As organisations rush to adopt artificial intelligence, many still treat it as purely a technology challenge. However, the real barriers to long-term success lie elsewhere.
According to Mike Le Galloudec, Principal Engineer at Oakland, businesses risk focusing on AI tools themselves rather than the underlying processes and data that ultimately determine whether initiatives succeed or fail.
Measuring Outcomes Over Adoption
Over the past two years, AI adoption has become a boardroom priority. Organisations have invested heavily in copilots, assistants, and generative AI platforms to drive productivity, improve decision-making, and find new efficiencies.
But as investment continues to accelerate, many leaders are beginning to ask a fundamental question: where is the value?
“A lot of organisations are still measuring AI maturity through adoption rather than outcomes,” says Mike. “Deploying an AI tool across the business might improve usage figures, but it doesn’t necessarily mean you’re creating value. The important question is whether AI is helping solve meaningful business problems.”
This shift reflects a broader evolution in how enterprises view AI. While early conversations focused on what the technology could do, attention is increasingly turning to how it can be successfully implemented within the realities of large organisations.
The AI Success Triangle: Technology, Process, and Data
“Most organisations aren’t operating in a greenfield environment,” Mike notes. “They’re dealing with legacy technology estates, regulatory requirements, security concerns, governance frameworks, and fragmented data spread across multiple systems. Those realities have a huge impact on how quickly AI can deliver value.”
While AI capabilities evolve rapidly, organisations frequently underestimate the core foundations required to support them.
“AI is only one point of a triangle,” Mike explains. “The other two points are process and data. You can have access to the best AI models in the world, but if you don’t understand your business processes or your data isn’t reliable, AI won’t solve those problems for you.”
This is where many businesses encounter unexpected roadblocks. Eager to explore cutting-edge capabilities, teams often skip past long-standing issues around data quality, clear ownership, and operational complexity.
“Technology tends to amplify what’s already there,” says Mike. “If your processes are inefficient or your data is inconsistent, AI can accelerate those problems just as easily as it can improve outcomes.”
Silicon Valley Tools vs. Enterprise Realities
The scale of this challenge is reflected across the broader market. Research from Wakefield Research, conducted on behalf of Informatica, revealed that 97% of organisations using or planning to use AI struggle to demonstrate clear business value. Furthermore, two-thirds have been unable to move even half of their AI pilots into full production.
At the same time, major technology providers continue to launch consumer-focused AI innovations, such as Apple’s recent unveiling of Siri, a more capable, personalised assistant designed to manage everyday tasks across apps.
For Mike, launches like these highlight a fundamental divide between personal productivity tools and enterprise needs.
“A lot of the AI tools coming out of Silicon Valley are solving Silicon Valley problems,” says Mike. “They’re excellent at helping individuals automate routine tasks or improve personal productivity. But enterprise organisations are often trying to solve much harder challenges involving governance, compliance, security, disconnected systems, and complex decision-making processes.”
Consequently, enterprise leaders are discovering that AI success is less about deploying brand-new software and more about modernising existing operations.
“Enterprise organisations aren’t struggling because they can’t summarise an email,” Mike explains. “They’re struggling because they have hundreds of systems, unclear data ownership, regulatory obligations, and business processes that have evolved over decades. Those are significantly more difficult problems to solve.”
Building a Sustainable AI Strategy
The AI landscape is now entering a more mature phase as organisations move beyond initial hype and evaluate projects through measurable outcomes.
“We’re starting to see organisations move beyond the initial excitement and ask tougher questions about where AI is genuinely delivering value,” Mike says. “That’s a positive development. AI is an incredibly powerful capability, but it isn’t a silver bullet. Success depends on how effectively it’s integrated into the wider business.”
For organisations aiming to unlock sustainable value, the key is balancing investments across technology, data management, and process improvement.
“The organisations that get the most from AI won’t necessarily be the ones deploying the most tools,” Mike concludes. “They’ll be the ones that understand their processes, trust their data, and have a clear view of the outcomes they’re trying to achieve. AI has a major role to play, but it’s only one part of the equation.”
Expert Bio
Mike Le Galloudec
Mike LG is Principal Engineer Advocate at Oakland, a data engineering consultancy. Half his time goes on building agentic AI solutions; the other half on explaining them to people who’d rather not hear the phrase “agentic AI.”
He’s delivered production AI for enterprise clients, including Network Rail, where he built their first production-grade AI application, and has designed complex agent architectures for Financial Reporting Operations on Microsoft Azure. He advises organisations across rail, transport, utilities and financial services on the messy reality of AI adoption, including rescuing Copilot deployments for which nobody planned governance.
Before any of this, Mike spent eight years teaching secondary physics, which turns out to be decent preparation for explaining large language models to a boardroom. He has a Physics degree from Oxford, a career pivot through data science, and a neural network tattoo to keep him honest. He can also be found running in the mountains, which he insists is relevant preparation for enterprise AI delivery.
He speaks regularly at conferences alongside Microsoft, most recently at Data Decoded in Manchester, and runs Oakland’s YouTube channel and a tech TikTok with over 30,000 followers, built on cramming dense data topics into short-form video.