Your CSMs Don’t Need More Tools. They Need a Copilot.

Ask a CSM what they'd do with ten extra hours a week and you'll get some version of the same answer: spend more time with customers.

Then watch how their week actually goes. Logging call notes in the CRM. Updating success plans nobody reads until renewal. Copying data from one system into another. Chasing signatures on renewal paperwork. Rebuilding the same deck for the fifth time this quarter.

We hire CSMs for judgment, empathy and commercial sense, and then we bury them in admin. For years the answer has been another tool. The stack grows, the admin grows with it, and customer time shrinks.

What we built instead

On my current team we went the other way. Instead of adding a tool, we built an AI hub that works alongside each CSM across the whole customer journey. Less another dashboard, more a copilot sitting in the passenger seat.

It connects to our CS platform and our CRM, so it works where the data already lives. Day to day, it logs calls and pulls out the key points, risks and next steps. It keeps success plans current as an account changes. It spots at-risk signals and kicks off the right save plan or playbook. It surfaces customer-specific insights a CSM can bring into the next conversation, which is really about telling stories with data. And it drafts renewal and upsell paperwork so a commercial conversation doesn't stall on admin.

None of that replaces the CSM. It takes the work that pulls them away from customers.

What changed

I can't point to one tidy metric, and I'm wary of anyone who says they can. What I saw was CSMs taking on larger books of business without drowning, and a noticeable improvement in how the team felt about their workload. The time that came back went where it should, into customer conversations. If covering more accounts is your goal, I've also written about scaling CS with AI while keeping it human.

There was a quieter benefit too. When the system handles the admin consistently, the data gets better. Call notes are complete, success plans are current, and health signals are accurate. That made forecasting and coaching sharper, because we weren't relying on whoever remembered to update the CRM on Friday afternoon. It also made predicting risk from the data much more realistic.

Where I draw the line

The most useful thing we worked out was where automation should stop.

Automate the record, not the relationship. Anything about capturing, organizing or moving information is fair game: notes, updates, reminders, paperwork, first drafts.

For risk, expansion and strategy, the AI's job is to put the right information in front of the CSM at the right moment. A flag that an executive sponsor has gone quiet. A usage trend worth raising. The CSM decides what to do with it.

And some moments stay human, full stop. The call where a customer admits their project is failing. The conversation with an executive about next year's priorities. Telling a customer something they don't want to hear. That's the job, and I've written more about why AI should be a partner here, not a replacement.

If you want to try this

You don't need a transformation program. Have your CSMs track a normal week and see where the hours go; the admin load is almost always bigger than leaders assume. Automate the most repetitive, lowest-judgment task first, which is usually call logging, because early wins buy trust. Connect to the systems the team already uses, since anything that needs a new tab won't get adopted. Bring CSMs into the design, because they know exactly where the friction is. And measure how the team feels, not just what it produces. If CSMs don't feel the difference, it isn't working. For smaller first steps, here are a few ways to start with the tools you already have.

The CS leaders who do well over the next few years won't be the ones with the biggest tech stack. They'll be the ones who gave every CSM a copilot and got out of the way.

If you're thinking about building something like this, I'm glad to share what worked and what didn't.

Previous
Previous

Stop Running QBRs Like Reporting Sessions

Next
Next

GRR Is the Metric Your Board Should Actually Care About