Many businesses are testing AI tools, but only 25% of AI initiatives are delivering expected ROI, according to IBM. Most AI transformations are failing during or right after experimentation. Long before adoption and long before outcomes.
Those initiatives don’t fail because of technology. They fail because:
- People can’t keep up. Businesses focus on tech, but the missing link is human capability. Remember the 70:20:10 model of learning? Forbes and Boston Consulting Group introduced an updated version of that framework as it relates to AI. They noted that successful AI embedding relies on 10% algorithms, 20% technology, and 70% people and processes. Tech is only one piece of the puzzle. The far greater share is in people and processes.
- The pilot couldn’t scale. Testing on a small cohort is critical, but success in a pilot does not automatically equate to success across a global enterprise. To move beyond isolated proofs of concept, you need to transform your AI investments into repeatable processes and systems. They need to fit into your workflows and your tech stack.
- The AI initiatives were never aligned with business and learner needs. Very few companies have made the leap to enterprise-level optimization where AI consistently delivers business outcomes. That’s because there’s missing buy-in and the initiatives weren’t aligned to measurable, organization-wide goals or challenges. Any outcome would be abstract or qualitative at best.
So, what works in the real world to address these challenges?
At LENS 2026, a panel of leaders from Yara, HubSpot, Pearson, and GSK who have successfully made the move from experimentation to practice took the stage to share their insights. Explore their top in-practice lessons and what efforts ultimately made the biggest impact:
Ground AI Initiatives in Business Friction, Not Idealism
It’s easy to get swept up in what AI can do, but successful implementation starts with what the business actually needs it to do. If you start with the technology features and hype, you’ll likely apply it to insignificant problems. You might generate excitement, but you’ll miss outcomes.
Leanne Jefferson, Global Head of Learning and Development at Yara, emphasized the importance of aligning AI initiatives with real business friction. For example, “In a global organization like Yara, scaling high quality leadership support like coaching is a challenge.”
That made coaching a great opportunity to test AI capabilities—in this case, through Degreed Maestro. Instead of just the “wow” factor, they secured approval for the AI tool based on use cases that resonated for the business, which in this case, was leadership development.
The choice to keep the focus on resolving a clear pain point is what ultimately secured needed stakeholder buy-in and enthusiasm, despite ongoing company changes and a proclivity toward being risk-averse.
Transform L&D Into a Cohesive AI System
As AI matures, the role of L&D must also evolve, because learning is, if anything, more critical than ever for business success. According to McKinsey, nearly 90% of leaders are seeking a significant change in how to develop employees.
“Although we don’t know what the future of an AI world might look like, there isn’t a positive outcome that doesn’t involve human development,” Zoe Botterill, Head of Learning and Development at Pearson, said.
Forward-thinking companies are leaving static virtual learning behind and embracing self-architected, responsive systems. At HubSpot, this shift was so profound that they haven’t made a traditional e-learning course since April 2024, and it’s even shifted traditional job roles.
“Many of us who were learning experience designers are now AI learning systems architects because they’re not creating e-learnings,” Jackson said. This new role focuses on connecting systems and amplifying the work of early adopters within the business rather than trying to do everything alone.
As part of building a scalable system, the production and distribution of learning content also has to adapt to keep pace with ever-evolving change. The time to prepare quality learning experiences is shrinking because the information they are based on evolves so fast, and business timelines are moving at light speed.
To address this, Botterill suggested moving into a more iterative workflow that allows professionals to adjust content in flight: “We’re going to launch, we’re going to adopt a product mindset, we’re going to iterate and learn as we go.”
Emerging AI capabilities are supporting this new way of working through responsive learning experiences and content libraries that update automatically.
Create a “Gym” for Practicing Skills
One under-estimated benefit of AI capabilities, especially for talent development, is the psychological safety it can provide for practice. When employees need to practice applying skills, but aren’t quite ready for the high stakes of a boots-on-the-ground scenario, tools like AI roleplays and one-to-one coaching can be a great option to bridge the gap.
“Nobody likes to practice roleplays in front of folks… This was a way that they could practice almost using it like a gym,” said Antonia Jackson, Senior Manager of Learning Innovation and Technology at HubSpot.
She noted that when employees used Degreed Maestro for difficult conversations, they felt more open to hearing feedback from AI than from a manager because it felt neutral and non-judgmental.
“You have to be able to fail because it’s in the failure that you find the goal, the treasure, and the lessons,” Jackson said.
This safe practice environment allows people to fail comfortably or “fail forward,” and become more capable before feeling comfortable with the vulnerability of trying out a skill in front of a colleague or manager. Not only is this iterative and safe practice more effective, but it also takes less high-value manager coaching time to develop capabilities.
Governance Is the Understated Secret to Scaling
When it comes to innovative AI pilots, many leaders view governance as a blocker. But robust data privacy and risk assessments actually build the trust required for wide-scale adoption. Any early misalignments in permissions or integrations can cause chaos when the project is scaled and released organization-wide.
Carlo Jose, Global Head of Learning and Talent Technology at GSK, would say that tech pilots are the perfect time to pressure-test governance and systems and make them more scalable: “I’m quite a big fan of breaking it now while we’re in testing stages so we can then figure out how we can evolve that.”
His better-to-break-it-now-than-later mindset involves engaging people in diverse functional roles to find the flaws in the tool before it is scaled across the business. Not only are flaws in governance and tech systems much easier to correct before a full launch, but launching with a fully fleshed out system and governance at scale also sets your workforce up for success.
Reimagine the Skills Needed for Modern Work
Skills are changing and so are the ways we develop and measure them.
Looking at skills through the lens of AI transformation fundamentally alters what skills businesses value and how they prioritize building them. For example, Botterill said Pearson focuses on three “power skills” for an AI world:
- Learning to learn: Understanding how to learn deeply and effectively.
- AI fluency: Building AI workflows into daily routines.
- Adaptability: Maintaining performance as technology pace increases.
At Yara, the learning team found that leadership development was not only a winning business use case and a priority for the “people and processes” investment, but it’s also the initiative that saw the most positive response. According to Jefferson, by the end of their pilot with Maestro, “The leadership coach was the highest rated feature.”
The Reality of Aligning AI Implementation to Business Outcomes
The companies seeing real impact aren’t experimenting more. They’re systematically redesigning how work, technology, and capability fit together.
AI transformation isn’t a “one-and-done” project. Whether you’re aiming to launch fast like Pearson or taking a deep-testing approach like GSK, the common thread across all these orgs is alignment. Align your strategy with desired outcomes, your technology with daily workflows, and your learning with real work.
Even when experimenting with AI, human development is the most critical part of the equation. Technology alone isn’t scalable or sustainable; unlocking AI value at scale demands a deliberate approach to transforming how your people learn and work.
Vamos manter contato.
Quero assinar a newsletter mensal que oferece insights exclusivos, anúncio de eventos e novidades sobre as soluções da Degreed.