Can MSPs Actually Bill for Copilot Cowork?

I’ve been mulling over a question that doesn’t get asked enough, and I think it deserves a hard look: when Copilot Cowork lands with pay-as-you-go billing, do small and midsize MSPs actually have the skills to handle it? Not the product. The billing. Because from where I sit, that’s the part most of us are least prepared for.

For years, the SMB MSP model has run on something beautifully predictable: per-seat licensing. A client has thirty users, you sell thirty licences, you mark them up, and everyone knows what next month’s invoice looks like before it arrives. That predictability is the whole foundation. It’s what lets you quote a managed services agreement with a straight face. Consumption billing pulls that foundation out from under you.

We’ve Never Had to Read This Kind of Meter

PAYG is a different animal. Usage goes up and down. Costs follow. Suddenly you’re not selling a fixed thing, you’re selling access to a meter that ticks based on what people actually do inside Copilot. And here’s the uncomfortable truth — most of us have never had to read a meter like this before. We don’t have the muscle for it.

Think about the questions a client will ask the moment their first variable bill arrives. Why was it higher this month? Which users drove that? Was it worth it? If your answer is a shrug and a forwarded Microsoft invoice, you’ve got a problem. You need to pull the usage data, make sense of it, and explain it in plain English. That’s not a skill most SMB MSPs have built, because we’ve never needed it.

Configuring It Is the Easy Part

Turning Copilot Cowork on through the admin centre isn’t the hard bit. Microsoft will make that straightforward enough. The hard bit is everything that wraps around it — setting spending limits so a client doesn’t get a nasty surprise, deciding who gets access, and putting guardrails in place before usage runs away from you.

Then comes reporting, which is where I think the real gap shows. Can you stand in front of a client at the monthly meeting and show them, clearly, what they consumed and what they got for it? You’ll be living in the Microsoft 365 admin centre and the usage reports, and you may well end up pulling that data into Excel — perhaps asking Copilot itself to summarise the month’s consumption into something a business owner can read in thirty seconds. If you can’t produce that story, the client will assume the worst.

This Is a Discipline We Have to Learn

What worries me isn’t the technology. It’s that consumption billing is a genuine discipline, and it’s one the SMB MSP world has largely skipped. The cloud providers have been doing variable billing for years. Most of us serving small business have not. We’ve been comfortable in fixed-price land, and Cowork is going to ask us to grow up fast.

So I’d put the question back to you honestly. Could your business take on a client with Copilot Cowork tomorrow, configure it sensibly, manage the spend, and report on it with confidence? If there’s hesitation in that answer, you’re not behind — you’re normal. But the MSPs who close that gap early, who learn to read the meter and tell the story, are the ones who’ll own this conversation. The rest will be forwarding invoices and hoping nobody asks why.

Do an Audit of Who’s Actually In Your Corner

A few weeks ago I was on a long drive home from a client meeting and I started running through the names of people I’d actually spent real time with over the past year — not the LinkedIn list, the actual list. The ones who’d had my ear, my weekends, my energy and, in some cases, my best thinking. By the time I got home I realised something a bit uncomfortable. A handful of those names didn’t really belong on it anymore. Not because they’re bad people. Because they’re not pulling in the same direction I am.

That’s the part nobody likes to say out loud.

The question worth sitting with

There’s one question I think every business owner should ask themselves once a year, and it’s blunt: is the time I spend with this person making me sharper, or just making me comfortable?

Comfortable is the trap. Comfortable feels like loyalty, like history, like the easy lunch where nothing hard ever gets said. Growth tends to live somewhere else.

I’ve started using Copilot in Outlook to actually look at the data. I’ll ask it to pull together who I’ve met with most over the last three months and the answer is sometimes a surprise. The people I think I’m investing in and the people my calendar shows I’m investing in are not always the same list. Calendars don’t lie. They quietly show you where your time really goes, and once you’ve seen it laid out on a page you can’t unsee it.

Three filters I run people through

When I’m doing this stocktake, I look at three things, and I try not to overthink any of them.

