Microsoft Copilot enterprise guide: all 12 versions explained with workflows, ROI strategies, and rollout playbook
A solo deep-dive by Nate B Jones on how enterprises can move beyond basic email use to unlock the full value of Microsoft Copilot.
Summary
Nate B Jones opens with a DM from the CTO of a 6,000-person company frustrated that a six-figure Copilot investment is being used almost exclusively for writing emails. The episode is a comprehensive guide to all 12 distinct Copilot products — which are confusingly named and priced anywhere from free to $35,000 per year — and explains which version is appropriate for which use case. Nate walks through specific workflows across email, document creation, Excel data analysis, meeting productivity, cross-team collaboration, and advanced techniques like prompt chaining and bot-building. He uses Vodafone's rollout to 68,000 employees as a case study for how phased deployment, executive sponsorship, and cultural change management are what separate companies getting 20x ROI from those stuck in first gear.
Key Takeaways
FULL TRANSCRIPT
The Problem: Six Figures Spent, Used Only for Email
Nate B Jones: Last week, I had the CTO of a 6,000-person company send me this DM: "We're paying six figures for Microsoft Copilot licenses and almost everyone on my team uses it for writing emails. That's it. What the hell are we missing?"
So this is a deep dive on Microsoft Copilot. This is a complete guide to how to roll it out and make the most of it. I have been asked for this particular guide for months, and I have taken my time because of how complex it is. Strap in, stay tuned, get out your notebooks. This is going to be a long and complicated video, and it is totally worth it — because Microsoft is in almost every enterprise out there. The chances are, if you can't use ChatGPT, if you can't use Claude, you've got to use Copilot. So what do you do? I get that question a ton. This is the answer.
I want you to walk away with a clear set of use cases you can adopt, a clear sense of the organizational changes you need to go through to enable Copilot to get beyond the email use case, and a clear sense of which of the 12 different Copilot products to pick. I'm not kidding — there are 12, and we're going to go into each of them.
Here's where we start. Something like 90% of the Fortune 500 has Copilot. 90%. That's why this is such an important video. That's why I've taken my time on it — it's got to be correct. Microsoft's own data shows that most people aren't using the full capability set of Copilot. Now, if you're not using the full capability set of Microsoft Word — which I know I don't — that's okay. It's Microsoft Word. It's there to help you type stuff. But if you're not using the full capability set of Copilot, it's a big deal, because Copilot is literally intelligence on tap. And so if your workforce isn't using it, if your teams don't feel good about using it for things that are more than email, you're the ones missing out. Companies are literally paying for a Ferrari and driving it to the grocery store in first gear. That is what is going on with Copilot.
I've spent months looking into case studies. I've spent months talking to actual folks who are working on Copilot in these different enterprises that adopt it. I've interviewed users who are individual contributors — users in engineering departments, folks in product management, sales. I wanted to get a sense of how Copilot is actually being used and then how it could be used. This is the fruit of all of that.
You're going to know exactly which Copilot you need. You'll be on track to save several hours a week if you go through this video and actually apply it. These are going to be real workflows. I'm going to give you specific prompts, and I'm going to show you some case studies — including how Vodafone rolled this out to something like 68,000 people successfully.
The 12 Flavors of Copilot Decoded
With all of that for intro, let's jump into part one. I want to talk about the 12 different flavors of Copilot, because it is completely confusing and I get why people don't understand it. Frankly, Microsoft did something absolutely insane. They took the name Copilot and slapped it onto everything.
I'm going to read you the real actual Copilot list. Microsoft Copilot — free, comes built into Windows, designed for consumers. Copilot Pro — a $20-a-month power user version. That does not mean it's actually for professionals in a work setting. Copilot for Microsoft 365 — $30 a month, looks like it's tied into companies. Microsoft 365 Copilot Chat — somehow different, and free for business accounts. Both of these are tied to the Microsoft 365 business product line. GitHub Copilot — for developers, tied to the GitHub product line. Everybody got the Copilot branding. GitHub Copilot for Business — different from regular GitHub, about double the price. Security Copilot — this one costs a lot of money. I don't understand how you can name the product the same thing and have it cost so much more. You can have something that is free in Windows, and the same name — Copilot — can also cost $35,000 a year minimum. Copilot in Dynamics 365 for Sales — built into your CRM. Copilot in Power Apps — for building apps. Windows Copilot — different from Microsoft Copilot somehow. Copilot Studio — for building your own Copilots, so we can have more. And there are variants for customer service and field service.
