Cardinal's data team was already further along than most
Cardinal Group manages multifamily and student housing communities from Denver, and its data team is further along than most in the industry. It runs its own BigQuery warehouse, its own business intelligence, machine-learning models and AI agents that prompt on-site teams. Behind all of that sits the operational stack every property manager runs, a long list of point systems, each with its own API, and each holding history the team wants to keep even as the large PMS vendors consolidate.
Every vendor pull was custom work, and teams grow just to keep up
Most of those APIs, she notes, are event-based, not batch, and every one is custom development for a data team. The pulls were costing the team the thing it was built for.
“Building out all those pulls is not adding any value. What's adding value is having the data.”
She has watched where that leads in data teams that build everything themselves.
“They figure out a way to do it, but because the way they've done it is so piecemeal … they end up only growing their team for the sake of maintenance and not true new build.”
“All of a sudden your scale of maintenance and overhead versus new innovation starts to slip. And nobody thinks about that.”
The easy advice does not survive contact with an enterprise shop.
“When you're running an enterprise shop, people are like, oh yeah, just connect to the API and use AI. That's not how that works when you're trying to do something with enterprise-grade integrity.”
Rachel's rule is to build for differentiation and buy the rest
Her rule for what the team builds itself is short.
“Build the things that add business value and are a differentiator. Buy the things that speed you up to getting to that differentiator. And do a true assessment of ROI.”
“Build for differentiation, or where it's truly cost-effective. Buy when there is no differentiation.”
By that test, integrations with the point systems around the business are a buy. Cardinal chose Propexo Connect for them, starting with its inspection platform into BigQuery.
The first integration was not turnkey. Data arrived that did not reconcile, and the question became whether the numbers could be trusted. Cardinal's team validated against its own SQL while Propexo worked the discrepancies with the vendor.
“Your team didn't just say, okay, the pipelines are hooked up … You weren't just committed to hooking up the pipelines, you were committed to the integrity of what was coming through. I cannot emphasize how much work that took off of our plate.”
The first integration proved the value was in the integrity, not the connection
What came out of the episode was inspection data in BigQuery that Cardinal could trust, four times what the PMS integration had exposed, and, in her account, a different kind of relationship.
“The value of the data connection isn't just the connection. It's the integrity. And that was committed to being partners with us.”
The ETL work went with it.
“You guys have taken away a lot of that data engineering ETL work.”
The integration now feeds compliance reporting and the AI agents
Inspection and checklist data from the platform lands in BigQuery beside Cardinal's maintenance tickets and resident sentiment, and compliance reporting runs across the portfolio from it. The same data feeds the AI agents Cardinal built to prompt on-site teams toward what needs doing, and she is deliberate about where the agent stops.
“AI is not going to go do a unit inspection … A human literally has to, and we don't want to try.”
The inspection platform was the start. Cardinal's team is now ramping more of its integrations through Propexo, and the pattern holds for each one: the data lands in BigQuery and goes to work.
Grow the team for new work, not to keep up
Her hiring rule is the other half of the argument. A team that buys its pulls does not have to add a person every time it adds a system, and the people it does add go to new work.
“I so often see people that are like, oh, I need another person just to keep up. That is a terrible reason to get a new person. You did something wrong to begin with, if that's why you need a new person.”
The maintenance nobody plans for is where that difference shows up.
“The maintenance thing is a real thing. Everyone's trying to move so fast, they don't know what's coming for them 18 months from now.”
If the next hire on your data team is for keeping up, start where Cardinal did, with the pulls from your operating tools into the warehouse you already run. Talk to us about which ones.