AWS launches APEX, a shared agentic platform for Wood Mackenzie built on Bedrock AgentCore
Wood Mackenzie reports that 88 % of AI proofs‑of‑concept never reach widescale deployment and only about a quarter of enterprises have managed to scale agents beyond pilot projects, according to surveys through early 2026. The main blockers are architectural – lacking evaluation, observability, governance and identity – rather than model quality. To close that gap the company built APEX (Agentic…
Key points
- Wood Mackenzie found 88 % of AI PoCs never reach widescale deployment, with only ~25 % of firms scaling agents (2026 surveys).
- APEX, built on Amazon Bedrock AgentCore, offers managed, model‑agnostic runtime, auto‑scaling, guardrails, identity, observability, and a unified control‑plane studio.
- AgentCore’s consumption‑based pricing bills only active CPU/memory per second, eliminating idle‑time costs for agent workloads.
APEX provides a managed, model‑agnostic environment that can host any open‑source agent framework (Strands, LangGraph, LangChain, CrewAI, etc.) and any model (Claude, GPT‑4.1, Amazon Nova, Mistral, Llama). It automatically scales from zero to thousands of concurrent invocations, enforces built‑in guardrails and IAM‑based identity, and streams telemetry to CloudWatch. APEX Studio offers a single control‑plane where teams can build, test, deploy and monitor agents without touching underlying infrastructure, cutting development time from months to hours. Consumption‑based pricing charges only active CPU and memory per second, avoiding the idle‑time costs typical of agent workloads.
By standardizing the agent stack, Wood Mackenzie expects teams to focus on business logic rather than re‑creating authentication, memory and observability layers, accelerating the move from experimentation to production and laying groundwork for future multi‑agent systems.
The story so far
10 episodes →- AWS launches APEX, a shared agentic platform for Wood Mackenzie built on Bedrock AgentCorethis story
A shared agentic platform for Wood Mackenzie, on Amazon Bedrock AgentCore
AWS Machine Learning Blog · 17 September 2026
Loading the full article…
This text was published by AWS Machine Learning Blog and written by Shridhar Navanageri. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
Coverage and discussion
3sources- Migrating multi-model AI agents to Amazon Bedrock AgentCore runtimePrimary source · AWS Machine Learning Blog ·
- The new AgentCore runtime: Elastic, optimized, and consistently fast startsPrimary source · AWS Machine Learning Blog ·
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Agents & Tools
All →- Anthropic agents submitted 20 incomplete visa forms on State Dept. site, sources say · 1 src
- AWS launches Anchor Browser with AgentCore payments for AI agents · 1 src
- Geoffrey Huntley blocks agents from bypassing diagnostics library with Git hooks · 1 src
- Anthropic releases free crash course on building AI agents with Opus 5.5 · 1 src
- Meta's Muse AI permits explicit adult conversations, unlike other Meta apps · 1 src
Comments
via GitHub Discussions