At a recent Head Resourcing IT Leadership Forum on Agentic AI, we welcomed Bill Wilson, VP and Head of Applied Artificial Intelligence at NTT DATA, to share his perspective;
Bill Wilson:
“I’ve been asked several times recently about what makes agentic AI successful. Rather than revisiting the well-known analyst forecasts on the impact of agentics, I want to focus on what I’m seeing in practice. These forecasts have certainly fuelled a surge of inquiries from CXOs — particularly now that NTT DATA is recognised as a Leader in CX consultancy services.
Many AI initiatives have failed to deliver meaningful ROI, so executives are understandably cautious. Ironically, that’s part of what makes agentics so compelling: it addresses many of the shortcomings that caused earlier AI efforts to falter.
One major challenge is business adoption. When you drop AI into an existing process, employees may or may not choose to use it. But when you design an agentic system where employees orchestrate or manage the process, adoption becomes a natural part of their work. That said, this isn’t a “flip the switch” moment — cultural change takes time. Employees need to see AI as a collaborator rather than a threat. Successful adoption also requires process transformation — otherwise, you’re just automating a broken process. At NTT DATA, we bring together process design, technology, and people expertise to ensure that transformation is holistic and sustainable.
A second challenge is the confusing technology landscape. We have an opportunity not to repeat past mistakes, but this time around, technology interoperability is key. To unlock the benefits of agentics, we need a foundation that allows agents to work together seamlessly. We call this the flywheel effect: as you create more agents, you should be able to reuse and combine them to deliver compounding value. The risk is agent sprawl — too many disconnected solutions, leaving humans to manually integrate them. Unfortunately, the conditions for agent sprawl are already here, driven by pressure from senior leaders and FOMO within development teams.
A third challenge is skills and mindsets. Low-code and drag-and-drop tools for agentic development are emerging rapidly — we’re partnering with Out Systems in this space, and all the hyperscalers are building similar capabilities. But this isn’t just about technical skills; it’s about cultivating the right mindset to get the most from these augmented engineering tools.
Despite these challenges, we’re already seeing significant benefits from agentics with our clients. One example is our work using agents for complex knowledge tasks — such as designing new software. Multiple agents collaborate, tackling different parts of the design process and iterating based on one another’s outputs. A human steps in at the end to finalise the design, but we build feedback loops so the agents learn from human input. This approach — now running in live environments — offers a glimpse into the future: enterprises where human and agent intelligence work together in true symbiosis.”
Thanks so much to Bill for sharing his perspective.
Curious about what agentic AI means for your talent, technology, or transformation plans? Let’s talk.