In certain instances, we may earn revenue when a reader makes a purchase or completes a form through our pages or on a partner’s website. We adhere to a strict 
Is artificial intelligence (AI) delivering value, or are organizations still chasing hype? That question is surfacing in boardrooms across every industry. While the headlines continue to celebrate breakthroughs and moonshots, business leaders are quietly asking where the return is on their AI and cloud investments.
After years of experiments, proofs of concept, and slide decks full of promises, the mood has shifted. Today, the pressure is to deliver measurable outcomes.
When Techopedia sat down with Leon Butler, CEO of IBM UK and Ireland, he didn’t hold back. “AI is not a bonus anymore. ROI is the benchmark,” he said. That single line captures how far the conversation has come. This isn’t about technology theatre anymore. It’s about whether AI is working.
Key Takeaways
- Only 25% of AI projects meet their ROI expectations, and persistent strategy gaps remain.
- AI scale requires trust, not black-box models or hype cycles.
- Start with outcomes, not algorithms, and work backward from the business result.
- Smaller, domain-specific models outperform giant, generic ones in enterprise use.
- AI ROI grows when systems are designed, not patched together reactively.
The Age of Experiments is Ending
According to IBM’s new CEO Study, two-thirds of UK and Ireland executives now believe their competitive edge hinges on a clearly defined generative AI strategy. 65% say their organizations are already adopting AI agents and are prepared to scale. The message is clear: we’ve moved beyond pilots. The mandate is execution.
However, that same study reveals a stubborn gap. Only 25% of AI projects have delivered their expected return on investment. And just 16% have scaled across the enterprise. The excitement around AI is real, but so is the frustration.
Leon Butler attributes much of this gap to a lack of foundational strategy. “Too many systems are hybrid by default. They’ve evolved reactively, not intentionally,” he said.
That misalignment creates fragmented environments where data is difficult to trust, workflows are disjointed, and AI struggles to perform effectively.
This is where IBM’s philosophy of being “hybrid by design” comes in. It’s not just branding. It’s a fundamental approach to building systems that scale on purpose. Think of it like engineering a bridge to carry high-speed trains. You wouldn’t just slap new parts onto an old footpath and expect it to hold.
What a Strategy-First AI Approach Looks Like
Why do some organisations thrive with #AI while others struggle to see returns? @IBM’s latest CEO study provides clarity: successful companies embrace five critical mindshifts. Which mindshift would create the biggest impact for your organisation?
🔗 https://t.co/J0CTgt0dGK pic.twitter.com/Pl1OkkHhmx
— Northdoor plc (@Northdoorplc) May 21, 2025
For IBM, strategy starts with clarity around the problem. Butler explained:
“You have to be outcome-focused from the beginning. Work backward from the result you want.”
That sounds simple, but it’s where many AI projects go wrong. The goal is often fuzzy. The data isn’t ready. The team is experimenting in isolation.
IBM has approached this differently. Internally, the company has realized $3.5 billion in productivity gains. $2 billion of that comes directly from AI and automation. “That’s not a projection. That’s real value,” Leon said.
At IBM, AskHR handles 94% of employee queries without human intervention. This isn’t a chatbot gimmick. It’s an orchestrated set of AI agents that pull information, trigger processes, and deliver answers that matter.
However, value doesn’t just come from building. It comes from governing. IBM has focused heavily on explainability, transparency, and control. “Trust is a prerequisite for scale. You cannot fake it,” Butler added. That means building AI models that can be audited, understood, and tailored.
The Case for Smaller, Smarter Models
One of the more distinctive decisions IBM has made is to resist the trend toward ever-larger general-purpose models. Instead, it has invested in what Leon described as “smaller, open-source, domain-specific models built for business.”
Butler told Techopedia:
“Enterprises don’t need a 750 billion parameter model trained on the entire internet. They need something transparent, efficient, and fine-tuned to their data.”
He used a visual analogy that stuck: “Imagine a test tube filled with liquid. You don’t know what’s in it. Now, you pour your data in. You shake it up. Would you drink it?”
That’s how many companies are treating black-box AI. They add their proprietary data to opaque models and hope for the best.
IBM’s alternative is the Granite family of models, which are smaller, explainable, and designed for high trust and high performance, built to be tuned, not mystified.
Productivity Gains That Add Up
It’s one thing to talk about strategy. It’s another to show it working at scale. At IBM’s recent Think conference, the company announced a new wave of AI agent capabilities. With WatsonX Orchestrate, teams can now deploy agents in under five minutes across 80 business apps.
These aren’t theoretical. These are live use cases in HR, procurement, finance, and customer support. Agents that not only complete tasks but also understand the workflow around them. One IBM client saw over 65% time savings on routine projects and reported a 170% return on investment.
Those numbers sound high until you see what’s being automated: scheduling, reimbursements, approvals, recommendations. It’s the kind of quiet work that makes everything else move faster. Or slower if it’s poorly done.
While this type of automation was once reserved for large enterprises, the tools are now becoming accessible to mid-sized companies as well.
The People Factor: Skills & Trust
None of this works without people. As Leon Butler pointed out, “60 percent of UK workers are not currently skilled for AI-driven roles.” That’s a reality check. As AI scales, so must talent.
IBM has committed to upskilling 30 million people globally by 2030, with a specific goal of upskilling two million in AI by 2026.
Through its IBM SkillsBuild program, the company offers free training across a wide range of disciplines. It’s open, online, and built with career progression in mind. But reskilling is only half the battle. Trust matters as much. From governance frameworks to explainable AI to ethical data use, IBM is building systems that not only perform but also perform responsibly.
This is becoming a competitive advantage in itself. Edgar Randall, UK&I Managing Director at Dun & Bradstreet, in IBM’s CEO study, said:
“Real transformation requires trust, not just speed. And that trust comes from strong governance, transparency, and the quality data that fuels AI.”
Quantum Is Coming, But AI Is the Now
Toward the end of our conversation, Butler briefly touched on IBM’s other major bet with quantum computing.
The company expects to reach systems more powerful than today’s supercomputers by 2033, with error-free quantum systems arriving as soon as 2029.
Already, IBM is working with companies like AstraZeneca and HSBC on use cases in drug discovery and financial modeling.
Quantum may be the next horizon, but AI is the challenge in front of us today. And for most businesses, that challenge is no longer about curiosity. It’s about delivery.
The Bottom Line
The biggest takeaway from this conversation is that simple strategies still matter. It matters more than ever. Because the companies that derive value from AI are not necessarily the ones with the flashiest models; they’re the ones with the clearest plans.
They’ve cleaned their data, connected their systems, trained their people, set tangible goals, and measured actual outcomes. This is not the work of a single pilot. It’s the work of intentional architecture.
So, the next time your team talks about launching another AI experiment, maybe pause and ask a harder question. Is the foundation ready? And most of all, is this system built for scale or built on sand?
