Steven Song

Founder, Diald — AI underwriting for CRE. $210B+ analyzed. 20 yrs: SOM architect → family-office CRE investor. LA/Seoul.

[Steven Song @StevenPSong]

A month ago I said we'd torn our own product down to rebuild it — and that I'd show the work along the way.

This is what we built.

The data was never what was missing in commercial real estate. The vantage point was. Diald is the vantage point — one system for the 95% who've been running on spreadsheets and the 5% who never loved what they had.

Live today. Serious underwriting. $20 dollars a month. Link below.

[Steven Song @StevenPSong]

Register for webinar: https://t.co/RU7NJBL96t

[Steven Song @StevenPSong]

One week ago we opened the rebuilt Diald. Here's the decision that got us there — and what the first week did to it.

When we shipped the old version in January, we thought we knew our user: institutional investors. The acquisitions teams, the analysts — the people the legacy stack already serves. Which is to say: we, too, were building for the 5%.

Then the signups came in. Roughly two-thirds weren't institutional at all. Family offices. HNW owners. Business owner-occupiers. People with real portfolios and no analyst. And the share kept growing.

We had two options: keep building for the user we'd imagined, or rebuild for the user who showed up. We chose the second — paused go-to-market in the middle of our best growth quarter and tore down a product that was working.

Last week was the verdict. The roster of first accounts reads like the thesis: family offices. A bank's credit desk, stress-testing borrowers' deals. A spec-industrial developer. An owner-operator running his own multifamily. Acquisitions teams too — sitting next to people who've never had an analyst to hand a deal to.

The most expensive gap in software is the one between who you built for and who arrives. Week one says we closed it from the right side.

If you'd rather see it than take my word for it: August 26, Cheyne Bloch and I run it live — an address in, a memo out, nothing edited. 70+ people are already in the room.

Registration in the comments.

[Steven Song @StevenPSong]

Human judgement still matters. With @Edward13Cohen and @commobserver

[Steven Song @StevenPSong]

@Diald_ai is AI-driven, but real estate runs on the convictions of investors, brokers, landlords, and everyone who touches it.

I joined @Edward13Cohen 's @commobserver podcast to discuss why AI shouldn't replace judgment, but only sharpen it.

Full episode: https://t.co/t37TMgqHA8

[Steven Song @StevenPSong]

Family offices and independent operators now control more than half of all U.S. commercial real estate value. But for these cohorts, properly evaluating a deal feels like finding a needle in an enormous haystack — finding data scattered across zoning records, permit filings, and market reports, with no institutional research team to pull it together.

@wealth_mgmt just reported on how our new platform is designed to close that gap.

Instead of leaning on generic metrics, Diald assesses whether a neighborhood's narrative actually holds up against the data — drawing on safety data, infrastructure, regulatory shifts, and local sentiment. The platform also analyzes a specific property type in a specific location, since the same block can be a strong bet for multifamily and a weak one for office.

Our redesign also brings institutional-grade analysis to investors no matter their experience with the industry. Users can input their data, ask questions, and customize variables using plain language, which Diald turns into a full financial model through an intuitive interface.

Check out the full piece here: https://t.co/I8mVqVpXP4

[Steven Song @StevenPSong]

The striking part: the priciest high-growth stage ($0.95 at $2–5M) still beats the cheapest 'Others' stage ($1.00 at <$250k). An efficiency advantage that survives a 20x in practice size isn't a budget line — it's a system. Curious whether the study decomposes the gap: channel mix (referral-weighted?) vs. conversion efficiency.

[Steven Song @StevenPSong]

@gregofisch Probably both. Underwriting is just paranoia with citations.

[Steven Song @StevenPSong]

@NewsLambert The actual cliffhanger is the org chart. CFO absorbing COO in the same week as 7% layoffs puts capital allocation and operations in one brain. And a first-ever chief legal &amp; policy officer tells you where Zillow expects the next five years to be decided.

[Steven Song @StevenPSong]

@TomInvests @Diald_ai We designed to that test. The score moves for exactly two reasons — exit cap or hold period. Baseline is Diald’s market read; the delta is the underwriter’s stance, attributed inline to whichever lever changed. If a movement ever needs a manual, that’s a bug.

[Steven Song @StevenPSong]

A month ago I said we'd torn our own product down to rebuild it — and that I'd show the work along the way.

This is what we built.