The first is how I feel walking away from the conversation. If I leave a coffee buzzing with ideas and a list of things to try, that’s a signal. If I leave flat and quietly wanting a nap, that’s also a signal. The body keeps a fairly honest scoreboard, even when the head is trying to be polite about it.

The second is whether they’re actually moving. Movement doesn’t have to be loud — I know plenty of quiet operators who are building something serious. But if someone has been telling me the same story about the same plan for three years running, that stagnation will start to drag on you whether you notice it or not.

The third is whether they’ll tell me I’m wrong. The people who only ever agree with me have very little to offer me. The ones who push back, who ask the awkward question, who say “are you sure about that?” — those are the ones I keep close. Anyone who’s only ever told me what I wanted to hear has never once helped me improve.

The hardest cuts are the kindest people

The brutal part of this exercise isn’t the obvious ones. It’s the genuinely lovely humans who just don’t fit where you’re heading next. Lovely doesn’t equal useful. You can wish someone well, mean it completely, and still recognise that the season of regularly being in each other’s diaries has run its course. That’s not betrayal. That’s adulthood.

I keep a private Loop page with my thinking on this — a few names, a few notes, no judgement attached. Copilot helps me draft the reflection prompts when I’m not quite sure what I’m trying to articulate. It’s not a kill list. It’s a way of being deliberate about where my hours go next year, because nobody else is going to be deliberate about it for me.

If you’ve never done a Friendventory, do one this month. Block an hour in your calendar, pour something decent, and walk through the names honestly. The people in your corner shape the business you build next. Pick them on purpose, because by default you’ll just keep whoever was around when you started.

Need to Know podcast–Episode 367

This episode ultimately reflects on how organisations must adapt to an environment where solutions are no longer neatly balanced between simplicity and capability. Instead, businesses need to reassess priorities, stay informed, and make deliberate choices about which innovations deliver real value. For SMBs, the challenge is not just keeping up—but identifying what’s truly “good enough” in an increasingly complex cloud-first world.

Brought to you by www.ciaopspatron.com

you can listen directly to this episode at:

https://ciaops.podbean.com/e/episode-367-goldilocks-gone/

Subscribe via iTunes at:

https://itunes.apple.com/au/podcast/ciaops-need-to-know-podcasts/id406891445?mt=2

or Spotify:

https://open.spotify.com/show/7ejj00cOuw8977GnnE2lPb

Don’t forget to give the show a rating as well as send me any feedback or suggestions you may have for the show

Resources

CIAOPS Need to Know podcast – CIAOPS – Need to Know podcasts | CIAOPS

X – https://www.twitter.com/directorcia

director@ciaops.com

CIAOPS Blog

Join my Teams Shared Channel – CIAOPS

CIAOPS Merch store – CIAOPS

Become a CIAOPS Patron

CIAOPS AI Dojo

CIAOPS weekly news update – CIA Brief – CIAOPS

CIAOPS Labs – The Special Activities Division of the CIAOPS

Support CIAOPS

Get your M365 questions answered via email

Join my email list

A special thanks to the CIAOPS Patron community for making this podcast possible. You can find the benefits of a subscription to the community and become a member at https://www.ciaopspatron.com

AutoJack: How a single page can RCE the host running your AI agent

Azure Sets a New Performance Record for LLM Training Benchmark at Extreme Scale

Forrester names Microsoft a Leader in the 2026 Extended Detection and Response Platforms Wave report

Copilot Cowork is now generally available

Microsoft Defender email security benchmarking: Key insights from one year of data

Stay productive in new Outlook for Windows with these 5 features

What’s new in Power Platform: June 2026 feature update

What’s New in Notebooks | June 2026

Mercedes-AMG PETRONAS F1 Team responds to the intensity of race weekends with Microsoft

AI brands as bait: How threat actors are using the AI hype in social engineering

AI Is Starting to Feel Like the Petrol Bowser

I filled up the car last weekend and did that thing we all do now — glanced at the price per litre, winced a little, and worked out whether to fill the tank or just put in enough to get me through the week. It struck me, standing there, that I’ve started having the exact same conversation about AI.

