How Miro's champions run their week with Claude Cowork

Claude Cowork spread across Miro, the company building a collaborative workspace where agents and teams converge, on the strength of what its champions are building with it. That includes an influencer content reviewer, an automated project tracker, and a thought partner that pressure-tests half-formed problems before there's anything to draft. We spoke with three of those champions about the workflows they run, how they keep quality high as more colleagues build, and where the value has surprised them.

Anthropic: What's surprised you most about how Claude Cowork gets used?

Kate Hostetler, Senior Product Marketing Manager at Miro: I wasn't surprised how quickly Claude Cowork spread. What surprised me was where the value turned up. I expected an output machine: something to churn out drafts, dashboards, and briefs faster. And it does all of that. What I didn't see coming is how much of the real use is thinking, not producing.

I started talking through half-formed problems with Cowork and having it pressure-test my logic, flag what I'm missing, and push back instead of just agreeing. It's become a thought partner for the messy, upstream part of the work: figuring out how to approach an ambiguous project before there's anything to draft. That's changed where the team reaches for it: less the last step in a process, cleaning up and formatting, and more the first step, when you're still figuring out what you're trying to do.

Anthropic: How were you handling that work before Claude Cowork?

Hostetler: On any given project, the context I need is scattered across multiple teams and tools: Slack threads, meeting notes, emails, docs, and Miro boards. I spent more time gathering, reconciling, and distributing information than actually doing the thinking. Claude Code's learning curve is steeper, and the terminal is intimidating, so what I initially automated mostly stayed locked to me. I couldn't easily hand things off or bring my team along, which meant I was running fast on my own, and that only gets you so far.

Claude Cowork closed that gap. The superpowers are there, but it's collaborative and far more approachable. The time back is real, too. Between triaging Slack and email, prepping for meetings, and keeping projects documented, my automations easily save me several hours a day.

How did you evaluate Claude models?

I've personally cross-tested every frontier model, and what separates Claude, and Claude Opus 5 specifically, is judgment. Claude is strongest in long-running, multi-step work where other models lose the thread, holding the full picture without dropping standards or details. Claude also challenges me when my logic is loose and flags when the strategy is off. It disagrees, it has taste, it knows when to push. It tells me what I need to hear, not just what I want to hear. That's what I want in a strategic partner.

Anthropic: Nicole, what first led you to try Claude Cowork?

Nicole Judson, Influencer Marketing Manager at Miro: For Miro's global event series, I managed nearly 90 influencers across three events with the goal of treating each one as a tentpole moment. That translated into having all influencer content live within 48 hours. The bottleneck was specific: the errors I'd catch were often small (capitalize "Sidekicks," use "Canvas 26" not "2026 Miro Canvas" to name the event), but multiply those across nearly 90 creators in a compressed timeline and you're spending an enormous amount of time on things that don't require creative judgment. I saw a LinkedIn post from a creator marketing leader using Claude Cowork to take a first pass on influencer scripts against a brief, and it clicked immediately. 

Anthropic: What specifically did you build?

Judson: The Influencer Content Reviewer became the backbone of our review process. An influencer would submit their script or caption, and I'd run it through as a first pass alongside the event brief: what they were supposed to cover, how to reference products, naming conventions, and what was off-limits before the announcement. The output cut review time in half. It was reliably clean on brand consistency and brief adherence.

Where I still stepped in was anything requiring creative judgment: does this sound authentic to the creator's voice, does the story land, and is the framing on-brand without being stiff? Those aren't things you can fully brief into a system, but the mechanical layer came back solid every time.

Anthropic: What did the results look like once the events wrapped?

Judson: At each event, I wasn't just behind a screen reviewing content. I was on the ground with influencers and managing logistics. Without Cowork, keeping up with review volume at that pace would have been a serious lift.

The numbers tell the story: at our Canvas 2025 event, we hit 64% of influencer content live within 48 hours. Across our three Canvas 2026 events, after implementing Cowork, we averaged 99%. Cowork was central to making review fast enough to hit that window.

Anthropic: Betty, what's a Claude Cowork workflow that's become part of your regular routine?

Betty Woods, Scaled Success Lead for Global Programs at Miro: I built a workflow to manage and track all of the different projects my team is working on. It used to be a highly manual process where I had to update a Miro table line by line. I was spending my Friday mornings acting as a historian, just documenting things that had already happened. It was an administrative drain that took over 90 minutes.

