Claude Gets Us Answers Faster. Without a Foundation, They're the Wrong Answers.

Everyone wants Claude (Anthropic) agents in their GTM stack. But what happens when the foundation underneath them is a mess?

We're finding out firsthand.

We've connected Claude to HubSpot. Different executives ask questions and come back with different revenue numbers, metrics, and conclusions.

Instead of debating what to do, we end up debating what the facts actually are.

And we're discovering why.

We have four different fields that can represent deal revenue. Old fields that should no longer be used are still being pulled into reports. Duplicate accounts have different AEs assigned to them.

Worse, our Claude integration had write access to HubSpot when it should have been read-only. People were unintentionally changing CRM data through Claude without the access restrictions we expected.

I was thinking about all of this after my conversation with Cathy Highley, an enterprise architect at ADM, for this week's episode of The Winding Road Show.

When Cathy arrived at ADM, she was surprised by the lack of common code libraries and reusable integrations. But ADM had grown through acquisitions. Creating consistency across those systems was no small task.

She spent years helping build the principles, standards, and guardrails that made the organization more effective.

Hearing Cathy talk about the importance of that foundation got me thinking about our own Claude efforts in GTM.

We spend so much time thinking about what Claude can do that we underestimate the work required to make it trustworthy.

Clean data. Common definitions. Clear ownership. Proper permissions.

In our case, Claude actually made things worse. We spent more time arguing about what should have been facts than making decisions based on them.

Claude can get us answers faster. But without a foundation we trust, we're just getting to the wrong answers faster.

My full conversation with Cathy is on The Winding Road Show.