Why the call is coming out
I mean, imagine building this state of the art automated machine, right? You run it for six months and then you realize it made 29 phone calls and generated exactly zero decisions. Yeah, that is a tough pill to swallow for sure. It really is. So welcome to today’s Deep Dive. For you listening, and I am specifically talking to you, the CEO and the head of sales, our mission today is to completely unpack the executive briefing on the new renewal strategy for Western Slope. Right, because we are completely scrapping the old playbook. Exactly. The biggest takeaway here is that our underlying engine, it’s actually incredibly sound. We just realized we had built the original plan for the completely wrong type of customer. Yeah, we took a playbook designed for Camellia, which had a single building and a staff that physically sat on site, and we tried to apply it to Western Slope. And they are totally different. Completely different. Western Slope operates 70 scattered homes managed by a centralized two person team. So obviously you can’t use the same communication or software strategy for two totally different operational realities. No, you really can’t. And I think to understand the new strategy, we first have to look at why the old one failed so hard, which means talking about the automated opening phone call. Right. The Camellia cohort data on those calls is just brutal. Because if you’re a sales leader, you rely on outreach metrics to forecast your pipeline, right? So walk us through the raw numbers from that Camellia cohort. OK, so out of 29 automated calls placed by the system, only two actually reached a resident. Wait, two out of 29? Two. And out of those two, exactly zero produced a renewal decision. Wow. OK, so I want to focus on the 27 calls that didn’t connect, because I feel like that breakdown tells the real story about the friction here. It does. So 16 of them hit voicemail immediately. Five hit those call screening robots, the ones that are getting super common on smartphones Oh yeah, I have that on my phone. Exactly. And the last two actually reached a human, but it was the wrong person entirely. So the phone channel is just functionally dead space for this type of outreach. Dead space. But honestly, there is a much more alarming data point for the executive team here. It involves the system’s internal tracking. Oh, right. The connected label. Yes. The system’s connected label, which is the exact metric feeding your dashboards to tell you a call, was successful. It was wrong. 10 out of 12 times. See, that metric failure is terrifying from a forecasting perspective. I mean, relying on that connected label, it’s basically like a delivery driver marking a package delivered when they really just threw it in a ditch on the side of the highway. That is exactly what it’s like. It creates this phantom pipeline. Yeah, it gives you this completely false sense of security that a resident has been contacted when it’s actually a total failure. Your dashboard is basically lying to you. It is. You think you have deals in motion, but you don’t. Meanwhile, the only channel that actually performed was text messaging. Right, because text actually gets read. Exactly. Texting had a 20% reply rate. It actually produced the decisions. But you know, let me just challenge this total abandonment of the phone line for a second. Let’s look at those 16 voicemails. Weren’t they at least doing some heavy lifting, like serving as a billboard, leaving a message that prompts the resident to go check their text? You would think so, but the timing of the workflow proves they weren’t. The system was actually configured to send a text right before the automated call. Oh, I see. Yeah, so the automated voicemail was literally just promising a text message that the resident already had sitting on their phone. So it’s totally redundant. Completely. It added no new information, created zero urgency, and honestly, it probably just annoyed them. Yeah, I’d be annoyed. Okay, so the data totally dictates that we cut the automated opening call. We need a new channel strategy. What does the MVP protocol look like moving forward?
Text first, email second
So the minimum viable product for Western Slope is text first, followed by an email on the exact same day, and absolutely no opening call. Got it. And what about voice entirely? Voice will now only be used for inbound calls, like if a resident actually dials the line themselves or if they specifically opt in later down the funnel. What about that branded caller ID line we talked about a while ago? That’s being shelved for now. It might be a phase two experiment, but it is firmly not part of this MVP. Okay, but let me ask you this. If texting has a 20% reply rate and produces the decisions, why bother adding email on the exact same day? Aren’t we just creating inbox clutter? Well, because a lease renewal isn’t just an impulse purchase, right? It’s a household contract. Text is unmatched for getting their immediate attention, but email serves a completely different functional purpose. Like having a paper trail. Exactly. It gives the resident a formal, static document. When someone is renewing a lease, they almost always need to discuss it with a partner or forward it to their roommates. That makes a lot of sense. Right. A text is like a tap on the shoulder, but the email provides the actual utility to forward the terms and say, hey, are we signing this? And crucially, sending that email costs us nothing. Right. So you’re treating the resident not just as a consumer, but as a stakeholder who has to run their own internal approval process. Exactly. Okay. So we’ve defined the outbound communication. Getting Western Slopes centralized team actually ready to process these renewals, that’s a totally different hurdle.
