Why Enterprises Need A Specialized Role Sitting Between AI And Operations

Customer-facing AI tends to expose the data architecture underneath it. A generative model can retrieve answers, summarize interactions, or guide agents through complex workflows, but it can’t compensate for stale knowledge bases, disconnected systems, or unclear ownership of the information it depends on. As enterprises move from AI pilots to production deployments, professionals are running up against a hard truth: model performance matters, but the data foundation and operating model determine whether AI creates leverage or simply accelerates existing problems.

We spoke with Andrew Hill, Chief Operating Officer of the coaching platform FiveLumens and Founding Principal of Huntington Group LLC, about where these deployments break down. Hill has run enterprise operations for more than 20 years, scaling global teams to over 1,500 agents across seven countries at TravelPASS Group and co-inventing patented cross-channel customer journey tracking technology at Lucency. That mix of operational and technical background gives him a hands-on view of what it takes to build, train, and maintain intelligent systems that actually work in production, and it's that type of position that he believes is crucial for AI adoption. "What most organizations are missing right now is that person sitting on the fence who can understand both the operations and technology sides and bridge that gap," Hill says.

When the tech works, but the process doesn't

AI isn't a panacea for enterprises, and throwing it on top of broken processes just scales dysfunction rather than curing it.

Hill points to a recent healthcare client that wanted to fix its knowledge management system with generative AI. His team cleaned the client's articles, removed duplicates, and wrapped the database with a natural-language layer that could surface answers and show supporting references to end users. As expected, the tech worked perfectly. It was the surrounding human employees who lagged.

Because the client took months to update their articles, the real-time search became useless. The deployment stalled, illustrating the friction of layering automation over contradictory or untraceable knowledge bases. Avoiding that outcome requires clear ownership of content updates and feedback loops, regardless of the use case the AI is meant to support.

He argues that AI rollouts deserve the same rigor as any other operational investment, which means documentation, planning, training, and testing. "You have to take the time to train and test and do everything like you would with marketing," he says. "Marketing is not throwing out a new campaign and saying, 'It didn't work.' They're going to test it out, they're going to A/B test like crazy to make sure it all works. We need to apply that same thing to technology."

Building toward a single source of truth

But even a perfectly trained team cannot save a fractured data pipeline. In Hill's experience, when a unified data foundation is lacking, AI systems often amplify existing gaps rather than resolve them. He highlights the architecture from his time at an online travel agency, where his team created a single operational layer to route and prioritize work across the business. That routing logic rested on a single, unified view of customers, transactions, and cohorts. Achieving that type of personalization often pushes companies to simplify their tech stacks and build toward a modern data environment where key metrics live in one place.

"Whether it's AI or even simpler technology, it's all the same. You have to have that single source of truth," Hill says. "Being able to treat the customer experience in a way where someone can interact with your business and immediately see, I know you're looking at this product in this market on these dates. That's creating a real experience."

The role most organizations are missing

With those organizational necessities in place, an organization can begin making serious inroads toward activating AI across departments. At the same time, a new challenge rears its head: with hybrid, agentic operating models, the line between human and machine work requires a practical organizational adjustment.

Hill suggests that many companies will benefit from creating a specific role that bridges operations and technology, rather than leaving that translation to chance. Someone in that position, as he describes it, lives on the boundary between teams: they understand the workflows and the customer and employee experience, and also know enough about the tech stack to shape how AI is applied. Their job is not just to suggest ideas but to work with IT, operations, sales, and marketing to implement changes while governing AI outputs and treating documentation as serious infrastructure, on par in rigor with serious data governance frameworks.

That same emphasis on documentation runs through Hill's view of operational readiness. "For me, the advice is that you have to start getting clarity with your operations, get clarity with what employees experience, get clarity with what the customer experiences within processes and workflows."

Treating AI like a new hire

Building a functional AI data foundation takes time, and Hill notes that progress usually comes from narrowing the scope. The role he describes (and inhabits in his own work) can help massage AI tech into workflows by understanding the training it needs and the molding it requires to be successful.

He likens this work to the hiring of employees. Organizations routinely give new hires time, coaching, and room to make mistakes. Yet when an AI system produces a bad outcome, tech buyers are quick to fire the bot for a mistake they would easily forgive in a human, rather than refining the prompts, data, or workflows feeding it.

"I always find it weird that we have humans that are very flawed and we allow them to make huge mistakes and we keep them around and we develop them," Hill says. "But when it comes to a technology, if it has one little flaw in something because we didn't train it right or didn't set it up correctly, it's the technology's fault. We need to apply the same methodology we do to hiring people that we do to technology."

A foundation worth building

Ultimately, Hill's approach to enterprise AI relies on focus. By choosing one or two key use cases, clearly documenting workflows, building feedback loops, and testing with real data before scaling, teams can bypass vendor hype. In his experience, those small, well-designed changes accumulate far more value than chasing dozens of features at once, and they give the technology something solid to stand on. "If you don't have that foundation," Hill says, "AI is going to do nothing more than make your problems bigger, faster."