This week I had the opportunity to attend the INFORMS Analytics+ Conference in Indianapolis—a gathering of industry professionals, academics, and thought leaders exploring the future of analytics, optimization, and AI. It was a mix of sharp insights, witty commentary, and honest reflection on where our field is headed. Below are some quotes and thoughts I’ve been mulling over since.

Asking Better Questions in the Age of LLMs

What’s valuable now is the question. Before it was the answer.

In the era of large language models like ChatGPT, we’re no longer limited by lack of access to information or even decent analysis. The challenge has shifted to framing the right question.

One practical tip I heard was to push AI during brainstorming. Don’t just ask for a single idea—ask for 10, or 50. Cast a wide net, then filter. The magic isn’t just in what the model gives you, but in how you guide it. Quantity can unlock quality.

Optimization Beyond the Dollar Sign

“We often forget about the things our company values that do not have a dollar sign directly in front of them. Such as time, quality, quantity, explainability, and complexity.”

As someone who builds optimization models, this was a timely reminder. Too often, we focus purely on financial objectives. But decision-makers also care deeply about things like interpretability, speed, and robustness—especially when models are deployed in the real world. Let’s make space for those considerations in our objective functions too.

When One Model Isn’t Enough

Lufthansa shared their approach to large-scale optimization, and it was eye-opening. Despite advances in solvers and compute, there are still problems too large to tackle all at once. Their solution? Break it down.

Planes, crew, and passengers are modeled separately. Each piece flows into the next, and while there’s communication between them, the process is inherently sequential. Sometimes decomposition isn’t just a technique—it’s a necessity.

Models Are Not Perfect Mirrors

“We don’t always know all the data that goes into the model, but we treat the model as if we did.”

A sobering reminder. We often assume completeness in our data and models that just isn’t there. Acknowledging uncertainty—and building systems that are resilient to it—might be more important than obsessing over precision.

From Punch Cards to Prompts

“Programming languages are becoming more like English over time.”

This historical arc—from punch cards to Python, and now to natural language prompting—shows how accessibility and abstraction continue to rise. Today, thanks to LLMs, we’re practically coding in English. This has implications not just for who can code, but for how we think about automation and creativity in analytics work.

Don’t Forget the Bill

“AI and the cloud cost money.”

It’s easy to get caught up in the allure of bleeding-edge tools, but we can’t lose sight of ROI. Cloud compute and AI services can quickly rack up costs. Just because a tool is cool doesn’t mean it’s cost-effective—or even necessary. Grounding tech adoption in value creation remains a core discipline.

Communication Is More Than an Accessory

“Students only get technical training, not presentation training.”
“Results are necessary, but not significant for success.”

These observations hit home. Technical skills will get you in the door—but communication is what gets your ideas across, your projects approved, and your impact felt. We need to train analysts not just to find the answer, but to sell it—clearly, convincingly, and with context.

AI vs. ML (and the PowerPoint Test)

“What is the difference between ML and AI? If it’s in Python it’s probably machine learning, if it’s in PowerPoint it’s probably AI.”

This tongue-in-cheek comment got a laugh, but it also hinted at a deeper truth: AI has become a catch-all term, often divorced from its technical meaning. As practitioners, we need to be precise about what we’re building—and honest about its capabilities.

Final Thoughts

The conference was a valuable pulse-check on the state of analytics. It reminded me that while tools and techniques evolve, some things remain constant: the need for clarity, the importance of framing, and the value of humility in the face of complex systems.

We’re in a fascinating moment where optimization, AI, and human judgment intersect. Let’s keep asking good questions.