AWS × Claude In-House — Serverworks On How Japanese Enterprises Win the AI-Agent Era

PIVOT's top-runner program hosted Serverworks CEO Oishi and AWS Japan executive Yasuda on the theme 'pushing the productivity ceiling with AI — enterprise strategy in the AI-agent era.' The opening frame: generative AI does not exist without cloud. Training requires large-scale compute; consolidating enterprise data into something AI can actually use requires an integrated platform like AWS. The two are structurally inseparable.

Inside that AWS stack, Anthropic's Claude is the model with the most momentum. The packaged offering — use Claude securely through Bedrock — is what Japanese enterprise customers are signing for. Serverworks operates Claude on Bedrock in production as an AWS partner, and the conversation is built on that real operational footing.

Oishi proposes AIDLC (AI-Driven Lifecycle), a new way to build software. An engineer and a domain expert sit together and use AI to build an entire application end to end in a single session. Tooling now supports the loop from spec through code, so productivity steps up dramatically. The approach is in customer deployment, sitting at the centre of Serverworks' offering.

AWS brings its own Kiro coding agent, and an April OpenAI partnership brought Codex onto Bedrock. Inside that multi-model lineup, Yasuda notes Claude is decisively the most popular. The honest counterpoint: using Claude naively gets expensive fast. Bedrock's value is letting you route simple coding or QA work through cheaper open models — some at zero inference cost — and reserve Claude for what actually needs it. That's the hybrid design that keeps the bill sane.

The most pointed moment was Serverworks' governance line: they will not build a 'CEO agent.' Encoding the founder's judgement into a long-running agent would mean his shadow stays bolted onto the company for decades, costing it the ability to adapt as the era changes. Don't preserve the CEO's residue — preserve the values, and put those into the AI. This is a design choice about what *not* to automate.

Closing, both speakers gave a message to viewers. AWS positions itself as the foundation and technology provider supporting customer experimentation. Serverworks positions itself as the AWS partner that holds a moat and a long-horizon plan. The shared frame: generative AI's effect on business performance arrives with delay, not this quarter. The companies that can hold the moat and keep executing over a medium-to-long horizon will be the winners of the generative AI era.