What to automate first, and what to leave alone
Automate whatever you are currently remembering. Recall is the most expensive thing a person does badly and the one thing software does perfectly.
The automation conversation in real estate is dominated by the most demoable things: social scheduling, listing videos, drip campaigns. Those are the easiest to show and among the least consequential.
A better priority rule: automate whatever you are currently remembering.
Why recall is the right target
Consider what a coordinator carrying fifteen files is actually doing. There are perhaps a hundred and fifty live obligations. Their job includes knowing, at any moment, which of those is closest to hurting someone.
That is not a skill that improves much with effort, and it is not a task humans are good at. It degrades with tiredness, interruption and volume, and it fails suddenly rather than gradually. The failure mode is a dropped item, and one dropped item looks exactly like excellent performance until the day it surfaces.
It is also the thing software does perfectly and cheaply.
Automate the tasks where a person's failure is invisible until it is expensive. That is recall, and almost nothing else on the usual list qualifies.
The priority order
1. Deadline extraction from the ratified contract. The highest-leverage automation available. It converts the single most consequential manual pass in the business, done under time pressure, into something that happens the moment the contract lands.
2. Recalculation when a date moves. Almost more valuable than the first, because this is where manual systems fail silently. Settlement moves four days and every dependent date is wrong until someone notices.
3. Escalating reminders on outstanding items. With elapsed time visible, so chasing happens on a schedule instead of on a feeling. Escalation-by-feeling produces both nagging and neglect.
4. Jurisdiction-specific document requirements. Given county, transaction type, and a few property facts, the required set. This is where conditional items stop being silently omitted.
5. Status visibility for parties. Most update emails exist to answer a question the client should never have had to ask.
6. Marketing. Genuinely useful. Sixth.
What to leave alone
Judgment. Whether to invoke a contingency, whether an appraisal dispute is worth pursuing, whether this buyer should waive inspection to compete. These are not data problems and they are not close to being data problems.
Anything irreversible, without a human in the loop. Generating a document is reversible and should run automatically; a file idling in an approval queue while deadlines run is a worse outcome, not a safer one. Sending something to a client, a lender, or a counterparty is not reversible, and that is where the human decision belongs.
The difficult conversation. A drafted message is fine. An automatically sent one, when a client is upset, is a way to lose them politely.
The test before automating anything
Two questions:
"If this runs wrong for a month before anyone notices, what does it cost?" High cost means automate it and instrument it heavily. Low cost means automate it and stop thinking about it.
"Can this be taken back?" If not, a person decides.
Those two questions sort almost every candidate correctly, and they explain why deadline tracking belongs at the top of the list and outbound sending never belongs on it at all.
The mistake worth naming
Buying marketing automation first because it demos well, then wondering why the business still feels chaotic.
The chaos is not in the marketing. It is in the fifteen files whose state lives in one person's head, and no amount of scheduled social content touches that.