Opinion
AWS Summit London 2026: The Adoption Gap Is the Real Story
Published on 17 Apr, 2026 by Jonathan
Every AWS Summit keynote produces a flurry of product announcements, and this one was no exception. But the slide that stuck with me wasn't a product launch. It was a statistic:
"64% of UK organisations have adopted AI, but only 24% are using it at advanced levels"
That gap, between adoption and impact, is the strategic question every business needs to answer over the next eighteen months. Everything else AWS announced is, in effect, a tool for closing it.
The three pillars, and what they actually mean
AWS framed the keynote around infrastructure, Gen AI, and skills. Translated into business terms:
Infrastructure is becoming a price-performance race.
The Trainium 3 announcement alongside the full Blackwell lineup (P6e GB200, GB300, B200, B300) signals that AWS wants to be the place you run AI workloads regardless of chip preference. For organisations making multi-year commitments, this is reassuring. You don't have to bet on a single silicon winner.
Gen AI is moving from chatbots to agents.
The most commercially significant reveal was Amazon Quick, an agentic application that goes from insights to action in minutes, bundling research, BI, custom agents, workflow creation, and end-to-end process automation. This is AWS directly competing with the ChatGPT Enterprise and Copilot category, but leaning on its data gravity advantage.
Skills are the real bottleneck.
The slide reframed this as moving from a "skills gap to skills advantage", and I think that framing is correct. The companies that win won't be the ones with the most AI licences. They'll be the ones whose people know what to do with them.
The productivity numbers are real, and they matter
The Kiro case study was the most quoted slide of the day: rebuilding the inference engine behind Bedrock was originally estimated at 40 developers over 365 days. With Kiro, it took 6 developers and 76 days. Motorway reported a 250% increase in deployments. ADP hit a 90% reduction in manual effort documenting legacy business rules using Amazon Transform.
The interesting question isn't whether these numbers are achievable. It's what they imply for your competitive position. If a competitor can ship a major rebuild in 20% of the time with 15% of the engineering resource, product roadmap assumptions need revisiting this quarter, not next year.
Three things worth paying attention to
Three announcements deserve a serious look from any leadership team planning their AI investments:
1. Agentic workflows are now a product management concern, not just an engineering one.
Tools like Kiro's custom agents and steering files mean that how you codify standards, specs, and CI patterns directly determines how much leverage you get from AI coding tools. That's a product decision about your own internal developer experience.
2. Lambda durable functions and the broader shift to long-running AI workflows
(up to a year, no additional infrastructure) change the unit economics of agent-based features. Capabilities that were uneconomic on a per-request basis are now viable.
3. S3 is quietly becoming the universal data substrate.
Object, tabular, metadata, and vector storage in one service, with Iceberg, Spark, Postgres, OpenSearch, and Kafka all sitting on top. For data strategy conversations, this reduces the argument for specialist vector databases considerably.
The uncomfortable conclusion
The announcements were strong. The demos were convincing. But the 64% | 24% slide is the one that should drive planning across the industry. Adoption is table stakes now. The advantage goes to whoever moves from pilots to production fastest, and that's a leadership and skills problem far more than it's a technology one.
Everyone is buying the same tools. What you do with them is the only thing that will matter.
Thinking about where your organisation sits on the 64/24 curve? We'd welcome the conversation. Get in touch at info@green-custard.com to map out the next step together.