Microsoft

Confidence 0.85 · 11 sources · last confirmed 2026-10-25

Microsoft is the cloud-and-developer-tools giant that owns GitHub (since 2018), partners deeply with OpenAI, and ships the developer surfaces central to the wiki’s agentic-coding corpus: Azure, Azure DevOps, Visual Studio Code, and the responsible-AI pipeline that proxies Copilot requests. Promoted to an entity page on 9 June 2026 via cross-page-presence promotion (precedent: LangChain, Andrej Karpathy): Microsoft is a plain-text mention across ~21 source pages, and the Agentic DevOps keynote — published on the Microsoft Visual Studio channel — is the first source centred on Microsoft’s own agentic developer tooling.

Backlinking note. source_count counts formal inbound source-page wikilinks (7 after the 2026-10-15 ingests), not ~21 (plain-text mentions). Backfilling the historical mentions into wikilinks is a deferred lint cleanup.

Role in the wiki

Microsoft is the platform-and-cloud vantage on agentic development — the vendor whose IDE (VS Code), cloud (Azure), ALM tooling (Azure DevOps), and safety layer (responsible-AI pipeline) host the agentic workflows that GitHub’s Copilot drives. The channel-author convention applies: Microsoft Visual Studio (the YouTube channel) maps to Microsoft as the publishing entity.

Visual Studio Code

Microsoft’s cross-platform editor — the keynote’s hub for agentic development. The four Copilot interaction modes: Ask / Edit / Agent / Plan (Plan is the newest — the agent states its implementation plan before acting, bounding the over-autonomy failure mode). Instructions files encode house rules + exemplars; agents run locally (CPU/GPU) or push to GitHub’s cloud; profiles swap toolsets; Ollama + the Continue extension enable offline/local-model agents.

Azure DevOps + the responsible-AI pipeline

Copilot agents are coming to Azure DevOps (private preview) — but routed through GitHub (“you still have to link it up to GitHub”). Every Copilot request passes through Microsoft’s responsible-AI pipeline, which validates input and the returned code (security vulnerabilities, NSFW) even when the chosen model is third-party (Anthropic/Google) — sending only enough across the wire and checking the response before it reaches the developer, in milliseconds. See responsible-ai.

Azure SRE Agent

Microsoft’s autonomous cloud-monitoring + incident-response agent (preview). Scope it to an Azure resource group; it performs incident management, can be permissioned to autonomously remediate certain conditions, and files health/security reports — framed as “a site reliability engineer on my team.” A worked instance of an autonomous agent in the IT-operations (not just coding) domain.

OpenAI partnership

Microsoft hosts and routes OpenAI models (and now the Codex agent inside Copilot), alongside Anthropic and Google models. The keynote frames model choice as a developer decision, with Microsoft’s proxy doing the safety validation regardless of model.

Satya Nadella / Build 2026 — the CEO-worldview source

The Possible interview (5 Jun 2026, post-Build 2026) is the wiki’s first platform-CEO worldview source and Microsoft’s second formal inbound source. Nadella frames AI as the future of the firm (not a technology), introduces token capital as the firm-level asset that compounds alongside human capital, and lays out Microsoft’s enterprise-AI stack as the manageability/security answer: the hill-climbing machine (evals/objectives as the new IP), Agent 365 (agent identity/sandbox/policy/observability via Entra/Defender/Purview — see responsible-ai), Foundry asserts (runtime boundary enforcement — see agent-harness), and the agentic development environment (the GitHub app as an inbox of agents with Canvas, for micro-steering of macro-delegation). The dictum “don’t use frontier models for non-frontier problems” anchors his token-efficiency argument. Co-designed silicon (Maya, Cobalt) and the platform-company-defined-by-value-created-on-top trust doctrine round out the strategy.

WorkLab podcast — the owned-media future-of-work vantage

Microsoft’s WorkLab podcast (host Molly Wood) is a separate publishing channel from Microsoft Visual Studio — owned media focused on the future of work broadly, not Microsoft’s own developer tooling. Wood (29 Jun 2026) is the wiki’s first WorkLab source: an interview with LinkedIn’s Chief Economic Opportunity Officer Aneesh Raman on AI-era career and organizational identity (“onlyness,” the 5 C’s, the electricity-bolt-on-failure analogy). The channel-author convention applies here as it does for Microsoft Visual Studio: WorkLab maps to Microsoft as publishing entity, with LinkedIn (a Microsoft subsidiary since 2016) as the guest’s employer — a third, futures-of-work-focused channel alongside the developer-tooling (Agentic DevOps) and CEO-worldview (Nadella/Possible) vantages already held.

