Meet Elena. She's the managing partner of a value-add multifamily sponsor in Denver: six assets, 38 LPs, $50M in committed capital, and a team of four. Reporting is nobody's job, which means it is everybody's job for three weeks out of every thirteen. You are not looking for AI to write your LP letter. You are looking to stop rebuilding the plumbing every quarter so the three weeks go back to asset management.
Over two hours you build Elena's reporting workflow and run it against a closed quarter. Accounting exports become a reconciled dataset. Fund structure and waterfall mechanics get encoded once. Variance becomes written commentary, asset by asset, including the deal that missed. The letter and capital account statements come out formatted and on-brand. Then the LP questions arrive and get answered against the same dataset in minutes.
The part you spend the most time on is the part most AI reporting demos skip entirely: making sure the numbers are right. A reporting workflow that produces a confident wrong figure is worse than no workflow at all, and the difference between a demo and something you can really run is a verification layer you trust enough to put your name on.
What you'll build
- A reconciled quarterly dataset built live from a trial balance, six inconsistent rent rolls, and a bank statement
- A fund-literate instruction set encoding waterfall tiers, an 8% pref compounding annually, and a JV ownership split
- A verification pass that catches a wrong figure before it reaches an LP
- A finished LP letter, capital account statements, and fund performance summary in your firm's voice
Paul is also a partner at PTB, a real estate investment banking boutique, and focuses on the intersection of capital markets, media, and alternative real estate. He has funded over $1B of real estate projects and platforms, and acquired and asset-managed over 8 million square feet of office, industrial, and multifamily assets in the US.
Contact the host →
Daniel Kronovet is a senior software engineer with ten years of industry experience. He specializes in machine learning and systems for organizational intelligence, with past work resulting in a patent and several academic papers. He is also a third-generation real estate operator who has sponsored, developed, and managed his own coliving project.
Contact the host →You'll learn how to
Turn accounting exports into a reporting dataset
Pull trial balances, operating statements, rent rolls, and bank reconciliations into one reconciled quarterly dataset, without changing how your books are kept or who keeps them.
Teach the workflow of your fund once
Encode entity structure, LP roster and commitments, and waterfall mechanics so capital account math comes out right every quarter instead of being re-derived by hand.
Keep the numbers honest
Build a review step that catches a wrong figure before an LP does, so the workflow earns the trust it needs to run unsupervised.
Draft commentary from variance
Move from budget-to-actual tables to written explanation, with the workflow flagging what moved, proposing why, and surfacing the two or three questions that genuinely need an asset manager's answer.
Write in your firm's voice
Turn an existing LP letter into a format specification, section order, metric hierarchy, disclosure conventions, and tone, so reports come back sounding like your firm rather than like a language model.
Absorb the long tail
Answer one-off LP questions and document requests against the same dataset in minutes, which is where more IR time goes than anyone accounts for.
What the workshop covers
The Quarter That Just Closed
What actually lands on a reporting desk: a trial balance from the accountant, six rent rolls in three different formats, a bank statement, a mid-quarter refinance nobody documented cleanly, and one asset behind on lease-up. You start from the mess rather than a tidy sample file.
Building the Quarterly Dataset (Live Build)
Getting the workflow to normalize inconsistent exports into a single reconciled dataset: consolidated results, asset-level NOI, occupancy and leasing activity, debt service coverage, and a flagged list of items that do not tie. You see the ties that break and how they get surfaced instead of quietly averaged away.
Encoding the Fund (Live Build)
Writing the instructions that make the workflow fund-literate: waterfall tiers, an 8% pref compounding annually rather than simple, the JV where the fund owns 62% of an asset, the management fee basis, the reporting calendar. Done once, reused every quarter.
Keeping the Math Verifiable
Where the workflow is allowed to calculate and where it must cite a source cell, how to build the verification pass, and what a good review step actually looks like when the person reviewing has forty minutes rather than a day. You run the pass live and catch what it got wrong, because it will get something wrong.
Writing the Commentary (Live Build)
Building the skill that turns variance into explanation for each asset: what moved, why, what management is doing about it, and what to expect next quarter. Including the underperforming deal, where you work on getting honest language rather than the softened non-answer AI defaults to.
Assembling the Package (Live Build)
Producing the finished quarterly deliverables: the GP letter in the firm's voice, the fund-level performance summary with charts, per-LP capital account statements reflecting the mid-quarter refinance distribution, and the asset appendix. Plus staging distribution and updating the data room.
The Long Tail
Three LP emails arrive. One asks how their capital account changed after the refi and how gross-to-net compares to the original projection. One wants K-1 timing. One wants a rent roll. All three answered against the dataset already built, before the fourth email arrives.
Running It Next Quarter
Turning a one-time build into a repeatable cycle: what to hand an analyst, what to keep as a partner-level judgment call, how the workflow adapts when the fund adds an asset or a side letter changes a reporting obligation, and where the real time savings land once the setup cost is behind you.
Format & access
One live session
Two hours on Zoom, with time for questions throughout.
Hands-on build
The workflow gets built live in Claude, the AI platform used for this session. Participants are encouraged to follow along in their own account (Claude Pro or Team recommended).
Materials & ongoing access
Recordings, slides, the LP letter and capital account templates, the skills files, and the number verification checklist land in your Thesis Driven account within one week, whether or not you attend live, and stay there for whenever you need them.
Frequently asked questions
Do I need a Claude account to participate?
A Claude Pro or Team account is recommended to follow along in real time, but not required. All instructions, skills files, and templates land in your Thesis Driven account so you can build the workflow at your own pace afterward.
Is it safe to put fund financials into AI?
This gets addressed directly during the build: which account types keep your data out of model training, what to keep local versus uploaded, and how to structure the workflow so LP-identifying information stays segmented. If your LPA or side letters restrict data handling, the session covers how to work within those constraints.
Does this replace my fund administrator?
No. What changes is where partner time goes: less of it spent formatting and chasing the same three LP emails, more of it spent on the judgment calls only a partner can make. Reconciliation and sign-off don't move.
Does it work for my fund structure?
The case study is a value-add multifamily fund, but the mechanics are structure-agnostic. Encoding a waterfall, a fee basis, and a reporting calendar works the same whether you run a closed-end fund, a series of single-asset syndications, or a programmatic JV.
Hear from our alumni
"Brad and Paul opened my eyes to how real estate deals get put together, where incentives lie, and how we might create business cases for proptech and climate tech. Great stuff!"
"A perfect introduction to the most important real estate concepts, distilled down in the perfect way to absorb and retain. Paul and Brad clearly thought a lot about how to actually educate, not just data-dump the group."
"Excellent course that guides you through a fictional case study with detailed explanation of every single step for all personas at every stage of a real estate deal. I highly recommend enrolling."
"Huge thank you for an amazing five weeks. The 'Selling into Real Estate Owners' course content is a goldmine for anyone building or selling in PropTech, and the weekly cohort discussions are a rare chance to learn directly from peers."
"A collection of engaging and collaborative sessions on how to launch and structure a venture into real estate. Paul and Brad put on a great class with an even better collection of participants."
As featured in
Automating Investor Reporting with AI
Live on Thursday, November 12, 12-2pm Eastern Time. $299.
Register for this workshopRecordings are sent to all registered attendees, whether or not you attend live.