So much of dev work happens in the context of a larger team, and now you can bring all of that work context into the GitHub Copilot CLI with Work IQ. Try it out: https://lnkd.in/gGPd-FWr

Like Comment
Transcript

Much of the context that AI agents need doesn't actually live in your application. It's actually in documents and emails and meetings and SharePoint your company data. Work IQ lets you bring the full context of your organization's data to any agent, like the Copilot CLI and Agentic AI assistant. Write in your terminal, and all you need to do is add a single MCP server. Once it's added, the agent can easily pull in any organizational data. All you have to do is ask. This is the missing piece of context for most applications, and it unlocks powerful workflows that you can use today. Here's a spec that was written in Word and it lives in SharePoint. I don't have to summarize this by hand, I just give Copilot the title of the document and a link to the repo and ask it to compare the spec to the implementation. Copilot. Finds the right spec through Work IQ, inspects the code and compares the two. It shows which requirements were implemented, which ones drifted, and what's missing. That's not just a search result, that's delegated execution. Now let's look at a harder problem. We had several meetings about a File Explorer version control feature, and nobody wrote down the architecture. So I asked Copilot to find the implementation plan. Draw it, Copilot reasons over meeting transcripts and emails. From work IQ and reconstructs the architecture and summarizes the decisions. These decisions were never formally captured, but they still exist under the hood. Copilot explores the problem space, plans the work, and reviews the result as it goes. It can select the right model per step without you having to wire up specific routing logic. You can also pick your model with a simple slash command. All of this without ever leaving the agent. Build real intelligence into your agent today with work IQ.

When developers spend less time managing context and more time creating value, the impact compounds across teams and organizations.

Satya Nadella This is a big step forward. Development is rarely a solo activity, and bringing real team context into Copilot CLI makes AI assistance far more practical and impactful. Work IQ feels like the missing layer that connects individual coding tasks with broader project goals, collaboration, and decision-making. Excited to see how this evolves for modern engineering teams.

Bringing more team and system context into AI tooling is a big step forward. One thing that becomes increasingly important at that point is having a fail-closed boundary between AI output and execution — especially when suggestions start affecting real infrastructure, costs, or safety. Context helps models propose better actions. Deterministic validation is what decides when those actions are actually allowed to execute.

The spec-to-implementation comparison is the enterprise unlock here. Most orgs have requirements in SharePoint, decisions in meeting transcripts, code in repos - all disconnected. The drift between "what we decided" and "what got built" compounds invisibly until something breaks. Surfacing that at the CLI - where developers actually live - beats dashboard reporting that nobody checks. The SDK making this embeddable is smart. Context shouldn't be a destination you visit. It should follow you.

This isn’t the first AI SDK. But it may be the first mainstream, enterprise-grade move that treats work context as a first-class input — not an afterthought. What’s being opened up here isn’t just Copilot, but the agent loop itself, grounded in how teams actually work. That transforms AI from a surface-level assistant into something that can reason across decisions, ownership, and tradeoffs over time. Productivity gains are easy to cite. The harder — and more consequential — step is making work legible to machines in ways humans can still trust, audit, and govern. When team context travels with the work, AI stops just helping people write code and starts helping teams remember why decisions were made. That’s the difference between short-term efficiency and durable progress.

"Great insights, Satya Nadella While AI is transforming biology, at NYUROX, we are pushing the boundaries by using Computational Physics to predict neuro-metabolic shifts. 🧠 We’ve built an engine that detects Alzheimer’s and neuro-decay 18 months earlier than traditional structural imaging (MRI). Our goal is to shift healthcare from 'reactive' to 'predictive' by mapping physics biomarkers. Building this from India, for the world. Would love to see how Azure's high-performance computing can further scale our physics-driven diagnostic engine! 🇮🇳🚀 #DeepTech #HealthInnovation #NYUROX"

The context problem is one of the most underappreciated challenges in enterprise AI adoption. Teams have been building powerful AI tools, but without understanding the organizational context - the why behind decisions, the history of choices, the team dynamics - they often miss the mark. Work IQ bringing that context into the development workflow is a smart move. We see similar patterns in government technology implementations - the tools that succeed are the ones that understand the institutional context, not just the technical requirements.

I first started using GitHub Copilot CLI back in September 2025, right after its public preview relaunch, both non confidential professionally and personal projects. It immediately became obvious that this wasn't merely another coding assistant or developer-focused utility. With well-crafted prompts, the appropriate model selection, and thoughtful guidance, it transforms into a near-autonomous computer management assistant — remarkably capable of orchestrating multi-step workflows, performing system administration tasks, manipulating files and directories, setting up development environments, running scripts, helping debugging issues, and handling a wide range of terminal-based operations. Furthermore with correct guidance and now top tier models it capable to emulate work performed by other more complex software. It is definitely a game changer in terms of productivity that many in the software engineering world are just starting to realize.

Satya Nadella We’re moving from "writing code" to "directing AI." The future isn't just about technical skills—it’s about having the vision to lead a fleet of digital minds to build something great

The real leap forward lies not in "stronger models," but in reconnecting the work context that is scattered across emails, meetings, and code. Context is productivity.

See more comments

To view or add a comment, sign in

More from this author

Explore content categories