Six hurdles before Western Slope
We have a ranked list of six operational blockers here. Yes. And these blockers exist entirely because we are moving to a centralized environment. So what’s blocker number one? Blocker number one is getting AppFolio write access. Right now, our system only has read only access to Western Slopes database. Which means we can look at the data, but we can’t touch it. I mean, for a sales op, that’s a massive bottleneck. Someone still has to manually type the final data in. Exactly. You can’t have an automated renewal engine if it doesn’t have the keys to write the contract to the database. Makes sense. Blocker two. Blocker two deals with visibility. We have to force their adoption of the actual renewals module within AppFolio. What are they using right now? Currently, they use this messy workaround where they combine the standard leasing flow with manual amendments just to handle a renewal. So if I’m the head of sales looking for my quarterly renewal projections, what does that workaround actually do to my reporting? It totally blinds you. Because they aren’t using the dedicated module, their native renewal reports are basically empty. The data is just scattered across leasing flows. We cannot turn our automated system on until they’re working inside the proper module. Otherwise the metrics will just never sync up. Okay. Blocker number three. This is a pure scalability issue, right? Moving settings to the company level. Yeah. So Western Slope has 70 separate properties. Right now, our software forces configuration at the property level. So they’d have to set up everything 70 times. Yep. Setting up enrollment toggles, time zones, policies 70 separate times. We are building a company tier so we don’t have to manage 70 disparate configs just to launch one customer. Wow. The operational drag on that would be insane. Which leads perfectly into blocker number four and this is about Kat and Bev, right? Yes. Kat and Bev are the two-person centralized team running this entire 70 property portfolio. We absolutely have to give them one all properties queue instead of 70 separate boards. Right. Because right now our software’s board forces you to look at properties individually. I mean, imagine your sales reps having to log out and log back into 70 different CRM dashboards every single morning just to figure out who to call. It would be a nightmare. They’d be playing 70 simultaneous games of chess just to do their daily jobs. One all properties queue is non-negotiable further sanity. Totally non-negotiable. Which brings us to blocker number five, addressing the communication loop with that team. We need to build escalations that actually chase. Chase. What do you mean by that? Well, the old playbook for Camellia assumed the onsite team was just too busy with foot traffic to answer system prompts so the workflow just fired off an email and moved on. But Western Slope’s centralized team, they actually do answer. They are at their desks. Ah. So we need an escalation sequence that waits, chases them for a response, and keeps the deal moving. Exactly. And the final blocker, number six, is treating month to month as a real product decision, not just an administrative afterthought. And this actually brings us to the actual revenue volume for Western Slope. Right. Let’s talk deal volume. Size per household. Western Slope has exactly 117 lease ends over the next 12 months. Size per household. Okay. 117. If you are forecasting pipeline, that is your total addressable market for this client over the next year. Yes. And the distribution is highly specific. Only 18 of those lease ends are in 2026. The real stress test comes next year, peaking at 38 lease ends concentrated entirely in May and June of 2027. Okay. So we have the communication plan. We have the six structural blockers and we know the exact volume. How are we staging this rollout so we don’t accidentally break their operations? It is a strictly controlled four-phase rollout.
Four phases
Phase zero is happening live this week. These are the immediate fixes to bugs and broken elements. Basically clearing the runway. Exactly. Then phase one happens next week. We launch the MVP at Western Slope. However, it stays dark. Yeah, it runs in the background, processing the logic. Yeah. But it remains entirely invisible to the residents until Fede gives the final go. Now, wait, if the software works and we have those 18 lease ends approaching in 2026, why keep the MVP dark? Why not just turn it on and let it generate revenue? Because of the sweeping changes we just talked about. The company level settings, the unified queue, the risk of misconfiguration is just too high right now. Oh, I see. Yeah. If a logic error sends a completely wrong lease offer to a live resident, we don’t just lose a deal. We actually damage Western Slope’s brand reputation. We need absolute certainty in the staging environment before Fede throws the switch on live outreach. It’s like you want the engine running flawlessly on the track before you put actual passengers in the car. Exactly. And once Fede gives that green light, we enter phase two. This spans weeks two through six. This is our dedicated learning period with Western Slope via a standing Wednesday call with Kat, Bev, and Jason. So we aren’t guessing what they need. No, we learn directly from them. And then phase three is scale. We take everything we’ve built and refined and we roll it out to our second customer, Situs Group. Situs Group. Nice. So they are the fast follower. Let’s reverse back to that phase two learning period for a second because those Wednesday calls with Kat, Bev, and Jason seem like the crucible for this entire strategy. Oh, they are.