For a while, AI felt free. You paid your subscription, you used it, and the meter never seemed to run. That era is ending. Token prices, usage caps, premium request limits — the cost of running AI is becoming visible, metered, and impossible to ignore. For a small business, it’s shifting from a novelty into a line item. And like petrol, it’s no longer optional. You can’t really run the business without it, but you can’t ignore what it costs either.

A running cost, not a one-off

The mistake I keep seeing is treating AI like a piece of software you buy once and forget. It isn’t. It behaves far more like fuel or electricity — something you consume, in varying amounts, every single day. Some weeks you’ll barely touch it. Other weeks, when you’re deep in a proposal or cleaning up a quarter’s worth of numbers, you’ll burn through it.

That changes how you should think about it. A running cost needs watching. It needs a budget. And it needs someone, every so often, asking the plain question: are we actually getting value for what we’re spending here?

Spend where the work actually is

Here’s where it gets interesting for a small business. You haven’t got an unlimited tank, so you have to decide where AI earns its keep. For most of the small operations I work with, the answer isn’t exotic. It’s the boring, repetitive, time-sapping work — the email triage, the first draft of a report, the summary of a long meeting nobody wants to rewatch.

That’s exactly where Copilot inside Microsoft 365 pays for itself. Asking Copilot in Outlook to clear and draft replies to a morning’s backlog saves real hours. Having it summarise a Teams meeting you missed, or pull the key figures out of an Excel workbook, turns an afternoon into a few minutes. The trick is to point your spend at the tasks that are costing you time and money today — not the shiny demos that look clever but never touch your actual week.

Economise without going without

The same way you don’t leave the car idling in the driveway, you don’t want AI burning through your allowance on low-value busywork. Be deliberate. Use the everyday Copilot features that already come with your licence before you reach for premium-priced add-ons. Give your people a quick steer on what’s worth asking and what’s just noise. And look at the bill — actually check where the consumption is going, the way you’d question a sudden jump in the power account.

None of this is about spending less for its own sake. It’s about spending on purpose.

Where I’ve landed

Petrol taught small business owners to think in terms of value per trip, not just price per litre. AI is heading the same way. The ones who do well won’t be those who spend the most or the least — they’ll be the ones who know exactly what they’re buying and why. Watch your usage, back the work that matters, and treat AI like the utility it’s quietly becoming. The meter is running now. Best to know what it’s running on.

The Clients You Need to Let Go

Last Saturday morning I sat down with a coffee and went through my client list properly. Not the polished version on a spreadsheet — the honest one. The list where you stop and ask yourself which names make your shoulders drop when they appear in your inbox.

Out of every hundred clients, maybe twenty fit that description. They weren’t bad people. But they were the ones who turned a thirty-minute call into ninety. The ones who treated every standard you set as optional. The ones who had a complaint queued up before they’d even tried what you suggested.

And here’s the part most MSP owners don’t want to admit out loud: those clients aren’t just slightly more work. They quietly run the place.

The Maths You’ve Been Avoiding

Sit down and add up the hours. Not the billable ones. The total ones. The Friday afternoon you spent re-explaining MFA for the fifth time. The Sunday email you answered because they wrote at 9pm and you didn’t want them to think you’d ignored them. The team meeting that ran twenty minutes long because someone needed to vent about a client who refuses to follow the playbook.

If you log it honestly for a fortnight, the pattern is uncomfortable. A small handful of clients eat a disproportionate slice of your week, your team’s morale, and your own headspace. The rest of your book — the ones who pay on time, listen to your advice, and treat your team with respect — get whatever scraps of attention are left over.

Copilot Can Show You What You Already Suspect

Here is where Microsoft 365 quietly earns its keep. Ask Copilot in Outlook to summarise the volume and tone of messages you’ve exchanged with a particular client over the last quarter. Open your Teams channels and let Copilot in Teams surface the recurring complaints by topic. Pull up the recap from your last quarterly review with that client and read it back without the emotion of being in the room.

