Great AI Is a Connector

Agent Workflow3 min read
Judd Hoffman
Judd HoffmanCEO and Co-Founder, Ethica AI

The agent lives the deal.

They know it inside and out, because they were there for all of it. The calls. The negotiations. The texts at nine at night. The showings, the second showings, the conversation in the driveway that changed everything.

The broker eventually sees the file.

Why is that a gap worth closing?

Because those are two different versions of the same transaction.

One version is lived: full of context, judgment calls, and the reasons behind every decision. The other is documentary: what made it onto paper, in the order it got filed. The first version is richer, and it lives almost entirely in one person's head.

That works fine right up until somebody needs to understand the deal who wasn't in it. A broker reviewing the file. A compliance question months later. A teammate covering while the agent is on vacation. Then the reconstruction starts: scrolling back through texts, calling people, trying to remember why a date moved.

What does that reconstruction actually cost?

The agent's time, first, and always at the worst moment.

But also accuracy. Memory fades, threads get lost, and the version that gets rebuilt is thinner than the one that happened. The details that made the deal make sense are the ones most likely to go missing, which is exactly why what a system remembers matters.

Nobody set out to lose that context. It's just that the lived version and the filed version were never connected in the first place.

So what should AI actually do here?

Connect the two. Not replace either one.

The agent keeps living the deal, because that's the job and that's where the judgment is. The broker keeps reviewing the file, because that's oversight and it matters. What changes is that the record builds itself along the way, from the work as it actually happens, so the file already reflects the deal rather than being a summary written afterward from memory.

That's the same coordination problem I described across the whole transaction team, narrowed to its most common seam: the person who was there, and the person who reads about it later.

What does good look like?

Nobody has to recreate the transaction later. The context is already where the next person needs it.

That's the play.

The short version

Quick takes

  • Why is there a gap between what an agent knows and what a broker sees?

    The agent lives the deal through calls, negotiations, texts, and showings, so the richest version stays in their head. The broker sees the file, which is what made it onto paper. Those are two different versions of the same transaction.

  • What does recreating a transaction after the fact cost?

    The agent's time, usually at the worst moment, plus accuracy: memory fades and threads get lost, so the reconstructed version is thinner than what actually happened.

  • How should AI connect the agent's view and the broker's view?

    By building the record along the way from the work as it happens, so the file already reflects the deal instead of being summarized afterward from memory. The agent keeps the judgment; the broker keeps the oversight.

Sources

  1. Transactions Are a Team Sport: Judd's earlier episode on coordination across the parties in a deal, the wider version of the agent-to-broker gap this episode describes.
  2. What Should Real Estate AI Remember?: On the memory that keeps a transaction's history intact so nobody has to reconstruct it after the fact.

Transcript

56 words

The agent lives the deal and knows the deal inside and out. They've had the calls, they've done the negotiations, the text, the showings, all of it. The broker eventually sees the file. Great AI should connect those two things so nobody has to recreate the transaction later. Great AI is a connector. That's the play.

Keep reading.

All posts →
  • A Formula One style pit stop mid-service in warm cream and sage green: a crew of dark green silhouetted figures swarming a race car in choreographed positions, jack operator, tire carriers, and fuel man each doing one job at the same instant, while the lollipop man stands out front holding the timing for all of them. Speed comes from coordination, not from adding people.
    Agent WorkflowAugust 17, 20263 min read

    Transactions Are a Team Sport

    Judd Hoffman on why the hard part of a real estate deal was never the number of people involved. It's keeping all of them moving together.

    Judd Hoffman
  • An original illustration of a real estate professional standing in a warm office, mid-shrug with a puzzled, searching expression, holding a half-open folder of papers, wondering where they left off after an interruption. Rendered in muted warm cream against deep forest green with the figure catching accent sage green. The confusion of returning to a transaction the software has forgotten, having to start over.
    Agent WorkflowAugust 6, 20265 min read

    What Should Real Estate AI Remember?

    Judd Hoffman on transaction memory: a real AI transaction assistant should remember where the agent left off, preserve the deal points and context, and flag what is still missing, so interruptions do not force the agent to start over.

    Judd Hoffman
  • A relief pitcher walking in from the bullpen gate across a sunlit outfield, rendered as a deep forest green silhouette with glove in hand, the packed stadium stands in soft sage and the infield waiting ahead. Called in exactly when the game is on the line.
    Agent WorkflowAugust 18, 20263 min read

    Show Up When It Counts

    Judd Hoffman on why an impressive AI demo proves nothing. The test is the hard day: deadlines moving, details slipping, nobody calling back.

    Judd Hoffman

Your next deal, done.

One call gets you started, whichever way you like to work.

New to Ethica? Activate in 2 minutes

Or call Ethica (714) 909-2027