For most of my career I’ve been the person in the room who decides how a system gets built. Salesforce programs for tens of thousands of users, integrations, data migrations, the architecture reviews where someone has to say “no, not like that.” I’m comfortable in those rooms. But over the last two years every client conversation started turning to AI, and I could feel the gap between talking fluently about something and understanding it.
So in December 2025, with a full project load and no obvious space in the week, I enrolled in the Post Graduate Program in Artificial Intelligence and Machine Learning: Business Applications at the University of Texas at Austin’s McCombs School of Business, delivered with Great Learning. Nine months, December to September. This week the certificate arrived, and I want to write down what the journey was actually like, because the LinkedIn version — a badge and a thank-you — leaves out the part that matters.
What the program is
Seven core modules with live faculty sessions and weekly mentored labs, five self-paced modules, and nine applied projects that each had to be submitted, graded, and defended. The arc runs from Python foundations through machine learning, advanced machine learning, neural networks, natural language processing with generative AI, AI agents for automation, and model deployment, with computer vision alongside. It is not a survey. Every module ends with a project on a real dataset, graded on whether the business recommendation holds up, not only on whether the model runs.
I finished with A+ in six of the seven core modules, 100% attendance, and a 4.28 GPA. I’m proud of that, and I’ll say plainly that the seventh module was the one that humbled me, and I learned more from it than from the six that went well.
What was hard
The time, first. There is no version of this that fits neatly around a consultant’s calendar. Saturdays belonged to class. Evenings belonged to projects. Deadlines didn’t move because a client escalation happened the same week. The only thing that worked was treating the program like a client engagement: blocked on the calendar, non-negotiable, with a weekly plan. My family absorbed the rest, and I don’t take that lightly.
Then the math. I hadn’t done serious linear algebra or calculus since my master’s degree. Gradient descent, regularization, the geometry of a support vector machine, why attention works — you can use a library without understanding these, but you can’t make architectural decisions without them. I relearned them at the kitchen table, usually late.
And the humility. In my day job I’m the person who decides. In this program I was the person whose confusion matrix was wrong at 11 pm and had to figure out why. That is a good thing to be reminded of at this stage of a career.
What changed
The most useful thing about doing this alongside real work is that nothing stayed theoretical. The ensemble-methods module became the way I think about evaluating LLM outputs. The NLP and transformers module became the retrieval pipeline behind RealApologist.com — millions of vectors on Cloudflare Vectorize, with Claude and Gemini orchestrated over it — a personal project I build in the margins. The AI agents module became the human-in-the-loop pattern I now push for whenever a client wants an agent to take an action on its own. The model deployment module became the discipline of treating a model like any other production service: versioned, monitored, rolled back when it misbehaves.
The deeper change is in judgment. I am now much more skeptical of “we’ll use AI for that.” I can tell the difference between a problem that wants a deterministic rule, one that wants a classical model, and one that actually needs a large language model — and I can explain the cost and the failure modes of each to a CFO. That, more than any single technique, is what I went back to school for.
What I’d tell my past self
Do it before you think you have time; you never will. Don’t audit — submit every project and let it be graded. Build something real alongside the coursework, because the coursework is where you learn the idea and the real thing is where you find out whether you understood it. And be honest about the module that didn’t go perfectly; it’s usually the one that teaches you the most.
Thank you
To the McCombs faculty and to the Great Learning program and mentor team, for a program that was demanding in the right ways. To my cohort, for the project threads at hours no one should be awake. To my family, for nine months of patience. And to the colleagues who asked, more than once, why I was doing this on top of everything else — this is why.