A successful Microsoft 365 Copilot rollout starts with business problems, not blanket licensing. Follow this five-step framework to identify AI use cases, select the right Microsoft tools, calculate value, and plan for lasting adoption.
Are you considering a broad Microsoft 365 Copilot roll out to most, if not all, of your company? Then read on. You will want all the information in this post so that you make the most informed decision.
Step 1: Reevaluate Microsoft 365 Copilot
First, let me be clear that the current, evolved version of Copilot is a very capable and, to me and many others, indispensable tool in daily work activities. I will also be clear that earlier versions of Copilot, as recent as 8-10 months ago, were experiencing growing pains. The AI responses were not on par with OpenAI and Anthropic tools. The promises of easy spreadsheet and presentation creation were not aligned with the actual experience. But, as I’ve written before, if you dismissed Copilot in the past, you owe yourself a re-evaluation. Copilot is more than just a chat interface. The premium licensed version now:
Let’s users choose from multiple models and versions within the same model. Make sure to turn on Anthropic models in the Microsoft 365 Admin Center. Note, you GCC tenants out there will have the OpenAI models, but the Anthropic Claude models are not yet available.
The integration with your work data has improved and is now called “WorkIQ”. Copilot is much more aware of what and who you work with and understands who you are and what is going on in your work environment. This native integration has always been a differentiator for Copilot, and it has now improved even more, connecting you to the right content without you asking. WorkIQ also allows for a much better search experience in Copilot. It is now the default tool to find disparate content.
There are more out of box agents created by Microsoft to make your job easier. Combined with the improved Researcher and Analyst agents, these agents help you more easily create content, find experts in your company, plan your day and manage your Planner tasks. These integrations allow you to get the full value out of your Microsoft productivity suite.
Personalization options have also improved. Copilot’s memory settings allow it to better remember your preferences and its personalization settings help guide it towards your style of writing so that generated content requires less editing by you…but always a review!
And then we have Copilot Cowork. Ah, Cowork. A remarkable set of integrated capabilities that turn Copilot into a fully functioning team as opposed to just an assistant. Cowork does require a consumption pricing model so there are additional cost and value considerations to be had.
Let’s finish off with Microsoft’s enterprise data protection promise. While this hasn’t changed, it is an important differentiator from other AI tools on the market, especially free ones.
Step 2: Understand the Microsoft AI Ecosystem
Great, we have now established that Copilot is an evolved and functional tool. However, it does not stand alone in the Microsoft AI ecosystem. There are three other tools to consider:
Copilot Chat, also known as the “Basic” version of Copilot, is part of your existing Microsoft license and provides quite a few core AI capabilities. While it is missing WorkIQ integration, it is a powerful multi-model chat tool. It allows you to build basic web-knowledge based agents. It can reason over uploaded content. It includes enterprise data protection and, for those organizations with less than 2,000 seats, integrates with the Office productivity suite. Very capable, and “free”.
Copilot Studio allows for the creation of custom AI agents. Agents can serve many purposes ranging from simple knowledge retrieval (chat with your content) to both prompted and autonomous task execution. Agents can integrate with your other systems and are great at executing focused activities. Where the Copilot tools assist individuals, agents assist entire process workflows.
Foundry rounds out the ecosystem. While Copilot Studio focuses on agents, Foundry and the Azure AI tools allow organizations to build complex solutions using professional code. Developers can choose from thousands of models and dozens of capabilities.
Why consider these tools when all you want to do is roll out Copilot? Because these tools don’t just provide you with new capabilities, but they also have differing cost models, some using a fixed fee (license) model and others a usage/consumption model. The diagram below highlights the differences. Note that Copilot straddles the payment models with the Copilot Cowork interactions requiring a consumption payment model.
| Solution | Best suited for | Data context | Typical cost model |
|---|---|---|---|
| Microsoft 365 Copilot Chat | General AI assistance and lightweight agents | Web, uploaded files, and permitted features | Included for eligible users |
| Microsoft 365 Copilot | Individual productivity in Microsoft 365 | Work IQ and Microsoft Graph context | Per-user license |
| Copilot Studio | Custom agents and process workflows | Connected enterprise systems | Capacity or usage based |
| Microsoft Foundry | Custom, developer-built AI applications | Custom data, APIs, models, and services | Azure consumption |
| Power Automate | Rules-based workflow automation | Connected applications and systems | License or consumption based |
| Copilot Cowork | Multi-step delegated work | Microsoft 365 and Work IQ | Copilot license plus usage |
Step 3: Identify High-Value AI Use Cases
Now that we understand the technical landscape, we need to switch our focus to the more important consideration, namely, what is AI going to help us solve for? What business processes are we going to optimize? What work activities are going to benefit from integrating AI capabilities? What customer experiences are going to be enhanced? What data insights are going to be made available for the first time? The answers to these questions will provide an organization with an entire inventory of possible improvements. Improvements that will provide real value to the organization through such measures as increased efficiency (and the value-add repurposing of that freed up time), improved customer satisfaction (and the additional sales that will come of it) and better business decisions through improved data access.
