Collecting Keys - Real Estate Investing Podcast

RESimpli 6.0 Broke, So Build Your Own CRM With AI Instead

By Mike DeHaan, Dan Austin, Dylan Koch ·

Every real estate operation runs on rented software: a CRM, an e-sign tool, a loan origination platform. On this episode, Mike DeHaan, Dan Austin and Dylan Koch dig into what happens when one of those vendors breaks your business overnight, why the math on renting software has changed, and how a housing market with sales volume at a 31-year low is squeezing flippers.

What went wrong with the RESimpli 6.0 rollout?

RESimpli pushed users from 5.0 to 6.0 and, in Dylan's words, everything broke. Texting stopped working. E-sign stopped working. The base functions operators rely on daily, lead organization, KPIs, reporting and seller communication, were unreliable, and there was no option to roll back.

Dylan's read is that the rollout was rushed to bolt on AI features, because a software company priced as an AI company sells for a higher multiple. His complaint isn't the new features, it's shipping on a Saturday night without verifying the core worked. Mike pushed back slightly, noting six years of accumulated infrastructure and data structures make that migration genuinely hard, but agreed it's not an excuse.

Mike's alternative: build the new version as its own product, offer free migration, then put a ninety-day time gate on it. After that you're on your own. Nobody's system breaks all at once.

Dan and Mike also noted who eats the damage. Not the big team with a staffer who babysits the CRM. The solo operator working leads, doing the books and running overhead at the same time, which Mike believes is most of the user base.

Can you actually build your own CRM or e-sign tool with AI?

Mike's position is that breaking your users is now the dumbest move a SaaS company can make, because replacing them is realistic. He built his own loan origination system with Claude. To match the features he liked in the platform he was leaving, he told Claude to look at what that platform does and rebuild it, and it produced a near carbon copy. He's still adding features weekly.

The e-sign story is sharper. Mike wanted PandaDoc's API. The rep quoted $1,000 per person per year, which with 12 staff meant $12,000, plus $2 per document through the API. He built his own instead. It took about an hour, and he later added templates and roles.

His method for Dylan, who's considering building a lighter version of his CRM: host it on something like Netlify, point Claude at the platform's own tutorial videos, tell it to build that exact thing, and add "get as far as you can without asking me additional questions." It runs for hours and gets closer than you'd expect. Mike also mentioned that the owner of a lending software product he used sold the company because he saw AI making him a dinosaur.

Does a ChatGPT demand letter actually work?

Dan has been using ChatGPT to write demand letters to companies he believes are overcharging him. A pest control company charged $1,000 to handle rodents at one of his rentals, put up half-hearted chicken wire that left visible holes, and the animals got back in through the same spots. Buried in a 40-page document he'd skimmed on his phone as a quote was a recommended $100-per-month recurring service, which they then started billing. When he asked to cancel, they said he owed $1,000.

He had ChatGPT break down the contract and write a demand letter. He sent it that day and got an email from the owner canceling the account with nothing owed. Mike had his own version, filing through a small claims service against an agent who cold-texted him claiming he'd opted in. None of this is legal advice, and results depend on your contract and state, but the pattern both hosts describe is that vendors fold fast when someone reads the agreement back to them.

Why aren't houses selling right now?

Mike said his lending base rates jumped a full percentage point in two weeks. Dylan looked up conventional pricing before recording and found highly qualified borrowers quoted in the low sevens, meaning a typical low-down-payment or FHA buyer is likely closer to 7.75 plus mortgage insurance. Mike doesn't think most people can qualify for that payment and expects eights.

Dylan has three Spokane listings he believes are priced correctly and is barely getting showings. They'll still be profitable, just well below what he underwrote. He cited sales volume at a 31-year low, worse than 2008. The difference, Dan said, is equity. Owners aren't upside down on 100% financing loans, so there are no forced sellers. Mike called it a stalemate: inventory piles up, nobody looks, and nobody is motivated to cut price.

On the rental side, CapEx is stacking. Dylan was quoted $14,000 for a rubber roof and had a leak misdiagnosed as a roof problem when the real culprit was a furnace condensation drain line dumping into the attic. Dan put his CapEx spend at roughly a thousand bucks a unit. Mike's summary: houses cost money, and the longer you own them the more they cost.

The through-line here is control. Vendors break, contractors misdiagnose, and markets stop cooperating. Listen to the full episode for Mike's build process with Claude and more on where the hosts think rates go.

Frequently asked questions

Is it realistic to replace a CRM with AI-built software?

Mike built his own loan origination system and e-sign tool with Claude by telling it to replicate the platforms he was leaving. The hardest pieces are 10DLC texting, calling and anything that moves money.

How do you avoid overpaying on a repair diagnosis?

Dylan replaced a roof that didn't need replacing because the actual leak was a furnace condensation drain line in the attic. His takeaway was to do the diagnostic work upfront instead of trusting the first bid.

Where are rates and sales volume?

Dylan found highly qualified borrowers quoted in the low sevens, with typical buyers likely higher, and cited sales volume at a 31-year low.

AI & TechMarket Updates

Educational content from the Collecting Keys podcast. Not financial, legal or tax advice.