← All Issues
My Techie Broker

Issue 004 · June 15, 2026

If you're new here — Avance is the school where Arizona agents learn what AI tools actually save time, what's snake oil, and how to build automations that run without you. Three ADRE-approved CE courses online. Plus our co-director Danae Marquez teaches a hands-on automation class in person throughout the Valley.

Free 90-second demo of Ada (our AI instructor) at avanceschool.com/meet-ada. No signup.

Every Monday we send this — what's changing in your practice, what's actually working in AI, and one workflow you can build today. Replies welcome.

Now, this week's issue.


ShowingTime is back in Flexmls. And it's now your broker's choice, not yours.

On June 10, ARMLS rolled out a long-awaited change: ShowingTime and Aligned Showings are now both available as integrated options inside Flexmls, with appointments syncing between them.

The catch: your designated broker picks one as the default for your brokerage. Once that pick is made, you may need to update your active listings to align with your broker's selection — or your showing requests won't route the way buyers and showing agents expect.

This is the practical end of a fight that started in December 2023, when Zillow sued ARMLS and Metro MLS for removing ShowingTime in favor of Aligned Showings. The two sides settled in June 2024. Two years later, the operational compromise is finally live.

Three things to do this week if you haven't already:

  • Find out which platform your designated broker selected as the default.
  • Audit your active listings to confirm they're configured to that platform.
  • Get familiar with both interfaces — buyers will book through whichever they prefer.

ARMLS said they'd send step-by-step instructions to brokers on launch day. If your broker hasn't forwarded those yet, that's your prompt to ask.


Foreclosures aren't exploding. Stop watching the videos that say they are.

Mike Orr at Cromford Report came out swinging on June 6: "If you see videos on YouTube claiming that foreclosures are exploding, please don't click on them. You would just be encouraging their nonsense."

His read on the actual Maricopa County residential data — for trustee sales, which is the legally correct term most agents call foreclosures — is that we're still roughly 50% below normal.

Here's what the numbers actually show. Current monthly pace: 400-500 Notices of Trustee Sale, 100-150 Trustee Deeds completed, around 300 notices that get cancelled before going through. Compare that to 2004, the last "normal" year: 1,000 notices per month and 400 trustee deeds. We're nowhere close.

Yes, numbers are creeping up between 2024 and 2026. Yes, the trend line is real. No, it's not a crash signal.

Adjust for population growth and the picture gets clearer. Maricopa County has gone from roughly 3.5 million people in 2004 to 4.7 million today. That's a 34% bigger homeowner base before raw foreclosure counts even start to feel proportional to historical norms.

Two practical takeaways for your client conversations:

  • Stop forwarding the doom clips.
  • When a buyer asks if now is a "crash moment," you have the actual Maricopa numbers to ground the answer.

When the tech stack isn't the problem.

Jon Brooks, co-founder of a 280-agent Florida independent brokerage, published an essay on June 12 that's worth your five minutes. His argument: most agents who join a big brokerage for the tech stack discover that the tech wasn't actually their problem.

His specific observation, after talking with agents who've left big firms: "Half of them barely touched the technology they were paying for, and it never moved their production anyway. What frustrated them was responsiveness, access and whether anyone at the top really knew their market."

That maps to something we've been hammering in this publication. A tool isn't a solution. An automation isn't a substitute for judgment. And — Brooks's addition — a tech stack isn't a substitute for a broker who picks up the phone.

The cost of operating without local leadership doesn't show up in the tech budget. It shows up in deals that stall because nobody senior knew the answer, transactions that get harder because regional contacts have never closed in your market, and agents who feel alone inside companies with thousands of names on the roster.

Worth asking yourself, regardless of your current brokerage: when's the last time someone in leadership at your firm actually answered a question that wasn't on a script? If you can't remember, that's the data point.


What separates a strong BINSR prompt from a weak one.

Before we go further: every AI-drafted BINSR response is a starting point, not a send-ready document. Your designated broker still reviews structural and material items before anything goes to the seller.

Most agents who try to use AI for BINSR response drafting build something approximately three sentences long: "Write a BINSR response for the following inspection items..." Then they paste the list and ship whatever comes back. That's a weak prompt. It generates generic language, doesn't account for buyer priorities, and ignores the statutory framework that makes a BINSR response defensible if the deal goes sideways.

A strong BINSR prompt — one that produces output you can actually use, on both the AAR Buyer's Inspection Notice and Seller's Response and the MLS Choice equivalents — has six structural elements. If you're going to build one yourself, here's what it needs to do:

1. Force three variants, not one. A single-variant prompt lets you draft on autopilot. A prompt that demands full ask / middle ground / walk-away threshold forces you to decide where your buyer actually stands before drafting language.

2. Tag each variant with a likelihood estimate. Low / medium / high probability the seller accepts. AI can't predict for certain, but forcing it to estimate makes the variants comparable and reveals which items are realistic to push on.

3. Distinguish cosmetic items from material adverse facts. Cosmetic, paint, minor repairs, moderate items where buyer preference is clear — AI handles these fine. Structural defects, electrical concerns, plumbing-major, foundation issues, or anything that could constitute a material adverse fact — those trigger broker callback, not AI drafting. The prompt must make this distinction explicit.

4. Include suggested form language, not just strategy. A strong prompt outputs language that drops into the form with minimal editing. A weak prompt outputs general advice you still have to translate into form-ready text.

5. Acknowledge both AAR and MLS Choice form contexts. The AAR Buyer's Inspection Notice and Seller's Response is the form REALTORS use. MLS Choice licensees use comparable forms with the same underlying statutory framework. A strong prompt produces output that works for both.

6. End with a one-sentence rationale per variant. When you present three options to your buyer, you need to be able to explain why each one is built the way it is. Forcing the AI to articulate the reasoning surfaces the logic and lets you decide which framing matches what you'd actually argue.

A prompt that does these six things produces BINSR responses that hold up under broker review. A prompt that skips any of them produces output you'll end up rewriting from scratch.

Limits and broker callback. AI handles drafting structure and language. AI does not handle judgment on whether to walk from a deal, how to read the seller's emotional posture, or whether a material adverse fact has been disclosed adequately on the SPDS. Those decisions are yours, and your designated broker's. A.R.S. § 32-2153 sets out multiple disciplinary grounds for misrepresentation, misleading dealings, and unfair conduct — the liability stays with you.


Turn on Lockdown Mode in ChatGPT before your next client document upload.

OpenAI recently rolled out Lockdown Mode to ChatGPT accounts — an optional security setting that limits how ChatGPT interacts with external services, designed to reduce the risk of data leakage from a technique called prompt injection (where someone hides malicious instructions in a webpage, document, or file your AI ends up reading).

OpenAI explicitly notes Lockdown Mode isn't for everyone — it disables or limits some useful features in exchange for tighter security. But if you use ChatGPT to summarize client contracts, draft buyer/seller communications, or analyze financial documents — turn it on. Agents handle exactly the kind of sensitive data the feature was designed to protect.

To enable: Settings → Security → Advanced security → Lockdown Mode toggle.

While you're there, check active sessions — OpenAI now shows you every device with your account logged in, and you can close any you don't recognize.

A small feature that keeps client data where it belongs.


One question before you go.

What's one tech tool, platform, or workflow you've adopted in the last six months that actually saved you time? Hit reply with a sentence — we'll feature the best ones in a future issue.

Josh Marquez
Designated Broker, HomePros
Director, Avance School of Real Estate (S22-0004)


Sources referenced in this issue