Do you get the idea why it's taken so long to make this video? People say, "How do I use Copilot?" and in my head I'm thinking, "Which of the 12 are you using?"
There are companies that have bought Copilot Pro thinking it was the business version, because it makes sense — it says "Pro." But it's not the business version. There are companies that have bought Microsoft 365 but don't have the right base licenses, so Copilot doesn't work properly. Dev teams have bought Copilot for Business when they just needed individual licenses. The product surface confusion with Copilot is as bad as the confusion with ChatGPT and picking model names. They're both wildly harming the capability of the product and its ability to deliver value by being unable to name things correctly.
Everybody's confused, but smart companies can still get massive advantage because fundamentally it is intelligence on tap. I tease them for the naming — and the naming is confusing — but we're going to go through each of those products and talk about what they do so you can figure out what you need. The difference is thousands of dollars per employee per year, so let's get this right.
Product One: Microsoft Copilot Free
Number one — Microsoft Copilot Free. It's built into Windows and designed for consumers. If you're in business, you're probably not using this. As an example, you can tell it to turn on focus mode, play jazz music, summarize a web page in Edge, help write a birthday message for mom, or explain a screenshot. This is the variant of Copilot that ends up in ads because it's the one average consumers have on their Windows machines at home. It doesn't work with files. You should not put confidential information into it. It does not have prioritized GPU access, so it gets slow during peak times. If you have Windows 11, it's already there.
Product Two: Copilot Pro
Copilot Pro is for individuals who are not professionals in a work setting — even though it's named Pro. It's roughly on par with the ChatGPT Plus offering, at about $20 a month. You get priority access to GPT-4 Turbo, which is an okay model. You get 100 image generations per day versus 15 on free. It works inside Office.com web apps, which is a real help. And you get early access to new features.
As examples of what you can do in Pro: you can generate social media images, get faster and better responses across long documents as a writer, and if you're a student, you get priority access during finals week when everyone is using it. If you're working on final drafts for a document and it saves you a few hours a week because it can handle the load, that adds up to real money. It easily pays for itself.
Product Three: Copilot for Microsoft 365
Copilot for Microsoft 365 is approximately $30 a month and is aimed at the business and enterprise layer. The key difference is this one will see your work data — it's tied into the Microsoft 365 ecosystem. That is the point. This is often the one people are talking about when they talk about Copilot at work.
Specific capabilities by app for this flavor of Copilot:
In Outlook, you can summarize emails from a specific person from last week about a specific project. You can get a draft reply accepting a meeting and proposing a different day — in five seconds.
In Word, you can say "Create a project proposal for X topic based on this template" and attach a DOCX file. You can say "Rewrite this section to be more executive-friendly." You can say "Add a risks and mitigation section using data from risks.xlsx."
In Excel, you can ask "What are the top three trends from the sales data?" and it will tell you things like how much something grew. You can say "Create a pivot table showing sales by region and month" and get an instant pivot table. As someone who has suffered through Excel, the fact that people can now create instant pivot tables by just typing a sentence — we live in a blessed world. I hated creating pivot tables when I was in marketing.
In PowerPoint, you can say "Create a five-slide deck from report.docx." It won't be gorgeously formatted, but it will have some design, some slide transitions, and you can tell it to add speaker notes based on the content.
In Microsoft Teams, you can type or speak during a live meeting and ask "What did we just decide about the budget?" You can ask Copilot to generate meeting minutes with action items, or summarize a channel's discussion.
I am just going through Copilot for Microsoft 365 — which is not the fanciest Copilot — and I'm already giving you so much more than just email. We're just barely getting started.
