Copilot Cowork: five questions customers are asking

On June 16, I published a post on the changes that the general availability of Copilot Cowork  will bring including the capability shift and new billing model for customers plus the five immediate actions that partners needed to take before usage-based billing kicked in for all end users (including Frontier preview users) on July 1.

Since then, I've spoken with partners and with our own Microsoft experts about the questions  customers are asking. This post covers what I feel are the most important of those questions, plus some suggestions on how they can be handled to guide a positive conversation around Cowork adoption.

1. "We used Cowork for free in Frontier preview. Why do we suddenly need to start paying?"

As I mentioned in my earlier post, the tranche of customers who used Cowork as part of the three-month Frontier preview was always going to be the low-hanging fruit. They’ve got used to Cowork and, in many cases, absolutely love it. But as all of us in the channel know, low-hanging fruit is not always synonymous with easy sales. Inevitably, some of these customers are now pushing back because Cowork was “free” and they resent the advent of usage-based billing (UBB).

The reframing here is actually pretty straightforward.

Frontier was a three-month preview programme, and like any preview, it was always going to transition into a paid model at GA. Microsoft made Copilot Cowork generally available on June 16 2026 and introduced usage-based billing through Copilot Credits. For tenants that had active Frontier Cowork usage during the preview, Microsoft provided a transition period from June 16–30 , with usage-based billing required from July 1 to maintain access.

During that transition period, selected Frontier usage was not separately charged, giving customers time to configure billing and decide how they wanted to govern Cowork adoption. Beyond those specialised tasks, customers should be reminded that the $30 Copilot licence still represents incredible value for money and can do so much more than many organisations realise (as you can demonstrate using the table below.)

This reframing leads naturally into discovery questions like: where did you find Cowork most useful during the preview?  Which teams and individuals made most use of  it?  Which teams or individuals didn’t use Cowork despite having access? Are you sure you’re making full use of  everything you have in your existing Copilot licences?

Once customers appreciate this difference between an agent (Cowork) and an assistant (Copilot) and how the two can best work together, the conversation shifts in a more productive direction. They are able to focus on the benefit of being able to choose the right tool for the right job rather than feeling they have lost anything that was “free”.

2. "I’ve heard the costs on this are going to be unpredictable. How do we control the bill?"

Many customers will not be familiar with metreed billing, and they need to understand that strong governance is essential to keep costs under control when using this model. This need not be an insurmountable barrier to sales. In fact, helping the customer put effective “tokenomics” in place (or at least understand the difference between licencing and UBB) to control their costs is a great opportunity to earn trust and build longer term relationships.

With that approach, you can afford to be upfront about the potential costs of poorly governed agentic AI.  On the upside, you can also let them know that with Cowork control is built in for them if they choose to use it.  Admins can set credit limits at tenant, group, and user level, configure alerts, and cap spend before it starts. They can also opt for prepaid capacity packs for predictable baseline usage with pay-as-you-go as a top-up for genuine spikes.

Partners who help customers get those limits in place before switching to Cowork position themselves as trusted consultants rather than sellers and earn the right to own the relationship going forward.

3. "Isn't this just Copilot Chat with a new name?"

As mentioned above, not all customers or prospects will immediately understand the difference between an assistant AI tool (Copilot Chat) and an agentic AI work environment (Cowork).

They will naturally be focused on their own desired business outcomes rather than the specific tools that deliver them.  This can often be the root of some pretty funademental objections (“I thought we were already paying for this?!”), so it will pay partners to be ready with a few simple bullets before taking a Cowork conversation any further.

• Copilot Chat advises and responds
• Cowork plans and acts
• Chat keeps the user in the driving seat: user asks, AI helps, user reviews
• Cowork takes the wheel: users delegates a multi-step job; AI plans, executes across applications, and hands back a finished result.

A quick recap like this gives your customer further context on why the two systems need to be billed differently with one on a licence and another on UBB and helps them make an informed decision on where and when they need to shift from working with AI (Copilot Chat) to getting AI working for them (Cowork).

When that difference is more clearly understood, you can probably expect the next question to follow shortly afterwards.

4. "Is it safe to let AI act on our behalf like this?"

Agentic AI execution introduces genuinely new security exposure that organisations will need to manage if they wish to take advantage of the productivity benefits that tools like Cowork will bring. These tools can not only answer questions but also access data and run processes autonomously. The security focus therefore shifts from what the AI says to what the AI does.

I’d recommend two starting points when it comes to fielding such questions:

• Cowork’s own architectural guardrails
• The customer’s operational best practise.

At the architectural level, Cowork has a strong story to tell compared with some other agentic choices.

According to Microsoft's own documentation, Cowork is governed through Microsoft Purview capabilities including DSPM/ DSPM for AI, auditing, sensitivity labels, encryption without sensitivity labels, Insider Risk Management, Communication Compliance, eDiscovery and Data Lifecycle Management. Microsoft’s GA blog also says prompts, responses and generated artefacts flow through existing Microsoft 365 controls and are governed, discoverable and retained securely, with DLP coming soon.

At the operational level, best practise is very simple. Customers should treat Cowork as a governed capability from day one rather than something to lock down only after something goes wrong.

5. "Is this realistic for a business our size?"

This one (or variations on it) focuses all of the preceding questions. It is a clear signal that the customer needs additional information about cost and security and value (in the shape of added functionality and whether that is worth paying for) when they are considering Cowork.

The short answer to the question is: “Yes. Usually.” The more comprehensive reply is that the governance required to make Cowork a secure and successful productivity tool scales down just as well as it scales up because it encourages:

• Choosing the right tool for the right task (Copilot vs. Cowork)
• Setting deliberate credit limits for anyone using Cowork
• Running a pilot before a broad rollout.

With a disciplined approach,  a 200-person business that picks three high-value scenarios for Cowork use and caps its credits should see ROI as readily as a multinational. The key factor isn’t the headcount: it’s the maturity of approach.

From the partner’s point of view, this is another opportunity to help customers with genuine consultancy rather than a simple sales motion. Knowing when to advise on limiting spend will earn you a lot of trust.