LSEG

LSEG has put its data into more AI platforms than any other incumbent. Since October 2025 it has launched MCP access in Microsoft Copilot Studio, Claude, ChatGPT, Amazon Quick and Gemini Enterprise, a Copilot connector in Excel and data shares in Databricks and Snowflake, backed by a ten-year Microsoft partnership. Its content — Reuters, I/B/E/S, Datastream, Worldscope and exclusive Morgan Stanley aftermarket research — is among the deepest in our coverage.

Workspace's own AI arrived late and is thinly documented. AI Search reached general availability in June 2026 with release notes conceding that some citations are missing; Deep Research and Company Intelligence followed in June and July. No accuracy evaluation, public trust centre, no-training commitment or MNPI controls were located, there is no expert content and no client document ingestion in Workspace, and pricing layers a Deep Research add-on, use-case licences and usage charges on top of the terminal.

We initiate at 5.96, Emerging, with an Improving outlook. Delivery surfaces and data are strengths; Commercial Value (4.17) and Enterprise Readiness (6.00) weigh on the score. Published security documentation and evaluation, and client data in Workspace AI, would lift it to Established.

Strengths

• Distribution. MCP in Claude, ChatGPT, Copilot Studio, Gemini Enterprise and Amazon Quick; Copilot in Excel; Databricks and Snowflake shares.
• Microsoft partnership. Ten-year alliance (2022) with native Teams, Excel and Copilot surfaces.
• Content. Reuters, I/B/E/S, StarMine, Datastream, Worldscope and exclusive Morgan Stanley aftermarket research.
• Agent traction. AI Search at 17,000 active users and Deep Research users up four times in a quarter (July 2026).
• Identity controls. Microsoft Entra SSO with Conditional Access and SCIM provisioning.
• Commitment. A ten-year Microsoft alliance and a dedicated Group Head of Enterprise AI.

Cautions

• Late native AI. AI Search reached general availability only in June 2026, with some citations acknowledged as missing.
• Disclosure. No public SOC 2 or ISO 27001 evidence, no-training commitment, AI subprocessor list or MNPI controls located.
• Client data. No document upload or client data in Workspace AI; only via Databricks or Snowflake.
• Expert content. None.
• Cost layering. Deep Research add-on, use-case licences for data in AI applications and MCP usage charges.
• Evaluation. No published accuracy measurement.