Product Four: GitHub Copilot
On the developer side — GitHub Copilot, approximately $10 a month for individuals and $19 a month for business. The idea is very similar to Cursor or Windsurf. It writes code alongside you. Depending on who you talk to, people say it makes them 55% faster on repetitive tasks. I don't care if it's 55 or 30 or 75 — the point is it's a repetitive task. If it's helping you autocomplete your code faster, it gives you that speed-up. You bring the design, the knowledge of the codebase, the ability to write the code — and Copilot acts like a smart autosuggest that runs really fast.
For folks who write code by hand, that's great. For folks using more of the agentic capabilities in tools like Cursor, this isn't really going to work for them. The developer ecosystem is fracturing quickly. You still have developers who write code by hand and find it better to write with an accelerated keyboard, getting Copilot-generated autosuggestions. But you also have people moving away from writing code at all — depending on AI through vibe coding, or using agents to run pull requests, or writing some code themselves and using agents for the rest.
I'm not going to have the "which development stack is future-aligned" conversation in this video, but we will have it soon. For now, what you need to know about GitHub Copilot is that it helps you write code you already know how to write — faster.
Product Five: Security Copilot
Security Copilot costs roughly $4 an hour for compute, which works out to approximately $35,000 a year minimum. Security teams with 10 or more analysts might use this. What it does is help you analyze suspicious user behavior across different logs, write incident reports, and explain malware code in plain English.
If you can explain malware code in plain English and write incident reports, and you're operating at true enterprise scale, this pays for itself immediately. You can say "Investigate if user John Doe's account is compromised" and it will actually run an investigation. Or you can ask it to write up and describe an incident that comes across your desk and accelerate the time to resolution. When you think about the millions of dollars lost during security incidents, the ROI is off the charts even at $35,000.
This is one of the things that makes large language models and AI hard to talk about in a single video. I have to talk both about how individuals are using this in their living room with Windows 11, and also how a security team at a Fortune 100 is using a $35,000 edition with the same name — one that enables them to get back online two or three hours faster, saving the company $10 million. Those are the same named thing.
Basic Workflows: Where Most People Start
All right, we got through a bunch of the different Copilots. I want to talk about some specific workflows that can help you deliver a win.
Yes, we really are going to start with email. If you've never done Copilot before, this is great. If you've done Copilot and you can already do email, we'll breeze through it fast enough that you can get to the other cool stuff.
Open Outlook with Copilot, click on the Copilot icon, and ask it to summarize your unread emails. Everyone has unread emails. You'll get answers like: "Ellis had a budget revision request. Carol wants her client meeting moved to 2 p.m." Now you know what needs your attention. That's value right there.
If you have a long complaint email from a client, just say "Draft a professional response acknowledging concerns, offer an X% discount, and propose a new timeline." You can do that kind of business engagement with your email inbox. And it does save time. Even in the case of the CTO who started this whole conversation — his team was saving multiple hours a week just from email. It wasn't nothing.
Document Creation
Let's move on to document creation. Take a task like a blog post. Let's say you're a marketer. You need to write a blog post about sustainable packaging. You can say "Give me five trends in sustainable packaging, a sentence each." Then: "Create a detailed outline for a thousand-word article." Then: "Write the first paragraph — help me see your hook." Tweak it. "Now write section one. Write section two." You're going to be done with that blog post in 15 minutes versus two hours. That's a huge savings for SEO-type document creation.
I will hasten to add: Copilot is extremely good at document creation, which may lead you to think it is extremely good at strategy. Those are different things. If you want a good strategic partner, that is not the workflow to go through. You still need to bring your brain if you are doing strategic thinking. What Copilot is doing is accelerating your ability to execute once you have clarity. I'm going to underline that like three times.
Data Analysis in Excel
Let's look at the data analysis piece. Let's say, like me, you don't love Excel formulas. You can be as high-level as you want. You can say "Hey Copilot, analyze this data. Tell me what's interesting." And you're in a spreadsheet. It will say things like: "Sales increased 23% year-over-year. July was an anomaly with a 45% spike. The western region is underperforming."
