All Blogs
Two professional services firms. Same Microsoft 365 licenses. Same Copilot subscription. Same price tag per user per month.
One is getting three hours back per employee per week. Senior staff are doing higher-value work. Client meetings are sharper. Proposals go out faster.
The other sent an IT announcement six months ago, called it an AI rollout, and hasn’t looked back. Eighty percent of their Copilot seats are sitting unused.
The difference isn’t the software.
We work with professional services firms across Western Pittsburgh every day. We see who’s getting real value from their Microsoft 365 investment and who isn’t. The gap almost never comes down to budget or company size.
It comes down to five things.
Why Do So Many Copilot Deployments Go Nowhere?
The pattern is almost always the same. Licences get purchased. IT configures access and sends a company-wide email. Leadership waits for results.
Three months later, adoption is low, ROI is unclear, and someone asks whether Copilot is actually worth the cost.
It is. Just not the way most organizations are deploying it.
Paying $30 per user per month gets you access to the tool. That’s it. What determines whether your team actually uses well it depends on what happens after you turn it on.
Which use cases you prioritize. How your team learns to ask better questions. Whether Copilot can see your actual work data. Whether anyone has thought through the security implications before scaling it company-wide.
Skip that work, and even motivated employees get inconsistent results, lose confidence, and quietly stop opening it.
That’s not a Copilot problem. It’s an implementation problem. And it’s the most expensive AI mistake we see professional services firms make, not skipping Copilot entirely, but paying for it and getting nothing back.
What Are the 5 Things Best-in-Class Companies Do Differently?
Across the professional services firms we work with, the ones getting measurable results share five specific behaviors. None of them require different technology. They require a different approach to using the technology they already have.
1. They connect Copilot to actual work data, not just the web
Microsoft Copilot Chat has two modes: Work and Web. Web mode searches the public internet. Work mode searches your organization’s own data like emails, meetings, files, Teams conversations, SharePoint, everything inside your Microsoft 365 environment.
Most organizations leave both modes available and let employees decide. Best-in-class firms set Work mode as the default for their operational teams and train staff on when to switch.
The output difference is significant. Work mode can surface the actual concerns a client raised in a meeting last month, the current status of a project based on recent email threads, or a draft that references your firm’s own prior work, not a generic version of what that document might look like.
In practice: “What were the three biggest concerns our client raised in our Q1 check-ins, and how have we addressed each one?” ask in Work mode, drawing from actual meeting notes and email threads. Not a generic summary. Your summary.
2. They use AI for research and briefing prep, not just drafting
Most teams discover Copilot through drafting: write an email, summarize a document, clean up a paragraph. That’s useful. It’s also the lowest-value application of the tool.
The firms seeing the highest ROI are using Copilot’s Researcher agent for preparation work: client meetings, board presentations, quarterly reviews, strategic planning cycles. Researcher pulls from your internal project history, recent communications, external news sources, and synthesizes it into a coherent brief. Work that previously took two to four hours of a senior employee’s time.
In practice:“Create a briefing for tomorrow’s quarterly review with [Client Name]. Include our recent interactions and any open items from our last meeting, their current challenges based on our notes, and relevant industry developments from the past 30 days.” — Researcher handles this in minutes.
3. They treat prompt engineering as a firm-wide skill not an IT responsibility
The biggest gap between firms that get results and firms that don’t is how their people ask questions. Vague prompts return generic output. Specific prompts with context, format instructions, and a clear audience return something you can actually use.
The difference isn’t subtle:
| Less Effective Prompt | More Effective Prompt |
| Write a summary of this meeting. | Summarize the three biggest unresolved decisions from last week’s project status meeting and suggest how I should address each one in Friday’s executive briefing. Keep it under 200 words. |
| Write about our Q2 performance. | Write a 300-word update for our leadership team on Q2 performance, focusing on the two service lines that grew and the one that declined. Tone should be direct and factual. Reference [file]. |
| Help me prepare for the client call. | I have a call with [Client Name] tomorrow. Based on our last three interactions and their current contract scope, what are the most likely concerns they will raise and what are the best responses to each? |
The firms getting ROI have trained their teams on this especially senior staff. It doesn’t take long. A half-day session and a shared prompt library gets most teams most of the way there.
