In this briefing
- 01Salesforce brings several Agent systems into one workplace entry point
- 02Kyndryl and Amazon Web Services place Agent systems in a continuing modernisation process
- 03OpenAI brings source-code storage and review into Codex
- 04Conductor's acquisition also triggers the exit of its cloud service
- →What to watch next
- ↗Sources and verification
- Salesforce already provides Google Workspace and Microsoft 365 connections, Conversation History, Skills and a US$50-per-user-per-month price option for Agentforce Coworker. Microsoft Teams and the desktop applications are planned to become generally available in November 2026 (month 11), while scheduled tasks are expected in December 2026 (month 12).
- OpenAI plans to acquire Pierre, which has already provided Codex with components for browsing and reviewing code. OpenAI says that Pierre's open-source projects will remain open source and that Code Storage customers will still be able to use the service after completion of the acquisition. The transaction has not yet met customary closing conditions, including regulatory approval.
- Meta has confirmed its acquisition of Conductor, but Conductor Cloud will become read-only on 13 November and will close on 20 November. Local Conductor is promised only new-model support and critical bug fixes until early 2027.
Salesforce brings several Agent systems into one workplace entry point
On 5 November, Salesforce announced an expansion of Agentforce Coworker. Agentforce Coworker uses Salesforce records, business metadata and connected data supported by Data 360 to plan steps, and Agentforce Coworker coordinates supported Agent systems, processes and business actions. Salesforce says that the relevant actions use applicable employee permissions, sharing rules and configured access controls, and that supported operations can use approval and audit capabilities.
Salesforce assigns different capabilities to different delivery states. Google Workspace and Microsoft 365 connections, Conversation History, Skills and a US$50-per-user-per-month price option are already available. Microsoft Teams, iOS, Mac, Windows and the Coworker MCP server are planned to become generally available in November 2026 (month 11). Daily Digest, Scheduled tasks and three industry mobile-application integrations are expected in December 2026 (month 12). Salesforce also says that pricing and availability may vary by region and customer agreement.
For enterprise adoption, one entry point can reduce the work required for users to switch among several Agent systems and applications, but one entry point cannot automatically unify the underlying records, permissions and approval rules. An enterprise needs to check the actor, target object, data source, approval condition and audit result for each business action, and the enterprise needs to place already available capabilities and planned capabilities in separate acceptance scopes. Per-user pricing can make budgets more visible, but per-user pricing cannot by itself prove that each business process has produced a verifiable benefit.
Kyndryl and Amazon Web Services place Agent systems in a continuing modernisation process
On 5 November, Kyndryl announced an integrated offering with Amazon Web Services. The offering combines Kyndryl Agentic Modernization services-as-software with AWS Transform to assess, plan and execute the modernisation of applications, infrastructure and core systems. Kyndryl says that the offering also combines engineering experience in cloud, security, compliance, FinOps and critical technology operations.
Kyndryl positions itself as a launch partner for AWS Transform, and Kyndryl describes the offering as a repeatable capability for continuing modernisation. The public announcement provides no formal pricing, common service level, complete regional list, target-system compatibility matrix or independent customer performance result. The progress percentages published by Kyndryl come from Kyndryl's own modernisation survey, so those percentages cannot replace the migration records of the target enterprise.
For enterprise adoption, an application-modernisation Agent system accesses source code, configurations, dependencies and critical business systems, so an enterprise needs to connect assessment results, planned changes, actual execution and human approval in one traceable record. Turning services into software can convert expert methods into a repeatable process, but the enterprise still needs to define which judgements, credentials, rollbacks and business-continuity results are the responsibility of Kyndryl, Amazon Web Services and the customer team respectively. A supplier partnership announcement can prove only that the offering has been made public; a supplier partnership announcement cannot prove that the offering already suits every legacy system.
OpenAI brings source-code storage and review into Codex
On 4 November, OpenAI announced its plan to acquire Pierre Computer Company. Pierre developed Code Storage and development tools including Trees, Diffs and Highlights. OpenAI says that Codex already uses Trees and Diffs to browse files and review changes. OpenAI also says that Pierre's open-source projects will remain open source and that Code Storage customers will still be able to use the service after completion of the acquisition. The transaction remains subject to customary closing conditions, including regulatory approval, and OpenAI and Pierre remain separate companies until the transaction closes.
OpenAI's Pando material published on 2 November explains the operating pressure behind this integration. OpenAI says that Codex allows more engineering work to take place in parallel, so one source-code repository receives more branch, push, merge, fetch and clone requests. OpenAI says that Pando already supports all development work in its single source-code repository. OpenAI also says that Pando reduced median push latency by nearly 80% in its internal environment and can deliver nearly 1 PB of Git data each day. These figures come from OpenAI's internal operating material and not from independent testing.
For enterprise adoption, a source-code management system is moving from being an external tool used by an Agent system to being a foundation for Agent collaboration, review, automation and audit. An enterprise needs to check the code-hosting location, branch concurrency, merge rules, continuous-integration triggers, human review, application programming interface compatibility and historical-export capability. OpenAI's retention of Git compatibility and open-source components can reduce some migration friction, but the acquisition that has not yet closed and the Codex product integration that has not yet been announced cannot be included prematurely as current capabilities.
Conductor's acquisition also triggers the exit of its cloud service
On 4 November, Conductor announced that Conductor was joining Meta. TechCrunch reported that Meta had confirmed its acquisition of Conductor, but Meta did not disclose further transaction terms. Conductor allows several coding Agent systems to work on software projects in parallel workspaces, and Conductor's Agent systems use external model providers that include Anthropic and OpenAI.
Conductor also published a specific exit timetable. On 4 November, Conductor stopped new registrations and application downloads. Conductor Cloud will become read-only on 13 November, and Conductor Cloud will close and begin deleting cloud-workspace data on 20 November. Conductor provides functions for moving cloud workspaces to local storage and exporting account data. Local Conductor will continue to receive new-model support and critical bug fixes until early 2027, after which Local Conductor will begin to wind down operation of the existing product.
For enterprise adoption, a supplier acquisition can expand product integration, and a supplier acquisition can also change registration, support, hosting and data-retention conditions immediately. An enterprise needs to check workspace-export formats, ownership of code and conversation history, local-runtime dependencies, service-closure notice periods and alternatives during procurement, and the enterprise needs to test regularly whether exported data can be restored in an independent environment. Conductor's migration functions provide a specific buffer, but the migration functions cannot prove that Meta's future product will retain the same interfaces, model choices or data-processing method.
What to watch next
- Can Salesforce publish verifiable results for permission denials, human approvals, successful execution, cross-system rollback and complete-task cost for specific business actions?
- After OpenAI completes the acquisition of Pierre, can OpenAI publish the specific scope of integration between Code Storage and Codex, customer migration conditions, service levels and methods for exporting source-code history?
- Can Meta and Conductor explain whether the new product will retain existing workspaces, model choices, data-processing boundaries and local-runtime capabilities?
Sources and verification
Golden Data has edited this briefing from the public materials listed above. The original sources govern facts and figures. The enterprise relevance sections are Golden Data editorial analysis and do not constitute an endorsement of any third-party product.
← Back to AI Daily Briefing