How do you find these opportunities, or use cases? The simplest approach is to simply ask your employees, teams and leaders what they would like to see improved. Don’t worry about whether their ideas require AI. All identified use cases are valuable. And don’t, whatever you do, ask them, “how are you going to use AI in your job?” Why? Because the probably have no idea nor do they have the time to figure it out. Just get their ideas and log them.
If you don’t have the time to interview a lot of people, then have technology help you out. One option is relatively old school. Create and publish a survey using Microsoft Forms. Ask simple questions about what folks would like to see improved and don’t require long multi-sentence answers. A simple statement is good enough for now. Want to “up your game” and move beyond surveys? Open the new Copilot Studio (copilotstudio.preview.microsoft.com) and use the prompt to create a “Work Improvement Gatherer” agent. The agent can be instructed to ask employees what they would like to see improved in their work. Make the agent use a friendly tone and limit its inquisitiveness so that those being interviewed don’t get bored or turned off by the process. Have the agent then populate a SharePoint list with the ideas and notify you that a new submission is available. There you go. Use cases!
Step 4: Match Each Use Case to the Right Solution
Now comes the fun part. For each of your use cases, you need to figure out which one of our technology options from Step 2 above provides the best solution. To figure that out, you can either give us a call here at eGroup and we can review the use cases with you and help you evaluate and prioritize them, or you can enlist the help of the very AI that we are looking to deploy. This latter approach will have you upload your use cases into Copilot and prompt it to read through them and provide its ranking of feasibility, complexity, effort and tech solution. Once Copilot has this information, it can stack rank the opportunities as well. Now you have turned your use case list into an actionable roadmap.
Remember, each technology has a different pricing model. Determine the costs to implement (a bit more on that in Step 5), and the benefits and you have all the pieces you need for your calculation. You can now calculate the value and ROI for each use case. Your calculations will result in the following insights (let me know if your results differ):
Everyone will benefit from the free Copilot “basic”.
Some will benefit from the Copilot “premium” license…but not everyone.
Others will need custom agents due to the more complex nature of their work and integration with other systems.
A subset of use cases will require advanced AI solutions.
And others can discard the need for AI altogether and simply use a Power Automate Flow. Do not dismiss these use cases since non-AI improvements can be just as valuable.
This mix and match approach will result in a more cost-effective way to deploy AI than to allocate a paid premium Copilot license to all. Trust me. We’ve done the math. Reach out if you want to walk through the details. The main lesson here is that you should treat paid Copilot licenses as options for solving use cases, not as a blanket requirement for all.
Step 5: Plan for Adoption and Organizational Change
Recently I have been a part of many discussions that focus on sustained AI value and adoption. Organizations want to make sure that their AI solutions are used to their fullest potential. Well, if you have followed the steps above, you have already mapped AI to the problem and there should be no guess work as to how AI can be used.
Many, however, make one significant mistake when deploying AI. They underestimate the effort needed to deploy the solutions. AI can be easy to turn on. Sometimes it just takes IT to deploy a license. Even agents, as evidenced above, can be easy to design and launch. Do not let the ease of a technical deployment obscure the effort needed for the human change. Here we finally come to the key realization – each of these solutions requires one or more users to change their behavior to adopt them. Individuals and teams must change their business process, often changing what they have been doing their entire work lives. This does not come easy. It involves training, monitoring and reinforcement. The overall process will likely slow down as new behaviors are learned before they speed up. The old saying of “go slow to go fast” is in full effect. The more complex the process, the more time it will take. Even basic behaviors can take 50 hours of effort to learn. These hours must be factored into the deployment, even for Copilot. A simple training session or certification exam is no replacement for a thorough implementation plan. Please give your workforce the time and space needed to relearn. Do so and you will recognize the value that you had planned for.
Here at eGroup, we focus as much on the human change element as we do on the technical implementation. Our Organizational Change Management (OCM) is a vital part of our customers’ AI solution deployments. So much so that, from Day 1, our OCM team led all our Copilot projects…and still do. Want to talk more about how to make sure that your AI projects achieve their goals. Reach out and we’ll review your use cases and apply the 5-step recipe to AI success.
Build Your AI Roadmap
Identify high-value AI opportunities, match them to the right Microsoft technologies, and create a practical roadmap for implementation, governance, and adoption.
