How to Implement Copilot in Your Organization

This was our advice on deploying Microsoft 365 Copilot a year ago (September 2023), two months BEFORE the M365 Copilot GA date of Nov 1. If you're a Gartner client, we hope you followed some of these recommendations to get ahead (and it's still not too late): ✅ Establish new generative AI skills and policies by evaluating Microsoft Copilot as a new technology stack rather than merely a productivity tool. ✅ Establish a M365 product team with direct oversight of governance of generative AI services that interact with the Copilot stack. ✅ Review and communicate to stakeholders key Microsoft online service terms and data protection and privacy commitments, all of which apply to M365 Copilot. ✅ Reinforce the need for information governance and access controls in M365 with stakeholders to ensure users don’t overshare information that could be exposed through the Copilot stack. ✅ Maximize adoption and reduce features overlap by coordinating with business unit leaders on use of Copilots and other generative AI tools from enterprise applications. ✅ Lead a coalition with your stakeholders to make your initial Copilot investments immediately valuable, and pave the way for an impactful and successful long-term integration of multiple generative AI technologies. ✅ Plan for a multivendor generative AI portfolio that includes Microsoft alongside other vendors, each likely with different approaches. ✅ Prioritize the rollouts to employees in a controlled way. A “big bang” rollout will cause confusion among employees, leading to a surge of support issues. ✅ To ensure ROI is achieved, plan and execute a meticulous rollout strategy that includes a series of communications, multichannel training and support, and a holistic change management strategy with buy-in from executives and business leaders. From: "Assessing the Impact of Microsoft’s Generative AI Copilots on Enterprise Application Strategy" https://lnkd.in/ewXescAp

In an interview with The Information, the CIO of Chevron indicated that about 20,000 employees are testing Microsoft Copilot, but, he said, “the jury is still out on whether it’s helpful enough to staff to justify the cost.” As a reminder, the cost of a Copilot license is ~$30 per user per month (although they probably pay less with that many licenses). Here’s my opinion on this: If a company can’t justify $30 for Copilot (or ChatGPT, Gemini or Claude), then it is more likely due to a lack of education, training and planning, than it is to a deficiency in the AI’s capabilities. This is both a challenge for the company licensing the technology, and a weakness in how the AI tech companies are selling and supporting the platforms. How do we solve this? Here is a five-step framework I’d recommend to businesses of all sizes: 1) Pilot with small groups in select departments over a 90-day period. Prove the value and create internal user champions, then scale it. 2) Prioritize use cases specific to employee roles and responsibilities. Break their jobs into bundles of tasks, and then assess the value of AI at the task level. Pick 3 - 5 use cases initially for each person that will have an immediate and measurable impact. 3) Provide generative AI education and training to maximize the value. Tailor learning journeys for individuals that include specific coursework and experiences in your core AI platforms. 4) Monitor utilization. Invest in the employees who are actually experimenting with and applying tech. Remove the licenses from employees who don’t use them. 5) Report performance versus benchmarks (before and after LLMs). In short, have a plan. The value is absolutely there when it’s rolled out in a strategic way, and part of a larger change management plan. 

I have been doing AI tool training and education for over two years now. One thing that I hardly ever see, but is so beneficial for an organization, is a prompt library. My advice: - Identify specific use cases where generative AI can help your teams save real time or get better results. - Create prompts that solve those use cases and make tasks easier. - Save those proven prompts into a shared library that's easy for everyone in the organization to use. - Spread the knowledge by encouraging the whole organization to use the library, so everyone benefits from the gains in productivity. It doesn't matter which AI tool your organization uses. It can be ChatGPT, Copilot, Claude, Perplexity, or a custom internal AI tool. What matters is how you use them. Learning how to prompt is only the first step. The real value comes from capturing successful prompts that solve important business challenges. Once you find prompts that truly boost productivity, don't keep them hidden away. Create a shared prompt library accessible to your entire organization. Make it easy for everyone to find and use these proven prompts, and watch how quickly productivity AND Gen AI excitement improves across the board. AI success isn't as complicated as most "AI Experts" make it sound. Identify, create, save, and share. That's it. #ai #promptengineering #chatgpt #aitoolsforbusiness #aitools #genai #genaiforbusiness

Getting started with Copilot agents is really simple. However, there are still lots of options as what type of agent you choose, and where you choose to build it, will define how complex that experience might be and what capabilities you can choose from. Here is a really basic overview of the different options and how complex they can be: ➡️ SharePoint Agents: Build in SharePoint (or there's a default one set-up for each site). Ground on SharePoint sources. Easily share in Teams group chats. Use with a Microsoft 365 Copilot license, PAYG, or as part of the current promo access available through June 2025. Complexity: 1 ➡️ Declarative Agents in the chat-based agent builder: Build from the "Create an agent" link in Copilot chat. Ground on your business files or the web. Use image generation or code interpreter. Share a link to give other users access and even use the agent in the Microsoft 365 apps. Available to users with Microsoft 365 Copilot licenses, access through PAYG, or, for basic agents, free access through Microsoft 365 Copilot Chat. Complexity: 1 ➡️ Declarative Agents in Copilot Studio: Everything available in the chat-based agent builder plus actions and extra data options. Build by building an agent under the main "Microsoft 365 Copilot" agent in Copilot Studio. This provides a subset of Copilot Studio capabilities for publishing an agent to Microsoft 365 Copilot/Microsoft 365 Copilot Chat. By default, Microsoft 365 Copilot licensed users have access to this builder experience, if you want to give access to others look at PAYG set-up. Complexity: 2 ➡️ Custom Agents in Copilot Studio: Use the full Copilot Studio capabilities including autonomous agents and access to a wide variety of knowledge and integration options. These agents currently do not publish to Microsoft 365 Copilot Chat, though this is an expected future feature. Publish the agent to a wide variety of channels, including Teams and websites. To use these, you need a Copilot Studio license (and message pack) or to set-up PAYG billing for Copilot Studio. Complexity: 5 ➡️ Custom Engine Agents: Build an M365 app backed by a custom AI experience, either driven through Azure AI Foundry or even 3rd-party offerings from outside Microsoft's services. You can publish your agent as a Copilot compliant app to be available in chat, or elsewhere such as Teams. Cost and licensing depending on the options you select, and the investment needed in running the connected services. Complexity: 10 When choosing from these options, Microsoft Learn is your friend in digging into the detail and understanding what options are available. Be sure to check out the help docs to make the right choice (link in the comments). 📲 Do you need help adopting or extending Microsoft 365 Copilot for your small- or medium-sized business? Drop me a DM to find out about the advisory, technical support, and training services I offer to help you pilot Copilot to success. 🖼️ from Microsoft Learn