You might follow up: "Hey, why is July so high?" Copilot might say "Column J shows there's a summer promotion running for July entries." Then you say "Great — create a chart showing monthly trends with the July spike highlighted," and you get a professional chart.
That entire workflow used to take me an afternoon when I was working through marketing data anomalies. It is now doable in a couple of minutes. This is no-code. It's not that hard. But I hear actual organizations using Copilot with Excel a lot less than I hear them using it with Word and Outlook. Excel seems to be underrepresented. Don't sleep on Copilot in Excel if you have Copilot enabled.
Meeting Productivity
Let's get into the meeting side of things. All of our meeting burden increased after 2020 — something like two to three times the baseline number of meetings, depending on how you measure it, and that hasn't really gone away. Meeting productivity matters more than ever.
For pre-meeting prep, you can type into Copilot: "I have a client meeting about project X in an hour. Give me an agenda — status update, challenges, next steps." It will spit out an agenda with time limits. Then you can say "What questions should I ask to better understand the concerns the client is likely to express?" and it will give you questions you can actually dig into.
You can get Copilot in Teams in a live meeting. If you joined late, you can ask "What have we discussed so far?" and Copilot will give you an update without you interrupting anyone. You've all been there — someone joins the Zoom or Teams meeting six or eight minutes late, they're a highly paid important person, so they get the update and everybody has to sit there while it happens. You can just ask Copilot for it instead.
Post-meeting, you can say "Create meeting minutes — give me attendees, key discussions, decisions made, and action items with owners," and you'll get formatted minutes very fast that you can send to everybody. Meeting minutes by themselves are not uniquely a Copilot thing — you can get them from Granola, from Otter, from OpenAI. What Microsoft has always sold is distribution and bundling. It's all in one place.
Research Assistant
You can use Copilot beyond just operating on documentation you already have. Web research is a great example. You can say "Find current pricing for the top five CRM software companies and create a comparison table." Copilot searches, returns what Salesforce costs, what HubSpot costs, and a few others, and creates a nicely formatted comparison table. If you are in sales, this is an example of how people are actually collapsing the buying funnel — people are making high-consideration software purchases by talking to Copilot or ChatGPT.
If you're looking internally rather than on the web — and remember, on certain flavors of Copilot, the 365 one for example, you can search all of your documents — you can say "Find all mentions of Q4 budget and summarize findings." Microsoft 365 has access to your documents, and it will dig them up and say "I found three documents: budget draft.xls, budget draft v2.xls" — we all have our weird names for our docs — and it will tell you what's inside. It mentions $2 million allocated. It mentions a 15% increase was approved.
This is a feature that works because of how powerful Microsoft has been for decades. There are other companies like Perplexity working on internal document search for your PC or Mac, but they have a hard road to hoe — they have to build a lot more, and you have to remember to use them, and you're still only an individual. The advantage of Microsoft really shines through here because they have your data, you built the data in their tool, they know where the data is, and they can just put Copilot over the top. This is why a lot of people are using Copilot — and hence this video.
Advanced Individual Workflows
Let's go from basic workflows to advanced individual workflows. We are past "Hey, can you draft this response to Sheila from sales?" Let's do something more fun.
Batch processing in email. You can say "Please group this list of 50 emails by topic." You'll get: Project X — 12. Budget questions — 5. Meeting requests — 8. It really will group them, which makes it easier to batch process your email. My brain works by topic, so this is very helpful. You can create prompts for common emails, invoke those prompts quickly, and knock out those batches. You can even have an end-of-day prompt: "Which emails from today need follow-up tomorrow?" and get a summary for the day.
Advanced blog post workflow. We talked before about drafting. Let's go further. You can find five recent stats about remote work productivity. You can outline and create a blog structure with an intro, three main points, and a conclusion. You can draft it, polish it, say "Suggest five SEO-friendly titles for the post," and then "Create three Twitter posts promoting this article." None of this requires code. But what it does is challenge someone who's only used AI for "draft this response" or "write this blog post" to think of AI as intelligence that underlies their entire workflow. What could you do with your time if intelligence was pushing on all of the different pieces of your workflow?