4. They apply AI to data analysis, not just communication tasks
Most professional services firms are doing data analysis the same way they have for years: spreadsheets, one-off reports, or an outside consultant.
Copilot’s Analyst reasons over multiple data files simultaneously — financial statements, billing data, client usage metrics, project performance records and produces insights and visualizations in minutes that would previously require hours of manual work or an outside engagement.
In practice: “Look at these three months of billing data and tell me which service lines are growing, which are stalling, and where we are losing margin. Flag any clients whose billing pattern has changed significantly compared to the same period last year.”
For firms that currently rely on outside consulting for this kind of analysis, or where a senior employee’s time is consumed by reporting, Analyst often returns its ROI in the first month.
5. They build governance before they scale adoption
Copilot can see everything your employees can see. That’s what makes it useful. It’s also why governance can’t be an afterthought.
A firm with inconsistent file permissions, unreviewed sharing settings, or no policy on how AI-generated content gets handled before client delivery isn’t just running a messy deployment. It’s running a confidentiality risk that gets bigger as more people start using the tool.
The firms that scale AI confidently define this early:
• Which data Copilot can access and which data requires explicit exclusion
• How prompts containing client information are reviewed before use
• How AI-generated outputs are reviewed before they go to clients or external stakeholders
• What employees should and should not ask Copilot to do with sensitive information
Microsoft’s compliance framework handles the infrastructure side. The policy and culture side is on you. Sort it out before you scale, not after something goes wrong.
How Does a Strategic Copilot Deployment Compare to a Standard One?
The five behaviors above are not abstract principles. They translate directly to measurable differences in outcomes. Here is what that comparison looks like across the dimensions that matter to a professional services firm:
| Factor | Standard Deployment (Licenses + IT Email) | Strategic Deployment (MBT Framework) |
| Adoption rate at 90 days | Typically 20–35% of licensed seats active | 70–90% active, benchmarked against structured rollouts |
| Prompt quality | Ad hoc; employees self-discover with inconsistent results | Firm-wide prompt training built into onboarding |
| Data connectivity | Web mode default; Work mode used inconsistently | Work mode set as default for operational teams from day one |
| Use cases targeted | Email drafting, basic summarization | Research prep, data analysis, client briefings, status reporting |
| Governance | Absent or informal | Data access policy, output review process, and compliance framework configured before scaling |
| Time to measurable ROI | Unclear; often never demonstrated | Typically visible within 60–90 days of structured deployment |
| Security exposure | Often unknown; permissions not reviewed | Permissions audited as part of onboarding |
What Is the Common Thread Across All Five Behaviors?
Every firm on the list above has the same Microsoft 365 subscription. The same Copilot license. The same underlying technology.
What’s different is how it was deployed.
The firms seeing results started with clear use cases tied to actual workflow pain points, not a company-wide email telling people to use AI more. They trained their teams on prompt quality, not just tool access. They connected Copilot to actual work data before launching it broadly. And they had governance sorted before scaling, not after something forced the issue.
That combination (technical configuration, workflow analysis, and change management) is what makes the difference. It’s also what most firms can’t pull off on their own the first time. Not for lack of trying. Because it requires someone who has done it before, across enough different firms to know where it breaks down.
Most DIY deployments end up back at 30 percent adoption six months later. The tool gets blamed. The real issue was never the tool.
What Does Waiting Actually Cost?
The five behaviours above explain what the best-performing firms are doing. The harder question is what it costs when a firm holds off while competitors who have already made the move are gaining ground. That’s the subject of the next post in this series.
The Bigger Risk? Doing Nothing: Why Waiting on AI Could Cost You Clients
When This Framework Isn’t the Right Fit
Not every firm reading this should go hire an implementation partner. There are situations where this approach doesn’t make sense right now.
If your organization is mid-migration like actively moving platforms, consolidating systems, or in the middle of a major Microsoft 365 reconfiguration then adding a Copilot deployment on top of that creates more problems than it solves. Finish the migration first.
If your firm operates in a heavily regulated sector with AI-specific compliance requirements that haven’t been legally reviewed, governance needs to happen before activation. Not at the same time.
