is usually where I start.