What this is
This week AWS added Claude Sonnet 5.5 to Bedrock (AWS's hosted model service — think of it as a "supermarket for enterprise models"). Sonnet 5.5 is positioned as cheaper than Opus 5.5 but faster than the previous generation — an "execution" model. By execution, we mean work where the goal is already clear and the only remaining job is to finish it: fixing bugs, generating SQL, building spreadsheets, drafting architecture diagrams.
Anthropic is laying out a clean division of labor: Opus 5.5 handles the work that requires judgment — chaining multiple PRs, security reviews, contract audits. Sonnet 5.5 handles the work where "the path is already clear" — alert response, Agent (AI assistants that autonomously complete multi-step tasks) monitoring, UI testing, fast in-IDE coding. The two are meant to be paired, not chosen between.
How the industry sees it
AWS's real selling point isn't the model itself — it's the enterprise-grade infrastructure underneath: data never leaves the AWS region, access control runs on existing IAM, auditing runs on CloudTrail, and the bill folds straight into the monthly AWS statement. For enterprise IT teams, this is the "fewer meetings, fewer approvals" pitch — no need to onboard a new vendor.
But the risks and counterarguments are real. On one hand, Sonnet 5.5's focus on well-scoped work means it's not a stronger model — it's a cheaper way to run established workflows. It can't replace real judgment work; enterprises that want AI to support actual decisions still need Opus or pricier tiers. On the other hand, by leaning its distribution weight onto AWS, Anthropic's long-term bargaining power will erode, and multi-cloud customers may find themselves edged out.
What it means for regular people
For enterprise IT: Companies already on AWS face near-zero integration cost. Companies not on AWS need to reassess their multi-cloud strategy — Anthropic has clearly made AWS its main battleground.
For individual careers: Sonnet 5.5 targets in-IDE coding and document generation, so programmers and consulting/analytics roles will feel the change first. "Drafting your own first version" work will speed up, but the part where you "decide whether that draft is worth anything" won't disappear.
For the consumer market: No direct consumer-facing product is coming in the short term, but as enterprise AI adoption accelerates, it'll indirectly drive up demand for data compliance and private deployment.