WorkLab’s second wiki source, Wood (10 Aug 2026), is a live recording from the Copilot Summit — so the venue as well as the channel is Microsoft’s, and the audience is Microsoft customers. Guest Allie K. Miller (Open Machine) is an independent advisor rather than an employee, but two of the statistics she cites are Microsoft first-party research: an 1,800-employee study finding a 22-point lift in critical thinking and a 30-point lift in trust of agentic AI where managers model AI use themselves, and the Work Trend Index split in which 65% of AI users fear falling behind while 45% say it feels safer to focus on current goals than to redesign their work. The episode’s durable contributions are structural rather than statistical — the frontier unit (a 60–120-person cross-disciplinary experimentation group on a higher per-head token budget), the Gremlin-mode ladder of usage depth, and the context layer prescription (“make your whole company queryable”).

WorkLab’s third wiki source, Wood (24 Nov 2025), was recorded at Harvard Business School and promotes the Frontier Firm AI Initiative, a research collaboration between Microsoft and HBS for which Sadun is a research leader. “Frontier Firm” is Microsoft’s own term for the AI-first organisation, so here the publisher is a partner in the research being discussed. The initiative’s own page, hosted by the HBS AI Institute, describes the collaboration: member firms run experiments with Harvard and Microsoft researchers, starting with sales, IT and finance.

Microsoft Advertising — the answer-engine owner selling into the shift

Longo & Marchiaro (HubSpot UNBOUND, Oct 2026) gives Microsoft’s view as the owner of a search engine (Bing, one billion monthly users) and an answer engine (Copilot): three eras of the web running at once, paid placement (Performance Max, AI Max) on conversational queries and AI surfaces, and Microsoft Clarity’s free AI-visibility metrics (citation rate, share of authority by domain). The session launched the first HubSpot–Microsoft Advertising integration. Its statistics are mostly Microsoft first-party research.

Open questions

• Full backlinking. ~21 plain-text “Microsoft” mentions across the corpus await promotion to wikilinks.
• Product sub-pages. Azure, VS Code, Azure DevOps, and the Azure SRE Agent are Dangling product mentions; promote on a second source each.
• The Microsoft–OpenAI commercial relationship. Referenced but not characterised; a dedicated source would let the wiki track it properly.

Mentioned in

• 2025-12-22-randell-gousset-microsoft-agentic-devops-in-real-life — the promotion trigger: VS Code modes, Azure DevOps, responsible-AI pipeline, Azure SRE Agent.
• GitHub — Microsoft subsidiary; the developer-platform half of the keynote.
• 2026-06-29-raman-wood-worklab-job-titles-dont-matter-2026 — WorkLab podcast; Aneesh Raman (LinkedIn) on AI-era career and organizational identity.

Its own transformation, told by its workforce-transformation lead (added 2026-10-17)

Katy George on WorkLab (July 2026): George, corporate vice president of workforce transformation, describes Microsoft’s internal practice. The first Copilot rollout to its own sales force was run “as a typical product launch” and people “tried it and then they stopped using it”; Harvard Business School wrote it up as a case in the Frontier Firm research. What replaced it: outcome goals, and redesign of tech, process and people with the people who do the work. Programmes named: Camp Air, a bootcamp that whole engineering teams attend together, and weekly role-based huddles in sales, taught practitioner to practitioner. Her team has written 100 internal case studies of AI transformation, whose common pattern is top-down goals with bottom-up citizen-developer builds. The internal audit team is her example of “Capability Add”: “every internal audit now delivers proactive risk identification.” Microsoft’s own account, on its own channel.

Tools and Weapons, and two earlier WorkLab episodes (added 2026-10-25)

• Roslansky (LinkedIn) on Tools and Weapons (Mar 2026): Brad Smith, Microsoft’s vice chair and president, interviews LinkedIn’s CEO about Open to Work; LinkedIn’s builder role and associate-builder entry route; Smith’s “schism” between AI that outperforms people and AI that empowers them, with Microsoft on the second side.
• Bornet on WorkLab (Oct 2025): a WorkLab episode from October 2025 on agents, the Frontier Firm’s traits, and the Work Trend Index finding that more than 80% of leaders plan to use agents to fill capacity gaps.
• Diasio on WorkLab (Dec 2025): EY’s Global Consulting AI Leader, introduced as a member of the inaugural class of the Frontier Firm initiative that Microsoft runs with Harvard. New joiners as “managers on day one” of agents, mindset–skill set–tool set, and the billable hour as EY’s hardest change.