Most organisations have a wiki full of hard-won how-to knowledge - how to request access to a SharePoint site, how to raise an IT ticket, how to onboard a new starter - and almost nobody can find the right page when they need it. If that wiki lives in Azure DevOps, you have two obvious but flawed options: send people to search the wiki manually, or point a generic Copilot at the problem and watch it confidently answer from its training data instead of your actual policies.

There’s a better path, and it comes in two halves:

1. The Azure DevOps Wiki Copilot connector indexes your wiki into Microsoft 365, so Copilot can search it.
2. A declarative agent with strict instructions turns those search results into an assistant that answers only from the wiki, cites the page it used, and admits when the answer isn’t there.

The second half is the one people skip - and it’s the one that earns your users’ trust.

A colleague pulls a late night and rescues a client demo that was heading for the rocks. You want to say thank you properly. In most places that means leaving the conversation you are in, opening the recognition site, finding the right form, filling in a handful of fields, and submitting. By the time you are done, the moment has passed.

Now imagine doing it without leaving Microsoft 365 Copilot. You type:

Give kudos to Alan Partridge for being excellent with client engagement.

A second later a small card appears inside the chat, with Alan Partridge already selected, a value chosen, and a thank-you message drafted for you to read over before you send it. Ask instead to see the wall and the same experience opens full screen, with the recognition feed, team filters, and leaderboards.

The compose card, rendered inside the conversation: Alan Partridge resolved, a value chosen, and a thank-you message drafted for you to review. Not a text reply describing one.

That card is not text, and it is not a link out to another site. It is a real React component running inside the Copilot canvas, built with the SharePoint Framework. In this post I follow that one sentence the whole way through, from the words you type to the card you see, and explain what happens at each step. The example is a sample I built, the Kudos & Recognition Wall, but the mechanics are the same for any SharePoint Copilot App.

You have a SharePoint document library packed with IT policies, each tagged with custom metadata columns: Document Status, IT Audience, Next Review Date, and a Document ID. You build a Copilot Studio agent and connect it to that library as a knowledge source. A user asks:

“Which IT policies are Active, targeted at Contractors, and due for review before the end of 2026?”

The agent returns a partial list. At the bottom of its response, it adds its own caveat:

“The search results do not include review date metadata for all documents, so this may not be a complete list.”

The agent already knows it’s failing. This is not a prompt engineering problem, and it is not a configuration mistake. It is a fundamental constraint in how Copilot Studio ingests SharePoint content, and it affects every organisation that relies on custom document library columns to classify and manage their content.

If you’ve been following the Microsoft 365 Copilot space, you know the APIs have matured rapidly. We now have six distinct APIs — Retrieval, Search, Chat, Meeting Insights, Change Notifications, and Usage Reports — each solving a specific integration challenge. But finding clear, practical guidance on all of them in one place? That’s been harder than it should be.

In this post, we’ll walk through every Copilot API available today, explain when to use each one, and provide real TypeScript examples you can adapt for your own projects. I recently published an open-source SPFx sample on the PnP community that demonstrates four of these APIs in action — I’ll reference that throughout.

In Part 1, we explored the fundamentals of Heft in SharePoint Framework 1.22. Now it’s time to unlock the full power of Heft by building custom plugins! This comprehensive guide walks you through creating production-ready Heft plugins with TypeScript, using a real-world version management scenario as a practical example.

This is Part 2 of a two-part series:

• Part 1: Understanding Heft in SharePoint Framework 1.22 - Heft basics, rig system, and webpack patches
• Part 2 (this post): Building Custom Heft Plugins for SharePoint Framework

SharePoint Framework (SPFx) 1.22 introduced a major architectural change by replacing the traditional Gulp-based build system with Heft - Microsoft’s modern build orchestrator from the Rush Stack ecosystem. This comprehensive guide will walk you through understanding Heft, getting started with the new toolchain, and customizing webpack configuration.

This is Part 1 of a two-part series:

• Part 1 (this post): Understanding Heft basics, rig system, and webpack patches
• Part 2: Building Custom Heft Plugins for SharePoint Framework

Master three distinct approaches to integrate Copilot Studio agents with SharePoint Framework applications - from zero-code deployment to full programmatic control.
This comprehensive guide takes you from simple SharePoint channel publishing to advanced custom API integration, showing you when and how to use each approach for maximum impact.
Whether you need rapid deployment or pixel-perfect branding, find the perfect integration strategy for your requirements.

A deep dive into building a sophisticated, multi-channel AI agent for project management. This guide explores the architecture, key technologies (Agents SDK, Semantic Kernel, Azure AI), implementation patterns, and strategic considerations for developers building enterprise-grade intelligent assistants.

The Power of Direct API Integration

In today’s AI-driven workplace, Microsoft 365 Copilot has become an indispensable tool for productivity. However, its true potential is unlocked when integrated with organizational knowledge sources. While OneDrive/SharePoint integration provides a simple solution, and custom engine agents offer maximum flexibility, Azure AI Search strikes the perfect balance - giving developers control over indexing and relevance without the complexity of maintaining full language model deployments.

In today’s data-driven world, organizations are increasingly looking for ways to leverage their vast repositories of documents to enhance productivity and decision-making. Microsoft’s Kernel Memory (KM) is a powerful tool that enables developers to build sophisticated AI-driven document chat experiences. In this blog post, we’ll explore what Kernel Memory is, how it works, and how you can use it to build a Document AI Chat experience.