The email example I shared — what makes it advanced is that you're thinking about your inbox differently. You're thinking about it in terms of topics and asking the intelligence Copilot provides to attack it differently. Same with the blog post — you're asking Copilot to help you through all of the stages of drafting and sharing, not just the writing. You'll see that end-to-end theme pull through with most of these advanced workflows.
Report writing. Summarize all of your project X documents from this month. That alone is worth hours, if you have clean data. Create a monthly report template with standard sections. Creating a template used to take me a long time because I had to think about and look up best practices and agonize over every section. Not anymore. Because of how Copilot is trained through reinforcement learning on business data, it is going to give you something that reflects best practice. You can add to it, but you're getting a pretty solid template. Then fill in the sections with the summarized information, get a one-page executive summary, and suggest three charts. If you have good data and a clear opinion on what you want to call out, you can write an exact report like that very quickly. Previously that would have taken a day or two.
I will emphasize again: it does not mean you are letting Copilot do the strategic thinking. Microsoft has actually branded it well as a co-pilot, because it means you are still the pilot. You are still the driver.
End-to-end sales analysis. We're not just going to say "Tell me what's interesting." We're going to say "Raw data — 10,000 rows of sales transactions. What's the overall trend?" It will say "Sales is growing 3.2% month over month." Then: "Which products are driving growth?" Product A is up 45%. Product B is down 12%. "Are there any concerning patterns?" Returns are increasing for product B, especially in the southern region — license expiration issues, churn issues, whatever it is. Then: "Create a pivot table broken out by region. And think with me about actions we can take."
This is where I want to call out that this is an advanced workflow partly because of the judgment you have to exercise. If you're asking for actions to consider, Copilot is going to give you best-practice actions — it will not necessarily give you creative actions. You need to think about the creativity you want to bring, and you need to make sure any recommendations Copilot generates are in line with your overall strategy. Again, you bring the brains to the exercise.
Team Productivity and Cross-Team Workflows
We've talked a lot about individual productivity. But one of the things the CTO wrote me about was wanting to see team productivity change. It's a vexed issue, especially if you're in the C-suite. How do you upshift teams? How do you move teams from buckets of individuals who say they save three or four hours a week on surveys — and you don't know where it went — to actually getting teams to move forward?
Build a shared prompt library. A lot of people are doing this. Marketing team examples: "Draft a customer win announcement email." "Write the webinar invitation." "Generate five LinkedIn post ideas for a white paper." "Create an email nurture sequence for new leads." "Analyze campaign performance and suggest improvements." This inside-team improvement doesn't just work for marketers. For sales: "Draft a follow-up email for the demo call" — and sales can customize that to particular product lines. "Create a proposal executive summary from our meeting notes to send to a key stakeholder." "Write a contract amendment for a scope change" — legal can weigh in on the prompt and give you the right framing so you get good draft language they can later approve.
You can create a Teams channel called "Copilot Prompts," pin successful prompts, tag them by use case like #email or #content, and do a monthly brown bag. That's a very simple way to start socializing prompts across the business.
Sales to implementation handoff. Sales can create a deal summary using Copilot and say "Extract all technical requirements from this contract so our engineering team can understand it." Copilot will go through it, extract the technical requirements — I'd recommend reading it again just to make sure — and engineering will get a clean requirements list. Engineering can then say "Create an implementation timeline based on these requirements." Engineering in most cases will look at that timeline, swear at it, go to lunch, come back, and then write the actual one. But even if you're upset with a Copilot estimate, it still moves the ball forward because it gets you off the blank page.
In most agile estimation meetings with developers, a lot of what we talk about is that estimating software production is inherently hard, and starting with an estimate somewhere helps us move the ball and shape our understanding. It helps the conversation move in the direction we want because we're no longer focused on getting something onto the page — we have an estimate now, and we can argue about it. The key value Copilot brings here is that it helps translate hard contract language into technical requirements engineers can work with. That by itself can save you days of meetings. I have seen complicated enterprise contracts founder because the engineering team was not brought in early enough. Copilot can help. That's an example of something that doesn't just save time — it can save an entire deal from churning.