I started wondering if Cowork could read through all of my messages across channels to summarize that information for me. Because Cowork integrates directly with Miro, I took it a step further and automated it. Now a scheduled task runs every Friday morning: Cowork pulls context from my Slack DMs, emails, calendar, and chat sessions throughout the week and suggests updates for the table. I spend a few minutes reviewing the suggestions, and then it pushes the updates directly to the board.

Now that Cowork handles the documentation in about 10 minutes, I use the saved hours to draft briefs for the next phase of a project, identify our key wins, and proactively solve blockers. Cowork took over the busywork of documenting the past so I can focus entirely on strategizing for the future.

Anthropic: Beyond the automations, how do you actually work with Cowork day to day?

Woods: I use Cowork as a thought partner about 80% of the time. My typical input is just a brain dump: "Here's the problem I'm trying to solve, and here's what I'm thinking. What do you recommend?"

The magic is in the Cowork project I've set up, which acts as the main holder of "who Betty is." I've loaded in our entire team's strategy, how our work connects to specific OKRs, and my exact communication styles for Slack versus email versus LinkedIn. Because of that baseline, I don't have to constantly remind Claude why I care about talking to an IT leader or why an upcoming product launch matters. It even acts as my career coach: I've built skills that pull coaching themes and wins from my internal and external meetings to track for performance reviews. Because it has all that deep context, what comes back is usable right out of the gate. I only really spend time refining when I'm crafting specific messages.

Anthropic: As more people across Miro build their own workflows and connect more tools, how do you keep quality and trust high?

Hostetler: Once a whole team is building, you can't personally check every output, so the standards can't live in your head or in a review step at the end. They have to live in the skills themselves, so they travel with the work no matter who runs it. Our shared skills have become a critical piece of team infrastructure. Before a skill rolls out team-wide, it gets a quality pass: does it cite sources, does it respect data boundaries, does it conflict with another skill? Anything that makes a factual claim has to trace back to a real source. For my team, Claude is required to add a link to where each claim came from, whether that's a Slack message, a doc, or a webpage. Making everything checkable matters more the more people are building. A human still owns the final call on anything external.

Judson: The trust I have in the output comes from knowing I’ve built in a high level of specificity, and humans never leave the loop. Cowork handles the volume, but every high-stakes call still comes to me.

Anthropic: What patterns do you notice in how teams across Miro adopt AI?

Hostetler: What teams get right is starting with a real problem, usually the repetitive one that eats their day. That's where the payoff is obvious and adoption sticks.

What they get wrong tends to be one of two extremes. Some overbuild. It's tempting to pile on skills and automations, but the trap is not maintaining them: skills that contradict each other, ones so long or detailed you don't actually know what's in them. That clutter makes Claude less reliable, not more. Others go the opposite way and never really customize Claude to their team. They use it as a personal chatbot, without the shared skills, projects, scheduled tasks, and artifacts that make it powerful, so they only ever see a fraction of the value.

The best version starts by zooming out: you keep each piece clean and simple, then chain them, automating the thing that triggers the automation that triggers your deliverables.

Anthropic: That's a lot to build toward. What's the right starting point for a team new to Cowork?

Judson: Start with the task you dread most, not the one that sounds most impressive. For me, that was content review: high-volume, tedious, low creative reward (and that's exactly where Cowork delivered). The other thing I'd say: involve the people who will actually use the workflow in the setup. I recorded a Miro talktrack walking my team through the project and had them test it with mock content before we went live. Adoption happens when people see a tool solving a real problem from someone they trust. The pitfall to avoid is thinking you can set it and forget it. The output is only as good as the context you give it, and that takes some iteration to get right.

Woods: My number one piece of advice is to nominate a few champions within your team to lead the charge. Having dedicated people who stay on top of updates, filter down what the team actually needs to know, and act as a first line of defense for questions makes adopting a powerful new tool feel a lot less overwhelming.

And don't let a lack of control paralyze you. Early on, I was actually afraid to connect my Google Drive or Gmail because I was worried about accidentally deleting something or sending an email I didn't mean to. That fear definitely delayed some of the workflows I could have built sooner. Once I took the time to read into how the permissions were structured, those concerns completely vanished. You pick which folders and tools Claude can reach, and it can't touch anything outside that.