Taking AI From Test to Real-World Impact: The TELUS Digital Guide

What does it really take to turn an exciting AI pilot project into something that works in the real world—without blowing your budget, getting lost in complexity, or losing sight of what’s right? That’s exactly what we explored in a recent chat with GT Tandon, Managing Director of AI & Data at TELUS Digital.

TELUS Digital is all about creating smart, seamless digital experiences for global brands. GT and his team aren’t just experimenting with AI—they’re using it to solve everyday challenges that real businesses face.

In our talk, we moved past the flashy headlines and buzzwords and got honest about what it’s like to build and launch powerful AI tools that actually work—especially in industries where the margin for error is tiny.

Taking AI From Test to Real-World Impact: The TELUS Digital Guide

Key Takeaways:

Modular AI enables small teams to build quickly using pre-built components and reusable pipelines.

Production AI needs governance controls because open source alone is not enterprise-ready.

Fix data quality issues first, or risk unreliable, biased, or unusable model outputs.

Utilize GenAI to automate compliance checks and identify regulatory gaps in real-time.

Establish governance groups to check for hallucinations, bias, and human review needs.

Why Just Five People Can Now Do What Used to Take Hundreds

Imagine this: if OpenAI were to roll out GPT-4 today, it might only need five people to manage it. Just five. That number tells a bigger story—something big has changed in how we build and scale AI.

“Back then, we were still deep in the research trenches... but over time, things have become way more modular,” said GT Tandon, reflecting on how far things have come.

Not so long ago, teams had to start from zero—researching, designing, testing—every single piece of the system by hand. It was slow, exhausting, and required a small army.

But today? Things are different. Thanks to reusable components, pre-built models, and plug-and-play data pipelines, teams can focus more on building cool things rather than reinventing the wheel.

Tandon put it simply:
“The parts are already out there—you just pick them up and start flying.”

That shift—from building every tiny piece from scratch to assembling pre-made components—is why creating powerful AI now takes a fraction of the time and people it used to.

The Hardware Behind the Magic: Chips, Memory, and Speed

But it’s not just about software. Hardware has evolved, too—and that’s a big deal.

“Earlier, the chips weren’t really built for the kind of heavy math AI needs,” GT said.

Back then, processors wasted power and time just trying to do the basic math behind AI. But now? Companies like NVIDIA design GPUs that are made specifically for the crazy calculations AI depends on.

📍 “AI is really just a lot of math. Old chips struggled with it. NVIDIA changed the game.”

Purpose-built hardware—paired with smarter algorithms—is making AI faster, cheaper, and easier to scale than ever before.

And it’s not just about crunching numbers faster. Improvements in memory access and how chips talk to each other are smoothing out the whole process. It’s like the difference between riding a bicycle and driving a sports car.

“Every single day, there’s a new piece of research pushing things forward,” GT added.

The pace of progress is wild—and it’s only getting faster. As AI infrastructure continues to improve, the slow, clunky parts of development are vanishing. In their place? More room to imagine, create, and innovate.

Why Open Source Sparks Curiosity and Creativity

Tandon spoke about how open source has become a powerful way to explore new ideas—but he was also clear about the limits.

He shared with Techopedia:

“Open source is amazing for experimenting and doing research… but when it’s time to actually build something for the real world, that’s when you need solid, reliable code.”

Open source tools give teams the freedom to try new things, play with fresh ideas, and test out early concepts without too much risk. It’s like a sandbox where creativity can really shine. But once those ideas move closer to becoming real-world applications—especially in high-stakes industries like healthcare or finance—things like security, stability, and trust can’t be ignored.

Tandon put it simply:

“Even if you’re using open source, you still need to ask—how is it being managed? Where is the data stored? Who’s responsible if something breaks?”

In other words, it’s all about knowing when to shift gears—from experimenting fast to building something safe and dependable.

Real-Life AI: Helping Doctors and Banks Do Better Work

One of the most exciting parts of the conversation was hearing how generative AI is already making a real difference—especially in industries where small changes can mean big impacts.

Let’s start with healthcare.

“There’s a huge need to help doctors and nurses make sense of all the medical documents they see every day,” Tandon explained. “AI can pull the most important details together into summaries that are actually useful.”

Hospitals deal with mountains of messy, unstructured information. GenAI is helping clean it up, spot patterns, and give healthcare workers the info they need—fast. That can directly improve how patients are treated and how smoothly things run behind the scenes.

In finance, things are just as intense, but in a different way.

Tandon talked about how much pressure companies are under to stay on top of complicated regulations. That’s where GenAI steps in—reading through the rules, flagging possible issues, and helping teams stay compliant.

“GenAI can understand the legal language, see where something might be out of line, and let the right people know before it becomes a problem,” he said.

He also mentioned other powerful examples—in cancer research, in insurance claims, even in everyday business operations.

“It’s everywhere now,” he said. “And honestly, it’s getting harder to find places where GenAI isn’t making a difference.”

Breaking Free from Pilot Mode: Why So Many AI Projects Stall

A lot of companies are still stuck in the “AI testing phase.” They’ve done the pilots, seen the demos, and they know AI is the future. But when it’s time to move from theory to practice? That’s where things start to fall apart.

GT laid it out clearly: there are three big reasons why businesses aren’t getting real results—or return on investment—from their often costly AI efforts. The culprits? Messy data, lack of skilled talent, and unclear leadership direction.

But here’s something GT emphasized with urgency: AI should never be built without ethics at the core.

“We’ve set up two dedicated teams—one focused on AI governance and the other on trust and safety,” he said.

These aren’t just for show. Every single AI use case they work on goes through checks—for things like hallucination risk, data accuracy, and whether the output is actually appropriate.

“Humans need to review the results. We have to make sure what’s coming out of these tools respects the people it’s meant for,” GT added.

Because at the end of the day, AI should be a tool that helps people make smarter choices—not something that makes decisions for them.

The Big Fear: Is AI Coming for Our Jobs?

It’s a question people ask all the time—and not without anxiety: Is AI going to take my job?

GT’s response? Honest and calming.

“When the locomotive came along, the horse-and-carriage industry shrank, sure. But then, all that energy shifted to building the railroads and running trains.”

Yes, things will change. But they won’t disappear.

AI isn’t here to wipe us out. It’s here to reshape how we work. GT added:

“There’s still so much work to be done. New roles will emerge. Entirely new job descriptions will be written. But no, AI isn’t going to just send humanity into early retirement.”

What Leaders Need to Focus on Now Through 2026

So what should today’s business leaders be focusing on as we head toward 2026?

GT didn’t mince words:

“Double down on your data and AI strategy. Don’t just chase every shiny new thing—ask yourself which initiatives are actually going to deliver results.”

And be smart about where you invest.

“Not every ‘build’ decision is a smart one. Same goes for what you buy. Be intentional.”

Yes, experimentation is still important—it’s how we learn. But doing it without a game plan? That’s just burning time and money. GT put it plainly:

“It’s still better to experiment than to do nothing. But you need a system to prioritize. Otherwise, it’s just noise.”

The Takeaway

AI isn’t some science experiment anymore. It’s out in the real world, and it’s growing fast. The tech is improving. We know what the common roadblocks are. Now, the hard part is execution.

GT Tandon’s advice is a wake-up call: successful AI doesn’t depend on flashy tools or futuristic hype. It depends on having clean data, ethical governance, skilled people, and leaders who know what they’re aiming for.

It’s time to stop circling the runway. No more endless pilots. The goal isn’t just to use AI—it’s to make it meaningful and valuable.

Because only then can we stop worrying about whether AI will replace us, and start building systems that work with us—with care, curiosity, and a sense of purpose.