I have a confession to make. I came to CCW Las Vegas expecting the usual mix of keynote optimism and carefully managed case studies. What I did not expect was to walk out of a Thursday morning session at Caesars Forum with three ideas I genuinely could not stop thinking about.
Be the 5% was the name of the Amazon Connect Customer session. It borrows its framing from an MIT article that made the rounds in enterprise technology circles, and the central argument is straightforward and rather uncomfortable: 95% of enterprise AI pilots never reach production. Not because the technology fails. Because the approach does.
Hannah Bloking, Head of Applied AI Acceleration Solutions Architecture at Amazon Connect Customer, and Andrei Papancea, Head of Engineering at AWS, presented the session to a room of roughly sixty people. I filmed them immediately afterwards, alongside Nate Brown, Co-Founder of CX Accelerator, who I brought in to represent the industry perspective and keep the conversation grounded in what CX leaders are actually experiencing day to day.
Three questions. Three perspectives on each. It is the most practically useful fifteen minutes I have filmed at a conference in a long time. And I say that as someone who has filmed rather a lot of them.
The Fear Nobody Names Out Loud
The first question I asked was about fear. Specifically, what CX leaders are most afraid of when it comes to AI, and whether that fear is justified.
Hannah’s answer comes from a vantage point that very few vendors can claim. Amazon runs AI across more than 150 internal service organisations. The lessons that scale teaches, about what production AI actually costs in time, resource, and organisational readiness, are not the lessons you read in a vendor whitepaper. They are the lessons you learn when your own customers are on the other end of the system.
Andrei’s answer goes straight to the leadership mandate problem. AI is ordered from the top. Nobody defines what good looks like. The project starts with energy and ends with a steering committee meeting nobody wants to be in. You will recognise this pattern if you have been in enterprise technology for more than five minutes.
Nate names the fear that most people in a room like this carry but rarely say out loud. The fear of being the person who championed AI, spent the budget, and had nothing to show for it twelve months later. It is a career risk as much as a technology risk. And it is shaping more deployment decisions than anyone is willing to admit.
Three Things That Landed With Me
I do not often editorialize this directly in a piece like this. But this session earned it.
Speed is a differentiator. The organisations that move deliberately but without hesitation are the ones that reach production. Waiting for perfect conditions is a strategy for the 95%.
Iterate and do not aim for perfection. The pilots that ship are rarely the most sophisticated ones. They are the ones that launched something real, learned from it, and improved. The ones still in the lab are still in the lab.
Minimise confusion. For customers, for agents, for the internal teams building the thing. Complexity accumulates quietly and kills adoption slowly. The 5% keep it simple on purpose.
Watch the Video
Hannah Bloking, Andrei Papancea, and Nate Brown. Three questions. Fifteen minutes. The most honest conversation about AI deployment failure I have heard on a conference floor in recent times.
Watch it at CX Today. And if you are sitting on an AI pilot right now, watch it before your next steering committee meeting.
See Amazon’s playbook here.
Reported live from CCW Las Vegas 2026