You will see it plainly. The same three issues raised four quarters in a row. The same standards politely ignored. The same tone in every second message. Copilot isn’t telling you anything new. It’s just laying out the evidence you’ve been too busy or too loyal to read properly.

I have started doing this every quarter. It takes about twenty minutes per client, and it removes the wishful thinking. You stop asking “are they really that bad?” and start asking the more useful question: “Why am I still putting up with this?”

The Conversation Itself

When you do decide to end the relationship, do it cleanly. Refund what is fair. Offer to hand over their data in a tidy package via OneDrive. Recommend a provider who is genuinely a better fit for them — not a punishment, just a different match.

I draft the offboarding email in Word with Copilot’s help, sit on it overnight, then read it again in the morning. If it still says what needs saying, I send it. No long justification. No door left ajar. A short, professional close.

What You Get Back

The first time I did this properly, three things shifted within a month. My team felt lighter. Monday mornings stopped starting with dread. And the clients I genuinely respect — the ones who do the work, follow the advice, ask good questions — got noticeably more of me.

That last shift is the one that matters. The people paying for your best work deserve your best attention, and you simply cannot offer it while a small group is quietly running off with the fuel.

Sensitivity labels and auto-labelling: put a name on your data

Most people meet sensitivity labels the wrong way.

They see encryption and Purview and compliance in the same sentence and decide it’s an enterprise problem. Something for banks. Something that needs E5, a consultant, and six months.

So they leave the whole thing switched off.

Then a client emails a payroll spreadsheet to the wrong “David” in the address book, and it becomes very much their problem.

Here’s what I want you to understand. A sensitivity label isn’t a lock. It’s a name tag you put on information so Microsoft 365 knows how to treat it — and most of the value shows up before you’ve encrypted a single file.

What are sensitivity labels, really?

A sensitivity label is a tag that travels with the content. Apply Confidential to a Word doc and that label rides along into SharePoint, OneDrive, Teams, and the email it’s attached to — even off your tenant, if you allow it.

What the label does is up to you. It can simply mark the document with a header, footer, or watermark. It can show a visual classification users notice. Or it can go further and encrypt the file so only the right people open it.

That range is the bit people miss. You don’t have to start with encryption. You can start with classification — and classification on its own changes how people handle a file.

Step-by-Step: build your first label

Everything lives in the Microsoft Purview portal at purview.microsoft.com, under Information Protection. Microsoft seeds new tenants with a default set, but build and publish your own so the names actually mean something to your client.

Turn on labels for files

Before a label will stick to documents in SharePoint and OneDrive, go to Settings > Information Protection and turn on co-authoring for files with sensitivity labels. Skip this and your labels won’t apply to files at rest. It catches everyone once.

Create the label

Under Information Protection > Sensitivity labels, select Create a label. Give it a name your users will understand, not a compliance codeword. Set the scope to Files and emails.

Decide what it does

Now choose protection. Content marking — a header, footer, or watermark — is the gentle option. Access control with encryption is the heavy one. For your first label, pick marking. You can add teeth later.

Publish it

A label nobody can see does nothing. Create a label policy, add your label, and publish it to a group of users. Now it shows up under the Sensitivity button in Word, Excel, PowerPoint, and Outlook.

That’s a working label. Manual, user-applied, and included with the sensitivity-label entitlement most of your Business Premium clients already hold.

Step-by-Step: let the tenant do the labelling

Manual labels rely on people remembering. People don’t. So the next step is auto-labelling — and this is where the licensing line sits, so be straight with clients.

Pick your method

There are two. Client-side auto-labelling prompts or applies a label while someone edits a document in Office. Service-side auto-labelling policies scan content already sitting in SharePoint and OneDrive, plus mail moving through Exchange, with no user involved at all.

Run it in simulation first

This is the setting that saves you. An auto-labelling policy runs in simulation mode — it shows you exactly what would get labelled across the tenant without touching a thing. My recommendation? Always simulate, read the matches, fix your conditions, then turn it on.