Design-partner questions
As a design partner, we need very specific answers from them on that very first Wednesday call. What are the questions we need answered? Okay, so first, what are their standard term options for the offer template? Second, who exactly has the authority to approve a counteroffer? Third, what is their exact month to month premium and required notice period? Fourth, how do they physically want the final offer delivered to the resident? Okay, and what’s the last word? The most important one. What does done actually mean to them? Oh, I love that. For anyone managing software rollouts, asking a client to define what done means is like the ultimate project management hack. It really is. Because if you don’t define it together, you’ll build a technically perfect system that totally fails their subjective expectations. Like, we might think a renewal is done when the resident texts back, yes, but Kat and Bev might think it’s only done when a countersigned PDF is sitting in a specific folder inside AppFolio. Exactly. Which is actually a perfect segue into the underlying technology powering all of this. We can redesign workflows all day, but is the engine actually ready for this?
Temporal verdict: solid
Let’s look at the temporal workflow engine. Right, the temporal verdict. What is the actual health of the system under the hood? The verdict is solid. Right now, there are 16 running renewals in the system. There are zero stuck activities. And critically for the engineering side, there is zero non-determinism. Okay, pause. Zero non-determinism. That is very heavy engineering jargon. Translate that for the executive team. Why does a head of sales actually care about non-determinism? Fair point. In short, it means the software isn’t acting unpredictably. If A happens, B happens every single time. It is perfectly predictable. For sales forecasting, that means your conversion metrics aren’t being skewed by random software behavior or logic branching off in weird directions. Oh, got it. So it’s like having a perfectly functioning sports car engine, right? Where the dashboard warning lights were just unplugged, the engine was always firing correctly. We just couldn’t see the dashboard properly? Yes. The gap we had was purely about observability alerts that were turned off. It wasn’t about correctness. The logic was sound. And that observability gap has already been fixed in phase zero. Okay, that is a huge relief. But how do we actually prove the engine is sound? I know we audited the data, but how?
Fable vs Opus: the two-model comparison
We executed a two-model AI comparison. We pitted two independent AI models against each other. We did. Fable and Opus. They independently inspected the system, and then their findings were reconciled against live production data. And what did that reconciliation actually catch? It caught some vital discrepancies. It confirmed the true call count was exactly 29. But more importantly, it caught the true 12-month lease and volume for Western Slope. It verified the exact number is 117 because the AI reconciled the data by sizing per household, not per person. That is massive. If you are projecting Q2 2027 revenue, sizing by household versus individual is literally the difference between hitting your quota and missing it entirely because Western Slope has multiple adults in these homes. Exactly. If we size by person, your pipeline would look artificially huge and the system would end up double messaging residents living in the exact same house. Which would just be confusing and annoying. So the AI reconciliation locked in the financial baseline. That’s great. But what happened when we actually stress tested the code in the lab?
Lab results
The lab results are super interesting. We replay the Camellia renewals through today’s code and we found two real bugs that are being fixed live right now under the bug fix rule. Oh, what were the bugs? The first one was an offer prepare timeout that had drifted. And because that window drifted, it was silently misclassifying the outcomes of certain renewals. Oof. Silent errors are total pipeline killers because they don’t trigger any alarms. And what was the second bug? The second bug was a workflow that goes completely quiet forever when a resident never replies. Wait, really? So if a resident just ghosts us, ignores every text and email, the workflow just quietly bathe up and recorded nothing. It just stopped. It doesn’t flag the file for Kat and Bev to follow up manually. No. And that is exactly why the lab replay exists. Getting zero crashes across everything we replayed is fantastic, sure. But finding silent logic failures like that where a lease just vanishes, that is what actually protects our centralized teams in production. Is there any limit to what the lab can test right now? There is one hard limit. There’s no full year of Camellia history to actually replay. The oldest renewal record in the system only goes back to 2026-04-30. So we are testing against a relatively short historical corpus. But even within that window, the structural integrity of the code held up. Zero crashes. So let’s bring this all together for the listener. Sitting where you sit as the CEO or the head of sales, this briefing is basically a massive validation of your tech investments, right? PropFlow, Clara, Temporal, and the overarching engine, they are technically sound. Yes, the underlying engineering was correct all along. We do not have a broken machine. The only thing we had to fix was dropping an ineffective playbook. We dropped the automated phone calls that generated zero decisions, and we are adapting our software configs to support a centralized team instead of an unresponsive on-site team. We simply had to tune the software to match the reality of how our new customer actually does business. Exactly. And now you have the exact volume metrics, those 117 households, the four rollout phases, and the operational blockers fully mapped out for your forecasting. And, you know, I will leave you with one final provocative thought as you plan your strategic roadmap for next year. Let’s hear it. We know the phone channel is dead space. Once Western Slope and Situs Group proved that a frictionless text and email first renewal process works flawlessly at a centralized level, how quickly will residents across the entire real estate industry start viewing traditional manual phone call renewals as an unacceptable nuisance? Oh, wow. Yeah, the standard is totally shifting. And this pivot puts you right ahead of it. Bring these insights into your next strategic meeting, and we will catch you on the next deep dive.