Customer story · September 2026
From four hours to seven minutes
A multifamily operator with 15,000+ units and $2.5B AUM already had AI on every analyst’s desk. Here’s how it transitioned from ad-hoc prompting to an agentic workflow that runs every first-pass underwrite, with a return it can measure.
The gap
Everyone has AI.
Very few have realized ROI.
Why?
Access was never the constraint. Every firm has a license, and someone on the team writing prompts that generally work. What almost nobody has is a system: AI wired into where the work actually happens, running every cycle on real data without a person babysitting every step.
A chat window alone can’t track down the “right” documents, reconcile inconsistent formats, and produce an investor-ready report on deadline. That requires clean data pipelines, encoding your firm’s business logic, and handling the exceptions nobody wrote down.
Orchestrating this system is hard, and it’s why most teams stall on their existing AI pilots.
Why firms struggle to do this themselves
Ad-hoc AI
Excel plugins, documents uploaded to an LLM
- Untracked changes to the model
- Prompts that drift from analyst to analyst
- A different financial model per analyst
- Costly errors surface late, downstream
- ROI is anecdotal, untracked across individuals
With Ollo
One workflow, built and run by our team
- Checks on every model change
- One consistent data pipeline
- Business logic from the whole team, every time
- You receive and review the output
- One pipeline, so ROI is measurable
1 JLL, 2025 Global Real Estate Technology Survey, October 2025. Based on responses from 1,500+ senior decision-makers across 16 markets; 88% of investors, owners and landlords report piloting AI, while 5% report having achieved all program goals.
The problem
A full-time analyst, spent mostly on passes.
- Who
- A multifamily operator
- Portfolio
- 15,000+ units and $2.5B AUM, screening up to 500 deals a year
- The work
- A first pass underwrite on every deal: pull the OM, rent roll, and T12 from the broker’s deal room, Key the figures into the firm’s model, and layer in the firm’s own assumptions on leasing, growth, and cap ex.
Each underwrite took an analyst 1–4 hours depending on the complexity of the deal. That’s up to 2,000 hours a year, roughly a full-time analyst, spent mostly on deals the firm would pass on. And because those hours are finite, the team could only underwrite the deals it had time for, not every deal worth a look.
The solution
The underwrite, end to end, inside the model the team already uses
We embedded with the team to map its process, buy box, and business logic, then built a workflow that runs the first pass from deal room to reviewed model.
Extract and populate
The agent pulls the OM, rent roll, and T12 the moment a deal hits the broker’s deal room, normalizes the figures however each broker formats them, and populates the firm’s proprietary model.
Layer in firm assumptions
It layers in the firm knowledge that move a deal, such as crime trends, school quality, and new supply, applying the assumptions the team would otherwise carry in their heads. An analyst reviews the output, every figure linked to its source page.
The value
A full-time analyst back
Extracted figures match the source documents, and every one links to the page it came from, so an analyst verifies a number in seconds instead of re-keying it. The judgment layer is where the work was: the model’s assumptions on taxes, capex, and opex came within 5% of what the team’s own underwriters would have used, on a process that took them hours. Time kills deals; a faster cycle qualifies out bad opportunities sooner and focuses the team on the 1–2% that may close.
Two key takeaways
Going beyond the numbers
Pulling figures from a rent roll is table stakes. A firm’s edge is how it weighs what no spreadsheet captures: zoning, tax exposure, crime trends, new supply. The workflow encodes that judgment and feeds it into the firm’s own model, so every deal gets the same local read.
Get to a no faster
When more than 95% of deals end in a pass, speed on any single deal matters less than how many deals the team can afford to look at. More looks at the same hit rate means more acquisitions.
Who we are
Ollo isn’t yet another software you log into and manage. We’re a team of MIT and Yale engineers that deploys agentic workflows on top of how you already work: no new platform, no change management, no data migrations. Solutions integrate into your existing processes from day one.
How it works
We build one or two of your highest-cost workflows on your real data, and you see it running before there’s any fee or commitment. Here’s what we ask of your team:
- Discovery workshop60 minUnderstand your process and system landscape, and define a scope and success criteria for the proof.
- Weekly check-ins30 minReview progress on what we’ve built and iterate collaboratively as it takes shape.
- Working sessionsAs neededOne or two deeper sessions with the person who owns the process, so we capture the edge cases and judgment calls that aren’t written down.
Your data never trains a model and never leaves the infrastructure we agree on up front. Every output traces back to its source. SOC 2 in progress. After the proof, we productionize, run, and maintain each workflow while your team reviews and signs off.
Get started
See it running on your own data.
First workflow running in about four weeks. No fee or commitment until you’ve seen it work.
Talk to us