Marketing to product. Marketing can say "Analyze the customer feedback we're getting from social media this month." You're going to need to get social media data into an Excel spreadsheet if you want a clean data set, or trust that your signal is loud enough on social channels that you'll get responsible feedback. If you just trust Copilot to go out and look at your social accounts, it tends to be recency-biased and tends to cut off its investigation after a few tweets or a few Reddit threads. So marketing needs to be responsible, understand how the tool actually works — it's a little token-lazy — and pull in all the data ahead of time.
But let's say you've done the good stuff and got the data in place in Excel. You'll get themes: UI confusion — X number of mentions. Feature X requests. Feature Y frustration. Product can then take that analysis from marketing, maybe throw a chart in — Copilot can help with that — and say "Help me prioritize these feedback themes. How do they align to our existing roadmap?" It can invoke roadmap.docx, put them together, and say "This could slot in here. This could slot in here."
That creative streamlining is something PMs have historically been frustrated by because it means revisiting a roadmap and changing it all the time. This makes some of that pain easier. It means marketing has more of a voice with product. It is one of the advanced workflows that ultimately pays for itself not just in time and alignment and reduced frustration, but in the right feature getting built for customers more often. Is it perfect? No — PMs are still going to argue about the roadmap. But it gives you a sense of what Copilot can do beyond just "help me with my email."
The Vodafone Rollout: Enterprise Case Study
I want to go into the Vodafone rollout, which I promised at the beginning. They have something like 68,000 to 70,000 employees across a couple of dozen countries. Across any team that size, you have different technical skill levels, multiple languages, and various job functions. So what did they do?
They started with a small pilot — all of this is publicly documented. Only a few hundred users in the UK, a deliberately mixed group: sales, support, IT, management. They had a six-week pilot with regular check-ins. They measured everything — the time, the quality, the satisfaction. They wanted to understand if this was actually helpful.
It was the pilot with a small number of users that sold leadership on the value of Copilot at scale. What they saw was it saved something like three hours a week per person. Most people wanted to keep using it. Most people said their work quality improved. Customer service in particular said response time was cut dramatically — something like 40%. Sales said proposals went out days faster. Ticket resolution times improved.
When you start to see in controlled environments that are small like that — real gains and people wanting to keep using it — it's an indicator that this is a generally available technology that can uplevel a large range of enterprise job functions if you let it.
What I love about the Vodafone example is they took the time to make sure in that controlled environment that everyone had access to the tool and the training, that everyone understood the expectations, and that everyone knew this was special and they would be focusing on it to make sure they adopted it successfully. I bet this would not have been as successful if it was just rolled out casually across everybody — if they didn't have training, if they didn't measure everything, if they didn't do weekly check-ins. The way you roll out the intelligence helps you see if the intelligence works or not, and helps people know how to use it. Otherwise you're just going to have people summarizing emails with Copilot and saying "Well, I save about an hour a week." And then CTOs email me and say "What are we doing wrong with Copilot?" And the answer is probably a culture change piece.
Phased Enterprise Rollout Framework
Phase one — Foundation, months one to two. I'm assuming here that this is a from-scratch rollout. It's not often actually the case, since 90% of the Fortune 500 or so have Copilot somewhere. But let's assume for clarity this is a clean rollout.
In the first couple of months, the CEO announces this is coming. They've identified a couple of dozen Copilot champions across different divisions — I'm assuming this is a big company. They've built an internal support site where you can get your questions answered. They've developed role-specific training for different departments, and those champions are owning that training.
You can see how you're starting to seed the change. The CEO is setting the pace, saying this is good and positive, and allaying fears about Copilot stealing your job. Anecdotally, talking to individual contributors, you have to talk about the job thing if you're going to talk about the AI thing. People have been so primed by the news to be afraid of AI stealing their job that they won't listen to anything else until you address it. And that must come from the leadership team — preferably from the CEO.
