Abstract
Version 2 answers the four questions buyers actually ask. Which assistant should employees use: Microsoft 365 Copilot, Gemini Enterprise, ChatGPT Enterprise or Claude Enterprise? Which models should do the work? Should those models be called directly or through Amazon Bedrock, Microsoft Foundry or Google Cloud? And which platform should run the agents and automation?
Each decision has a different owner, rate of change and cost of reversal. The guide gives a dated default for each, a model evaluation method and scorecard, a map of which cloud serves which model as of 29 October 2026, and a build-tier ladder from assistant features and low-code tools to pro-code runtimes and orchestration.
The platform review compares cloud, application-native, data-native and independent platforms, including Microsoft, AWS, Google, Salesforce, SAP, ServiceNow, Oracle, Snowflake and Databricks. The selection method starts with mandatory gates, then a weighted scorecard, ten scenario starting stacks, a five-week production-shaped proof and sixteen diligence questions.
Decision it supports
Which assistant, models, model-access path and agent platform to choose, and where native extensions are justified.
What you will produce
A four-decision record, a model evaluation, a hard-gate assessment, a weighted finalist scorecard and a five-week proof plan for one named workflow.
See it worked through in a guide
Claims package 1.1.0 · 26 October 2026. Synthetic worked example: team A, group disability, region 2. Read-only preparation of the assigned case; examiners retain every case decision. Hold candidate 1.0.0 after an unrelated employee’s claim is returned. In the synthetic decision, accept 1.1.0 for a monitored team A rollout only after prior-claims history is disabled and the remaining scope passes a fresh evaluation. All figures, evidence and sign-offs in that guide are invented; the paper provides the underlying method.
In the paper
Three passages, as published
No universal winner
There’s no one best agent platform on the market.
Roadmaps are not controls
A roadmap promise isn’t a production control.
The actual test
The winning platform isn’t the one with the broadest demo.
Next step
Discuss this paper against your own program.
A free sixty-minute working session with Richard on your top gaps. You leave with a one-page summary within three business days, and three next questions.