Mind the licence

Auto-labelling — both flavours — needs the E5-tier Information Protection entitlement, not the base Business Premium one. Manual labels are included. Automatic ones aren’t. Don’t promise a client auto-labelling on a licence that doesn’t carry it.

“So do I need encryption on everything?” No. Most of my labels never encrypt anything. They classify. Encryption is reserved for the one or two labels that genuinely need it.

Here’s a starter taxonomy worth copying:

Personal
Public
General
Confidential
Highly Confidential

Notice what’s missing? Encryption — on four of the five. The top label might lock files down. The rest just name the sensitivity so people, and the tenant, treat them accordingly. Classification first, control second.

Why this actually changes behaviour

A labelled file behaves differently. DLP policies can key off the label. Auto-labelling can find the credit-card numbers your user forgot were buried in an old quote. And both SharePoint and Copilot respect the access a label enforces — which matters more every month.

But the quiet win is human. When someone clicks Confidential and a watermark appears, they slow down. They think before they forward. The label is doing the teaching.

Set it up once. It keeps working while everyone’s asleep.

Sensitivity labels aren’t there to make compliance harder. They’re there to make a careless mistake hard to make by accident.

If you’re rolling out Microsoft 365 and your clients’ data still has no name on it — that’s the gap. Put a name on it.

Your first Copilot Studio agent — and the bits the demo skips

Everyone wants to build an agent right now.

It’s the demo that makes the room go quiet. You describe what you want in plain English, and thirty seconds later there’s a working assistant answering questions from your own documents.

Then the bill arrives. Or worse — it doesn’t, because nobody published the thing and it’s sitting in a maker’s personal environment doing nothing.

Here’s what I keep seeing. People treat an agent as a science project. They build one, screenshot it, post it on LinkedIn, and move on. The client never touches it.

That’s not a product. That’s a party trick.

If you’re an MSP, the agent isn’t the deliverable. The service around it is. Let me show you the bits people skip.

What is a Copilot Studio agent, really?

Strip away the marketing and an agent is three things: instructions, knowledge, and tools.

Instructions are the job description — what it does, the tone it takes, when to give up and send someone to a human. Knowledge is what it’s allowed to read: a SharePoint site, a public page, a handful of documents. Tools are the actions it can take, like logging a ticket or looking up an order.

That’s it. You’re not training a model. You’re scoping a very literal new hire.

And there are two doors into this. The Agent Builder baked into Microsoft 365 Copilot, and Copilot Studio proper. The choice between them comes down to audience and reach — a quick helper for your own Copilot chat, or a real one you publish out to Teams, a website, wherever your client’s people already work.

Step-by-Step: building your first agent

Portal only. No code.

Open the studio

Go to copilotstudio.microsoft.com and check the environment name in the top right. This matters more than it looks. Build in the wrong environment and you’ll be untangling it later. Pick the client’s environment deliberately, every time.

Describe what you want

You get a box. Type what the agent should do in your own words — that’s the whole guided, no-code start. Mine for a small-business help desk reads like this:

Help our staff with common IT questions — password resets,
VPN setup, who to call for hardware. Answer only from the
company IT knowledge base. If you're not sure, say so and point
them to the service desk. Don't guess.

Notice what’s missing? No model to pick. No temperature dial. No code. You’re writing a job description in plain English. And the load-bearing line is the last one: ask it once to admit when it doesn’t know. An agent that says “I’m not sure” beats one that bluffs with total confidence.

Add knowledge, narrowly

Point it at one good source first. The company FAQ. The IT policy site. Not everything you own — one trusted thing. Test it. Then add more. An agent fed the whole tenant on day one is an agent that confidently surfaces the wrong document.

Test like a real person

Use the test pane on the right. Ask the messy questions: “can’t log in”, “where’s the VPN guide”, “who do I call”. If it waffles, fix the instructions — not the model.

Publish it

This is the step everyone forgets. An agent does nothing until you hit Publish. And every change you make afterwards needs publishing again, or your users sit on the old version. Then add a channel — Teams and Microsoft 365 Copilot are the obvious two — and decide who’s allowed in.