The data was never what was missing in commercial real estate. The vantage point was. Diald is the vantage point — one system for the 95% who've been running on spreadsheets and the 5% who never loved what they had.

Live today. Serious underwriting. $20 dollars a month. Link below.

[Steven Song @StevenPSong]

You've written our design spec. Assumptions in @Diald_ai are tagged baseline vs. override, memos generate from an immutable versioned snapshot so the source→assumption→decision chain survives, missing inputs get flagged rather than silently defaulted, and the Confidence Score measures exactly one thing: how well-supported the exit cap is. Underwrite aggressively and the number on the cover drops. Put a real deal through it and tell me where the chain breaks — first month's free.

[Steven Song @StevenPSong]

@MichaelKitces @FPANorCal @t3techhub @AICPA We'll be at Future Proof Festival! Thank you for sharing this --

[Steven Song @StevenPSong]

Doors are open: https://t.co/oB6I2WA8Uk — $20 a month, first month free for the first 500 accounts, cancel anytime.

[Steven Song @StevenPSong]

Two weeks ago I wrote about a dentist who owns three medical office buildings — $1.4M of NOI running on QuickBooks, Dropbox, and an Excel sheet his accountant maintains.

I expected dentists to respond.

Instead, a commercial roofing president wrote that his industry has the same split: the national firms build for portfolio-scale clients, everyone else pieces it together. A founder who builds for the 1–4 unit residential world said the pattern repeats there. An investments director summed up the general mood: "Excel is going to be the death of me."

And the reactions filled in the rest of the map — a self-storage managing director, a healthcare-portfolio owner, a real estate CFO, a debt director who stress-tests other people's assumptions for a living.

But the line I keep coming back to arrived earlier, from an owner who signed up months ago: "I run my $40M CRE portfolio in Excel, because the real software is built for someone who isn't me."

The dentist was never a story about dentists. The stack is the story. Most of the people who own commercial real estate in America run it on some version of the same three tools, none of which was built for the decision they're actually making.

So this time the ask isn't "join a list."
Bring the building.

We open Monday, August 10. Email me what you have — rent roll, T-12, OM — at [email protected] this week, and during launch week we'll put your building through Diald together on a live screen share. Twenty minutes. Not a demo asset. Yours. First ones in get Monday morning.

DMs work too, if that's easier.

Waitlist: https://t.co/DcRUNBW4Bl
First 500 accounts get the first month free. Cancel anytime.

[Steven Song @StevenPSong]

August 10.
If you want in on day one: https://t.co/JZk9HHx6ag — one-click sign-in the moment we open, and the first 500 accounts get their first month free.

[Steven Song @StevenPSong]

The obvious question about a market this size: if 19 in 20 CRE owners have no real software, why has nobody built it?

The honest answer: for twenty years, the economics didn’t work.

Serving this market meant enterprise software — seat licenses, sales teams, implementation, training. That model needs customers who buy seats by the dozen and staff analysts to run them. The dentist from last week’s post doesn’t. Most of the 95% doesn’t. The revenue per customer could never carry the cost of reaching them.

So the industry concluded these owners “don’t need software.” What that actually meant: nobody could afford to sell it to them.

What changed isn’t the market. It’s the cost of the product. When analysis is generated by software instead of assembled by hand — when underwriting a property costs a query instead of a consultant — the economics invert. The owner who could never justify a five-figure seat license can justify a subscription.

For the first time, the bigger half of the market is serviceable.

August 10: Waitlist in the comments.

[Steven Song @StevenPSong]

August 10. If you want in on day one: https://t.co/JZk9HHx6ag — one-click sign-in the moment we open.

[Steven Song @StevenPSong]

Every startup claims to be the Carta of something. Bear with me — the parallel here is structural.

Before Carta, private-company equity was tracked in Excel and PDFs, by founders and law firms. The infrastructure that existed — NASDAQ, DTCC — was built for public markets. Most American companies ran on a system nobody designed for the people running them.

Not because the asset didn't matter. Because the owners were too many, too small, and too scattered for enterprise software economics to reach.

Sound familiar? 2.5–3 million entities own US commercial real estate. $15 trillion in value. The infrastructure that exists was built for institutions. Everyone else runs on the same Excel-and-PDFs system private-company equity ran on in 2011.

Carta built the missing layer and became the default for an asset class.

The CRE version of that layer is what we've been building.

August 10.