How CIOs can make AI work: layer in context (and a 3-step plan to do it)

Enterprise leaders all over the world can’t stop asking themselves: Why is AI taking so long to give us ROI? Will it ever? Their anxiety makes sense: more than 50% of executive leaders admitted it’s still "too early to tell" if AI has had any measurable financial impact (EBIT) on their business at all, according to McKinsey.

The problem isn’t AI itself per se. It’s the lack of operational context, argue two of our transformation experts at Celonis, Patrick Thompson (Global Senior VP of Customer Transformation) and Kerry Brown (Transformation Evangelist).

In a recent webinar hosted by CIO.com titled “The Winning Strategy for AI Success: Layer in Context,” Patrick and Kerry broke down exactly why AI is struggling to create meaningful RoAI.

• Watch The Winning Strategy for AI Success: Layer in Context webinar →

The real reason AI isn’t getting returns… yet

Tech stacks today are all over the place. Enterprises have CRM systems, ERPs, supply chain tools, human capital management software, and many others jostling for space.

As Patrick points out, “Those best-of-breed solutions just weren’t built to work exactly perfectly together. No one system or one platform can really run a business; it takes multiple systems, and in many cases, our customers have hundreds of systems.”

When companies channel data from these siloed systems into data lakes and try to run an AI agent on top of it, AI doesn’t fully understand what’s going on because it lacks the operational context of how the business actually runs.

“Context is the biggest challenge,” Patrick explains. “ The context layer is missing. The semantic layer where sources of data bring information into data storage platforms. What’s missing is the contextual model to make AI enabled.”

Without that missing context layer, AI might as well be taking business advice from a fortune cookie.

AI adoption is the problem, too

Kerry mentioned a statistic from an MIT study: 95% of AI use cases fail, and a massive “40% of those failures are purely due to adoption issues.”

“For years, I used to refer to people and processes as a relay race,” Kerry shares. “I grab the baton, I run with it for a few steps in the process, and I hand the baton off, and then when you look at all those systems of record and systems of engagement, where do those relay races intersect?”

By layering in operational context, you suddenly have a clear, role-based blueprint, says Kerry, “If you can be clearer around where the human makes a difference and where AI’s going to make a difference, both can be interrelated so that it’s not a mystery and a place of resistance and concern, but rather how to optimize resources, capability, and capacity planning across everybody.”

Beyond process mining: the digital twin

Advanced organizations over the past decade have used process mining to learn how their business ran. Then, to Process Intelligence, and now to a context layer that connects the data and consumption layers. Forrester predicts it will rescue 30% of failed AI projects this year.

Think of traditional process mining as diagnosing a single room in a house, advises Kerry, “Organizations would look at how to renovate a bedroom, kitchen, or bathroom – a particular area – and that use case would scale globally. But where AI is being considered, it’s really enterprise-wide. That enterprise-wide digital twin and our ability to provide a view of the entire home can give you a lot more choices and clarity on which choices to make…”

• Learn more about the Celonis Platform and the Context Model →

Patrick agrees that an enterprise-wide digital twin unlocks RoAI – Return on AI Investment – by providing agents with the process data and business knowledge they need to succeed.

“You can be successful with static AI where you just put in data as a repository and ask it a question," he says. “But when you’re talking about how businesses run in real life, in real time, all the way down to the user ID levels, the contextualization of those objects and the knowledge models and data models creates a digital twin of how your business is running. That’s what AI has needed, and that’s where the AI flip from failure to success is happening.”

CIO Hall of Famer Patrick Thompson’s 3-step “tech” plan for CIOs

1. Align with the rest of the C-suite

Don’t build a tech roadmap in isolation. Understand what your internal business stakeholders are trying to achieve.

2. Audit for context

Raw rows and columns in a data lake require immense manual engineering to map out. Prioritize tools that build a semantic layer on top of your data lake.

3. Build a composable stack

Structure your tech neatly. Systems of record at the base, data lakes next, a Context Model that creates a digital twin on top of that, topped off with an orchestration engine to coordinate multiple AI agents.

Forbes Technology Council member Kerry Brown’s 3-step “people & change management” plan for CIOs

1. Map out your demographics

Look at your business and IT roadmaps, and explicitly tie them to your people roadmap. Know where your hiring gaps and natural attrition points are so you can deploy AI strategically where it’s needed most.

2. Foster shared ownership

“People don’t hurt what they own,” shares Kerry. Involve employees early. Let them experiment, innovate, and contribute to how AI can augment their roles.

3. Create psychological safety

Shift the narrative from fear to opportunity. Use the data from your context layer to be highly surgical and intentional about how workflows will shift, giving HR and teams the runway to adjust gracefully.

Constant curiosity: the evolution of the CIO

CIOs have rapidly gone from tech owners to business strategists.

Kerry suggests that today’s successful CIO needs to adopt a mindset of “constant curiosity for constant change.” Instead of getting attached to an AI solution for three to five years, tech leaders are now “speed dating” – testing multiple solutions, seeing how they evolve, and adapting on a three-to-six-month horizon.

Patrick echoes that shift. Concluding that context bridges the gap between raw IT and high-level corporate growth.

“AI brings the CIO closer to the business more than ever,” Patrick says. “The CIO is becoming more of a ‘business CIO’ than ever before, enabling business strategy because of those insights… Making CIOs more strategic and relevant at the C-suite and board level.”

If you want your AI agents to cook, raw data is not enough. Feed AI operational context to eliminate operational blind spots, then watch your failure rates turn into RoAI.

• Watch The Winning Strategy for AI Success: Layer in Context webinar →

• Also, check out the other partnership articles, ebooks, videos, and more at celonis.CIO.com to learn how Celonis helps you secure meaningful RoAI →