Claude Dashboard When Your CRM Is Stale: A Small-Business Lender's Google Sheet Fix
By Alex Regueiro · November 15, 2026 · 5 min read
An AI coworker is only as current as the data you point it at, and one small-business lending rep was pointing it at a CRM that had gone stale years earlier. The fix started with changing the source, not the AI.
60-second check
Take the 2-minute AI readiness survey →
Talk it through
Book a call with the Reshape team →
The person in this story is a sales rep at a small-business lender, and the project was a Claude coworker setup with a dashboard on top of it. If you lead or work on a lending team, our page for mortgage loan officers and lenders explains how we approach these builds.
The situation
The rep had a Claude coworker and a dashboard in the works. On paper, the data lived in the CRM. In practice, the rep had not updated it since joining, and the person who held the role before had stopped tracking there roughly four years earlier.
That left a system full of old deals and old statuses. The rep also kept a sales tracker in Google Sheets, and that sheet became the source.
What was broken
A dashboard built on the CRM would have looked finished and been wrong. Stale data does not announce itself. Every score and task list would have come from records nobody was maintaining.
Claude had to be switched off the CRM before anything else would hold up. The problem was not the dashboard idea. It was the source underneath it.
What we built
We rebuilt the setup from the data up.
- A new source: we configured Claude to pull from the Google Sheets tracker instead of the CRM.
- A scoreboard: we built a scoreboard dashboard on top of that sheet, organized around monthly scores and weekly tasks.
- A fixed rhythm: the rep confirmed that the month is the window that matters most for measuring success, and weekly tasks give clear direction on what to do next.
- Planned next: line-of-credit logic so funded amounts display correctly, and a map of where lifecycle and term data live so renewals can be tracked in one place.
What changed
No result had been reported at the time of the call, so the change is in the foundation. The scoreboard now reads from the sales tracker sheet instead of the CRM, and it answers two questions on a fixed rhythm: how the month is going and what to do this week.
The line-of-credit logic is not finished. It cannot be validated without real example deals from the rep, which were still needed after the call. Until then, funded amounts on those deals should be treated as unproven.
Test your source before you build a dashboard
Pick five records you know well and compare them to what your CRM shows today. If more than a couple are wrong, do not build on it yet. Fix the source first or point the dashboard somewhere else.
Then choose the system your team actually updates, even if it is a spreadsheet. A simple sheet that is current beats a CRM that is complete but abandoned.
Frequently asked questions
Can Claude build a dashboard from a Google Sheet?
Yes. In this build Claude was configured to read a Google Sheets tracker, and a scoreboard dashboard was built on top of it. The result is only as good as the sheet, so someone has to keep the sheet current.
How do I know if my CRM data is too stale to use?
Compare a handful of records you know well against what the CRM shows. If statuses, amounts or dates are out of date, anything built on that data will be wrong in the same way, and a dashboard will only make the problem look official.
Why use monthly scores and weekly tasks?
The rep said the month is the window that matters most for judging success, and weekly tasks turn that score into specific next steps. One view shows how you are doing and the other shows the work to do.
How should line-of-credit deals be handled in a dashboard?
Line-of-credit deals need their own logic so funded amounts display correctly. That logic was still to be written here, and it needs real example deals to check against before anyone relies on the numbers.
If your reports or AI tools are reading from a system nobody updates, book a 30-minute AI follow-up and reporting audit. We check which source your numbers really come from and whether it can be trusted.