By the way, there's not a ton of evidence that AI is actually stealing jobs in aggregate right now. Ethan Mollick is one of the best-known AI academics out there. He studied AI extensively and noted as recently as early July 2025 that there was just not much evidence that AI is actually responsible for significant job attrition. Isolated instances, or expectations changing for job roles — 100%. Changing job descriptions, scaling up hiring for AI-specific roles — definitely seeing that. But we're not seeing large-scale attrition from AI. And part of why, I think, is the very reason I've had to make this video. If AI was as good as it needs to be to take our jobs, I wouldn't have to make this video — because you could ask Copilot and Copilot would tell you how to roll itself out. But it doesn't. And so we need to do this culture change.
Phase two — Scaled deployment. Let's say you have a six-figure employee count company. You're trying to get 10,000 users a week on. That's the pace you can do with Teams channels for Q&A, success story sharing, and daily Copilot tip emails. You have some operational support for these things that caps out at around 10,000 users, and you want to see how it's working.
By the time you get to full adoption — maybe 100,000, maybe 200,000, maybe like Vodafone's 68,000 users enabled — you'll have department competitions for best use cases, integrations with existing workflows, and continuous measurement and optimization to see if things are working.
If you take your time and roll it out like that, if you make it a serious organizational change initiative, it gives you space to celebrate and talk about your success factors. You can talk about the fact that the CEO and CTO personally championed this — that skin in the game mattered. You can talk about how the focus of your rollout was on helping each employee do their job better, not just saying "use AI." You can celebrate the champions in your departments. You can have weekly dashboards, bonuses tied to impact, success stories and tips shared in dedicated Teams channels, in brown bags, at all hands.
The larger lessons learned are not too hard. You want a phased rollout — if you're an enterprise, don't do a big bang. You want champions who are individual contributors on the front lines. They translate the tech to business value, and your job as a leader is to celebrate them. You need to address fears upfront — be clear that this helps you and doesn't replace you. You need to measure what works and what doesn't, be really honest about the places where it doesn't work so you can fix it, and celebrate the wins.
None of this is new. If you've been in the culture change business and leadership, this is what we've been saying for a long time. It's not different with AI. In fact, it's more important to do this with AI because AI feels more personal. We as a human species have never had to do shared intelligence at work. We are at the Henry Ford generation for figuring out how to work with machine intelligence at work. We need good leadership to make that successful. And like it or not, even if you love ChatGPT — and I talk about ChatGPT a lot — Copilot is where a lot of that is going to be worked out in the workplace, because of Microsoft's distribution advantage.
Advanced Features Most People Miss
Let's call out a few of the advanced features people miss as we transition toward the end of this lengthy video.
Copilot Vision is a really interesting one that people rarely use beyond email. You can take a screenshot of an error message and say "Can you explain this error and how to fix it?" Copilot will read your screen and explain it in plain English. You can take a screenshot of a PDF table and say "Can you convert this to Excel?" — unless the table is really small with tiny print, that will often work, and you'll get properly formatted data ready to paste. You can take a screenshot of complicated software or an onboarding screen and say "I don't know how to export data — how do I do it?" I have absolutely done this with complicated, poorly documented software and gotten so much help. You can get button names, menu paths — it's really, really helpful.
Prompt chaining. This is how you do advanced drafting with Copilot. As an example, think about five prompts that together create a business plan. Number one: outline the business plan for an AI consulting firm. Number two: expand the market analysis section in more depth. Number three: add financial projections for the next three years. Number four: create an executive summary based on the sections. Number five: suggest five potential risks.
Individually, these are prompts you might think of for drafting. But when you take them together, you're going on a chain of thought with the AI. You'll write it slower than I said it — and the reason is that you need to bring your own thinking to each piece in prompt chaining so that you're actually partnering with AI. It almost looks like a walk through the conceptual space you're exploring together with machine intelligence. Each step in the prompt chain helps you move your own thinking forward.
Cross-application report generation. In Excel: analyze the Q4 sales data. Copy those insights to Word. Then tell Word Copilot: "Create a professional report using these insights. Add an introduction." Then move to PowerPoint Copilot: "Create a five-slide presentation from this Word report." Then go to Outlook: "Draft an email to leadership with the report attached." So fast. You have to bring your brain — you have to know what you want to say — but it is possible to jump from doc to doc with no code needed. Most people aren't doing that jump.
