How Copilot CoWork is Exposing Marketing Teams' Blind Spots

AI digital assistants are evolving beyond being individual productivity tools. As the name suggests, Microsoft’s new Copilot CoWork is a team-level tool that works across the shared 365 environment. By pulling context from emails, Teams chats, shared files, transcripts and meeting notes, the wider team can work from the same AI-generated insights at once.      

Marketing teams can draw on data that’s been accumulated over many years, so (for instance) campaign planning can draw on insights that may have been missed or forgotten as the (human) team make-up has changed. You might argue this makes the AI a participant as much as a tool. 

Collaborative AI tools that work directly within documents offer the kind of automation that directly reduces manual effort across the marketing team. This is significant because it offers those much touted day-to-day efficiencies that will save users time and ultimately cut costs for the businesses that use them.  

There’s an adage in data, which goes “something” like this - rubbish in, rubbish out. However, the irony is there’s likely lots of gold scattered across organizational data that the AI won’t be able to find if systems and/or internal processes are disconnected. 

When AI tools don’t live up to expectations, it’s usually for this reason. AI doesn’t fix your foundations, it exposes them. 

One of the benefits of a Microsoft 365 set-up is that data exists within a single ecosystem and CoWork can handle a degree of mess. But it can’t solve poor filing, so  if teams aren’t, or haven’t previously, aligned on naming and storage conventions, there won’t be a single source of truth for the AI to work from. 

While the best case scenario would be that everything’s kept in a single ecosystem, the reality is that’s not how tech stacks work for marketing teams: CRM, marketing automation, analytics, campaign data, DAM will all sit outside of Microsoft 365. 

So, while CoWork will be able to offer a great picture of the planning, it won’t have the same visibility of what’s happening in the tools where that plan is executed, the performance is analysed, or of the customer data that’s captured. 

The risk is that whatever CoWork presents back to you will look confident and polished regardless - so be very careful if your stack extends beyond 365. 

The biggest challenge is CoWork’s inability to access anything outside the realms of the 365 ecosystem. That aside, the sorts of issues that come up time and again are inconsistent naming conventions across files and folders, poor version control protocols and outdated templates that haven't been retired. 

But the other major challenges are uniquely human -  knowledge trapped in individual inboxes rather than going into shared spaces, and approval chains that exist in people's heads - but nowhere in the system.

Most teams will be guilty of having unofficial, albeit well-established, workarounds that may have been in place for years. We might all ignore the file in the ‘finals’ folder on the shared drive because Sarah likes to keep track of the finals in her own folder… Good for Sarah, but how does CoWork know that?  

The honest answer is that nobody has yet figured out what a multi-AI ecosystem will look like. CoWork does now integrate other AI models but many teams are still using ChatGPT or Claude separately, plus their martech platforms will have their own AI agents built in. Making this complexity work seamlessly could take time.

In the meantime, teams need to be very clear on what tool should be used for particular scenarios: General-purpose AI is strong for broad research, analysis, and drafting, while platform-specific agents work better for workflow execution within the stack for things like automating approvals, generating asset variations, and optimising campaign targeting.

Conversely, the worst possible scenario is a free-for-all in which people just use whatever's in front of them and nobody tracks the overlap, the costs can quickly add up when multiple tools accumulate.

What we can safely predict is the AI market will consolidate as the big platforms keep absorbing more AI capabilities, and the general-purpose tools will become more integrated. 

As it stands, two-thirds of IT leaders have reported unexpected charges on consumption-based AI, but that’s not surprising. The overlap between general-purpose tools and platform-specific agents is growing, and it’s hard to keep on top of the total picture because the pricing models are so varied - credits, consumption, per-seat, bundled and so on.

Visibility is key because you can’t rationalize if you don't know what you've got. A practical starting point is to map every AI capability across your stack: what it does, which team uses it, what it costs or what pricing model it sits on, and what business outcome it supports. Next, look for the overlaps. You'll almost certainly find you're paying for similar capabilities in multiple places.

The harder question is whether to consolidate or keep both. The answer will depend on whether the platform-specific version does the job better than the general-purpose one. The extra cost can be justified when the specialist tool is genuinely better for that workflow.

Introducing a tool like CoWork won’t necessarily slow things down, but it won’t automatically speed them up either. The issue lies in managing people’s expectations. 

Within the context of a marketing team, slowdown doesn’t happen because of the CoWork, it’s more likely to come from the AI’s lack of visibility of anything outside of 365. Let’s say you’re working on a campaign review pack, CoWork can build you a polished looking document in a matter of minutes, but it won’t be accurate because much of the key data is held elsewhere. That means someone then has to retroactively fill in the gaps and review it for any errors or hallucinations. 

The reality is it would have been quicker for the team to fill in the pack manually as they went along. Using AI when it can accurately access all the right information will speed up workflows, but otherwise it can’t work miracles.     

CoWork lives inside of 365 so that’s the right place to start. And that typically means fixing bad habits - auditing shared file structure and naming conventions so the AI can find what it needs. Then document key workflows, even roughly, so CoWork has a process to follow rather than guessing. Finally, clean up data sources by retiring outdated templates and assets that could confuse the AI.

The next step is to agree on a policy for sign-offs, as a team agree which decisions need human sign-off and which can be AI-assisted. That might sound simple but, in my experience, it rarely exists in writing.

You then need to look at what happens at the boundaries of the 365 ecosystem, these handoffs are usually where the biggest process gaps lie. Documenting those, even at a basic level, will make a big difference to how useful the AI can be.

When it comes to evaluating success of the roll-out, it makes sense to prove value in a small area, learn from what the AI exposes, fix the foundations, then scale. Pick something contained, like campaign briefing or reporting prep, run CoWork against it, and see what it surfaces. That's how you build confidence and demonstrate ROI without trying to overhaul everything on day one.