CGS Exec on Enterprise AI: Experiment With Humans in the Loop

As enterprise organizations continue to determine how to leverage AI, many are finding ROI a difficult milestone to achieve.

We spoke with CGS COO John Samuel about how enterprises can better prepare for AI to get the most out of their deployments. 

Why C-suite leaders need to commit to understanding AI before deploying it

To Samuel, the sheer pressure to adopt AI at any cost is overwhelming leaders who might get unclear directives from the top of their organizations, with little understanding of how to leverage AI efficiently and securely.

“It’s great everything is AI right now, but it’s also unfortunate sometimes,” Samuel said. “CEOs and other leaders are being bombarded with messaging, and it’s natural for leaders to want this. At the same time, it can be hard to tell what is marketing and what is reality.”

To Samuel, the way forward is for leaders to take the time to learn more about AI and the various AI-enabled technologies on the market before directing others to use the tools internally or adopting them in customer-facing capacities.

AI adoption, he says, can’t happen in a vacuum or without a clear, top-down understanding of how a company views AI — and when and where it should be leveraged. Without it, employees and customers alike are often left frustrated.

“When you take on these things in isolation, you have friction,” said Samuel.

‘Go slow to go fast’: keeping governance and security at the forefront of planning

Samuel also thinks some organizations are moving too fast to adopt AI, leaving behind basic best practices in data governance and security in favor of what they feel is a necessary pace of innovation.

“AI came to light and suddenly some people completely forgot their fundamentals,” said Samuel. “There’s something to be said for going slow at first to go fast later.”

Part of that, Samuel says, is bringing leaders from legal and security teams into conversations at the beginning to embed governance from the start of AI-related planning.

“People are afraid of regulation and compliance, and businesses can’t operate without guardrails, so put those teams and leaders at the table as you’re deciding on AI so they’re involved from the beginning,” said Samuel.

According to Samuel, CGS established working groups that brought together leaders across the organization to develop plans for its internal AI use. He says the company has also instituted policies that enable its employees to experiment with AI tools and use cases, and to embed technology into their own workflows, with clearly defined expectations around data security and best practices.

“AI isn’t just an IT project; it should be approached more like the way people-driven policies are handled,” Samuel said. “Continuous improvement, feedback, and performance maintenance are still important with AI, the way they have been in the past with other types of technology.”

How enterprises can safely experiment with AI to achieve ROI and long-term value

CGS has adopted AI internally over the past year, according to Samuel, in part because teams realized they wanted to leverage technology at work that they used in their personal lives.

“I do think leaders kind of need to just embrace it at this point– it’s happening anyway, and it’s much better for you to have guardrails around that use and have visibility into where it is,” Samuel said of employee AI tool use, sometimes called shadow AI.

Both because of that usage and the market-wide adoption rate, Samuel warns against sitting on the sidelines in 2026.

“If you’re in an industry that can afford to wait, good for you, but honestly, I don’t know what that industry would be,” Samuel said. “I think there’s a real danger in waiting another year to even start.”

Those experiments, though, don’t need to involve total transformations across operations. Samuel says CGS considers three horizons when determining investments:

• The low-lift “quick wins,” like enabling Microsoft Copilot if you’re already using Microsoft 365 solutions, that take little added resource but can offer teams a way to use new tools
• The year-long objectives that require more integration between systems, but provide tangible ROI within a set amount of time
• Long-term transformational projects that overhaul existing systems and require layered processes, but promise significant returns on investment once completed

Organizations of all sizes can implement low-lift tools, Samuel says, to get comfortable with AI as a step towards the transformational projects he thinks will deliver long-term value to businesses worldwide.