Building your own bots. An HR onboarding bot built by HR without code can save HR several hours a week by just answering: "What's the vacation policy?" "I need to add my spouse to insurance." "Walk me through setting up my phone." If you're at enterprise scale, you get lots of those queries every single day. An IT help desk bot can run password reset workflows, software installation guides, and escalation to a human when needed. It might deflect half or 60% of your tickets. It can be very helpful.
Industry Case Studies
DWF law firm — Microsoft mentioned them as a case study. They used Copilot to reduce the time it takes to review contracts. Even though legal has a very high bar on accuracy and cannot mess anything up, you can still use something like Copilot to get an early read and reduce the time. The time taken went from something like seven days to seven hours, and they got more consistent quality because they uploaded the contract and asked the right prompts: "Identify obligations. Identify deadlines. Flag unusual terms compared to our standard. Create a summary in plain English." It's all about the kinds of prompting you bring. You can see how they're prompting in ways that give the AI room to succeed.
Healthcare. You can use Copilot for patient discharge summaries. Doctors might spend 30 minutes writing discharges. Copilot can pull from patient records, create the discharge summary including medications and follow-up instructions, and be done with it. It can result in a dramatic drop in the time doctors spend on paperwork and more time with patients.
Finance. You can feed market data into Excel and get a second perspective on market anomalies.
Retail. You can upload a product image and say "Write this product description." That one gets done a lot — not just with Copilot.
Getting Started and the ROI Reality
If you want to do this, it is not that hard. You can get Copilot in, try basic prompts, get into your email, get into Word and Excel in the first couple of days. If you want to get into Copilot Vision, if you want to get into chain prompting and some of the advanced stuff I talked about, just try it. This is one of those things I keep emphasizing with AI — just try it. You will learn if you are willing to try.
If you're working with a team of people and trying to move them forward, let your team try it. Celebrate the wins your team is getting with Copilot. Encourage your team by showing them that you value their time saved — that you will not just pile more work on, that you will not tell them they must not be that good at their job because they did it that fast. Actually celebrate the win.
Here's the reality. I'm going to go right back to the beginning. The CTO wrote me and said they're only using email — 6,000-person company. These companies are using 10% of a six-figure investment in some cases, whereas the ones that are using it fully are getting dramatically more return on investment. The swing I see here is insane.
I talked earlier about the Security Copilot that costs around $35,000. I talked about the ROI you get if you can reduce the time of an incident because you're actually better at triaging. That by itself delivers 20x ROI. Another way to get to that 20x ROI: get these cross-team workflows going. Get marketing talking to product with Copilot. Get sales working with engineering more successfully. You're going to hit that ROI.
Advanced Copilot usage like I'm describing is going to be table stakes relatively quickly. If you don't adopt it, you're going to be behind. That's another reason I'm choosing to make this video now. I could have sat on this and said "I need more information — it's not perfect." But no information set ever is. I'm making this video now because you need to adopt Copilot quickly if you have it in your business, and you need to adopt it successfully. Waiting won't help you.
Start with a workflow I outline here in the video. Measure the time you saved this week. Share the video with your team. Copilot mastery is a competitive advantage in the workplace by definition — because Copilot matters so much, because Copilot is in 90% of enterprises. Everything I'm describing is going to look like a basic requirement in about a year. So start now.
I have a full written guide available if you want it. If you want to see more videos on Copilot, let me know. This has been a long time coming — there was a ton to cover. I'm happy to dive deeper on specific industries and specific flavors of Copilot. I wanted to summarize it and give you the overall introduction first.
I've been excited to have meaningful AI conversations with people who use Copilot, so we can move from this world where people are just using it for email to a world where people can actually take advantage of the fact that we have taught the rocks to think — and now they're at work with us in Microsoft Word, and they can help us do smarter things faster. You still bring the brains to work. But Copilot will help.