AI Was Coming for My Job. So I Slayed the Dragon.

For most of the last year, I read the same headline in a hundred different forms. AI is coming for operations jobs. AI will replace the coordinators, the program managers, the Chiefs of Staff, the people who keep the trains running.

I have spent more than twelve years keeping the trains running. So I took it personally.

I had two choices. I could wait and see how the story ended, or I could walk straight at the thing I was afraid of and learn what it actually was. I picked the second one.

Step one: learn the dragon

In June I finished Claude 101 from Anthropic. By August I had completed the Google AI Professional Certificate and its full set of courses, from AI for app building to AI for research.

The certificates were never the point. What they gave me was vocabulary and a little confidence. I could finally tell the difference between what AI is good at and what people only say it is good at. And I noticed something I did not expect: almost every hard problem in those courses was an operations problem wearing a tech costume. Scope. Sequencing. Who decides. What happens when it goes wrong.

But knowing about something is not the same as doing it. I needed to build.

Step two: prove it

In July I entered the Slack Agent Builder Challenge. I had 12 days, no developer background, and an idea I could not stop thinking about.

At after-school tutoring nonprofits, one volunteer can be responsible for 20 to 30 kids. The hardest moment for a student is not during a session. It is 7pm on a school night, stuck on homework, with nobody to ask. Kids give up quietly, and the tutor never finds out.

So I built Lumi. Lumi coaches students with hints instead of answers. If a kid is still confused, it slows down and gets more concrete. And when a student is genuinely stuck, it quietly alerts a human tutor with what they need to know, so the tutor can step in. No child gets called out in front of anyone. A person always makes the final call.

I built it with Claude as my pair partner, one step at a time. There were nights it fought me. There were bugs I could not name, let alone fix, until I learned how to describe them. But I shipped a working agent before the deadline.

Here is what surprised me. The technical parts were hard, but they were learnable. The parts that decided whether Lumi would work were the parts I already knew how to do. Writing the plan. Cutting scope when the clock got loud. Deciding what the guardrails had to be before writing a single line, because the users were kids. Designing the handoff between the AI and the human. That is operations. That is what I have been doing for twelve years.

Step three: make it real

A hackathon proves you can build something. A product proves someone needs it.

Lumi started inside Slack, and that was the right place to prove the idea. But most tutoring programs do not run on Slack, and a tool kids need should not depend on which chat app the grown-ups picked. So I am now building Lumi into its own standalone app that any program can use.

I am doing it the way I would run any launch for a client: phased, with a clear critical path, and with the trust work treated as part of the product, not an afterthought. When your users are children, privacy and safety are the product.

The skill I had to dust off: strategy

Somewhere in the middle of all this, I noticed I was thinking differently.

For a long stretch of my career, the job was execution. Make the plan work. Keep everything moving. I was good at it, and honestly, I had not had to be truly strategic in a long time.

AI changed that. When a new tool can take on a whole category of work, you cannot just ask how to do the same thing faster. You have to step back and look at the problem from a different angle entirely. What would we do differently if this part of the work were nearly effortless? Which problems were we never able to touch before? Where does a person matter most, and where are we spending people on work that does not need them?

Those questions pulled me back into strategy. Lumi only exists because I asked one of them: what if tutors did not have to guess which kid needed them?

The harder part: bringing people with you

Learning the tool was the easy part. Introducing it to a team was not.

I have rolled out plenty of new tools over the years, and that has always been an interesting piece of the work. Every new system comes with some grumbling and a learning curve. AI is different, because a lot of people do not see it as a tool. They see it as a threat.

On the teams I work with, I had people on both sides of the fence. Some loved it and wanted to use it for everything. Some hated it and wanted nothing to do with it. Both reactions made sense to me, because I had felt both of them myself.

That forced me to lead at a different level. Rolling out a tool is a project. Helping people move from fear to curiosity is change management, and it is cultural work. It meant listening to the skeptics instead of steamrolling them, because they often see the real risks first. It meant being honest about what would change and what would not. And it meant showing, with real examples, that the goal was to take the tedious work off people's plates, not to take their jobs.

I am still learning here. But I am convinced this is where most AI efforts succeed or fail. Not in the technology, in the culture around it.

Walking the walk: the website you are reading

There is one more thing that came out of this journey, and you are looking at it.

I wanted a place to show what I had learned, not just talk about it. So I built this website with the help of AI. The case studies, the writing, and the agent demos you can click through on the Case Studies page are all part of the proof. I am not just talking the talk.

But I want to be honest about how it got built, because it matters. I do not believe AI replaces people, and this site is a good example of why.

AI did the build. People made it good. I am lucky to have friends who do marketing, business development, websites, and SEO for a living, and they showed up. They looked at drafts, told me what was not working, and gave me the guidance I could not have gotten anywhere else. I had a human team behind me the whole way.

AI kept helping well past launch, too. It helps me track my SEO, and it helped me find the right people to connect with and the right groups to join on LinkedIn.

Here is the one line I hold to. My posts are me. My writing is me. The build is where AI helps.

What I actually slayed

I thought the dragon was AI. It was not. The dragon was the story that people like me get replaced.

What I found instead is that AI rewards exactly the skills operations people have spent careers building. Knowing what problem you are actually solving. Breaking big work into steps a team can execute. Building in checks. Keeping a human in the loop where it matters. Bringing a team along through change. Getting it out the door on time.

AI did not make my experience obsolete. It made it more useful, because now I can build the systems I used to only design. And it did not replace the people around me either. Every step of this went better because humans were in it.

If you are an operator watching the same headlines I was, my advice is simple. Do not wait to find out how the story ends. Pick something small and real, and build it. You will be surprised how much of what you already know carries over.

If this sounds like your business

If you are a founder trying to figure out where AI actually fits in your operations, that is the work I do every day. And if you run a tutoring program and want to be one of Lumi's first pilot partners, I would love to talk. Reach out at valorierobles.com/contact.