By default it’s Authenticate with Microsoft, which means your tenant and no one else. Leave it there. “No authentication” hands a working link to the entire internet. For an internal SMB agent, that is never what you want.

The licensing question you answer first, not last

Here’s the bit the demos never mention.

Agents built in the Agent Builder inside Microsoft 365 Copilot ride along with the licence your client already pays for. The custom agents you build and publish in Copilot Studio consume Copilot Credits — and since September 2025 that’s the currency Microsoft meters them in. Pay-as-you-go, or prepaid packs.

“So I just describe it and it works?” Yes — and that’s exactly the trap. The describing is the easy part. The scoping, the publishing, and the consumption call are the actual job.

So before you build, settle one thing: is this an internal helper for a few Copilot users, or a published agent the whole company will hammer all day? The answer decides which door you walk through and whether there’s a consumption bill hanging off it.

Get that wrong and you’ve either over-built a toy or under-budgeted a product.

The point of an agent isn’t to look clever for ninety seconds in a demo.

It’s to take a repetitive question off a human’s desk and answer it the same way, every time, at 2am. A published agent earns its keep. An unpublished one is a screenshot.

Build one narrow agent, publish it properly, and put your name on the governance around it. If you’re not showing your clients this, someone else will — and they’ll be the ones charging for it.

The Hiring Shift Copilot Has Quietly Forced

A few weeks ago I watched someone forward a Copilot-drafted email to a client without really reading it. The tone was off. A figure was wrong. The client picked it up before they did. Nobody was being lazy — they were being efficient, in the exact way we’ve all been trained to be efficient. Get the task done. Move to the next one. That small moment has stuck with me, because I think it points at something much bigger that’s quietly reshaping how small businesses need to think about the people they hire, and the behaviours they reward.

For most of my working life, businesses hired people to do things. Send the email. Build the report. Process the invoice. Update the spreadsheet. Speed and responsiveness were the metrics that mattered most. If you could turn incoming requests into completed work quickly and reliably, you were valuable. That model held up because doing the thing was the hard part, and doing it well took real skill, attention and time.

The work that used to be hard isn’t hard anymore

Copilot has quietly upended a lot of that. Drop into Outlook and you can have a draft reply in seconds. Open Word with a half-formed idea and Copilot will hand you a structured first pass before your coffee has cooled. Excel will summarise a sheet, flag the anomalies, and suggest a formula it thinks you wanted. Teams will catch you up on the meeting you missed and tell you what was decided and who owns what. Every one of those tasks used to be where someone earned their keep. Now they’re table stakes.

What’s left isn’t the doing. It’s the deciding. Was that actually the right answer? Does the draft fit the specific client we’re dealing with, or the average client Copilot has imagined? Is the figure it pulled from SharePoint the current number, or one from two restructures ago that nobody got around to archiving? Should we even send this email, or is there a phone call hiding behind it?

We’ve been training people out of asking

Here’s the part I find uncomfortable. A lot of small businesses have spent years quietly training their people out of pausing. We rewarded responsiveness over reflection. Inbox zero over inbox thoughtful. Closing tickets over questioning whether the ticket made sense in the first place. Over time, some genuinely good people got very fast at moving work between systems without really engaging with any of it.

That worked when the slow, expensive part of the job was producing the output. It doesn’t work now. The slow part is the human bit — the read, the sense-check, the quiet moment of “this doesn’t sit right with me” that keeps a business out of trouble with a client, a regulator, or its own future self.

The maths has flipped

When output was expensive, mistakes were rare almost by accident. There was friction in the system, and friction gave people time to second-guess themselves. A bad proposal took half a day to write, so it usually got read twice before it left the building. A bad proposal now takes ninety seconds, and so do the next ten after it.

The businesses that come out of this next stretch in good shape won’t be the ones with the fastest Copilot adopters. They’ll be the ones that have quietly shifted what they recognise and reward — from doing the thing, to noticing when the thing isn’t right. That’s a hiring conversation, a performance review conversation, and most of all a culture conversation. And in most of the small businesses I see day to day, it’s well overdue.