Microsoft's low-code platform for building custom enterprise AI agents on top of Microsoft 365, Dynamics, and Azure. Independently assessed by BenchLytix using proxy methodology.
Scores are estimated from documentation until an agent is runtime-measured. Methodology v2.6.0 rev 1.27.
Week of 2026-05-04 · Manually assessed
Methodology v2.6.0Public-evidence proxy across four pillars. Not directly comparable to indie agent scores.
Manually assessed by BenchLytix · Week of 2026-05-04
Analyst est.Pillar scores are derived from publicly available documentation (SLAs, certifications, pricing pages, trust center). They do not represent live API performance measurements. Maximum composite: 79/100.
Public sources used to derive the proxy score. Last reviewed . Per the incumbent methodology, each citation maps to a sub-component in the rubric.
SLA: Azure default contractual SLA is 99.9% across most services. Copilot Studio runs on Azure with this SLA tier. Status page: status.cloud.microsoft and status.azure.com both show historical data >12 months back. Incident history: Azure status shows 3-5 P1-class incidents trailing 12mo (typical for hyperscaler scale). Changelog: Microsoft publishes monthly+ release notes for Copilot Studio + Power Platform. Reliability case study: not specifically a "reliability/uptime claim" case study — generic enterprise customer studies exist but are feature-focused. Conservative: FALSE. Pillar 2: no formal Copilot-Studio-specific latency benchmark from Microsoft or third party. Streaming + edge documented (Azure regions). No response-time SLA. → L = 30. Pillar 3: $0.01/credit pricing public, sufficient to estimate, limited free trial via M365 — qualifies under boundary case #6 (no sales contact required). Pillar 4: canonical 90 (per Pillar 4 audit 2026-05-04). Pentest false per strict reading.