If budget is genuinely constrained and a structured implementation isn’t feasible right now, there’s still a better path than doing nothing. Start with a single team, a single use case, and a focused month of prompt training. A narrow pilot almost always produces enough return to justify the broader investment.
And if your team is already using Copilot and seeing results like adoption above 60 percent, consistent prompt quality, governance sorted then this article isn’t for you. You already have what most firms are still trying to build.
See This Framework in Practice
On June 16 at 11:00 AM EST, Larry Schwartz is hosting a free live webinar with Julie Hodges, a Copilot Expert from Microsoft. The session covers how this implementation framework works in practice, with real examples from professional services deployments, an honest look at where it breaks down, and time for direct Q&A.
Copilot 2.0: From AI Hype to Practical ROI Tuesday, June 16, 2026 | 11:00 AM EST | Microsoft Teams Live and on-demand recording available
[Reserve your spot here]
Frequently Asked Questions
| Question | Answer |
| What is Microsoft Copilot ROI, and how is it measured? | The most reliable measure is time recovered. Take the hours per employee per week returned from routine tasks like meeting prep, email drafting, research, and reporting, then multiply by the employee’s hourly cost. That’s your baseline. Secondary measures include adoption rate, prompt quality, and how much of your team is operating under a defined AI policy. Microsoft’s own data shows structured deployments saving users one to two hours per day. Run conservative numbers on that. Forty-five minutes per day across a 50-person firm returns $56,000 to $75,000 per month in recovered capacity. Your monthly licence cost is $1,500. That’s the gap between a strategic deployment and an unused one. |
| What are the best practices for Microsoft Copilot adoption in professional services? | Five practices consistently drive adoption above 60%: connecting Copilot to organizational data (Work mode as default), training staff on prompt engineering as a standard skill rather than an IT function, identifying 3–5 specific high-value use cases before broad rollout, configuring data governance and access permissions before scaling, and assigning internal adoption champions who own outcome tracking. General rollouts without these elements routinely stall at 20–35% adoption. |
| Why is Copilot adoption so low at most companies? | Work mode connects Copilot to your organization’s own Microsoft 365 data. Emails, meeting notes, files, Teams conversations, SharePoint. Everything inside your environment. The output it produces is informed by your actual work, not a generic version of it. Web mode connects Copilot to the public internet. Useful for general research. Not useful for anything that requires context specific to your firm or your clients. Most standard deployments leave the choice to the employee. The result is inconsistent output and low confidence in the tool. Best-in-class firms set Work mode as the default for operational teams from day one and train staff on when switching to Web makes sense. |
| Do I need a partner to implement Microsoft Copilot strategically? | Not necessarily. A firm with an experienced internal IT team, a clear understanding of its data governance requirements, and the bandwidth to run a structured change management process can implement Copilot effectively without outside help. The question is whether that capacity exists. Most professional services firms of 20–200 people do not have a dedicated resource who can manage configuration, prompt training, governance policy, and adoption tracking simultaneously. That is where a partner changes the outcome. |
| What is the difference between Copilot Chat Work mode and Web mode? | Work mode connects Copilot to your organization’s own Microsoft 365 data. Emails, meeting notes, files, Teams conversations, SharePoint. Everything inside your environment. The output it produces is informed by your actual work, not a generic version of it. Web mode connects Copilot to the public internet. Useful for general research. Not useful for anything that requires context specific to your firm or your clients. Most standard deployments leave the choice to the employee. The result is inconsistent output and low confidence in the tool. Best-in-class firms set Work mode as the default for operational teams from day one and train staff on when switching to Web makes sense. |
| What is the Copilot Researcher and Analyst agent? | Researcher handles complex, multi-step research by combining your internal Microsoft 365 data with external web sources. It’s built for preparation-heavy work: client briefings, board materials, project status reports, strategic analysis. Tasks that used to take a senior employee two to four hours now take minutes. Analyst reasons over structured data files. Financial statements, billing records, project metrics. It produces insights and visualizations from data that would otherwise require hours of manual work or an outside consultant. Both live in the Microsoft 365 Copilot app under the Agents panel. Both require Work mode to access your internal data. |