AI Underwriting Analyst
How Westlake Realty Group cut per-deal underwriting from 6-8 hours to 20 minutes
Six to eight hours of senior time to underwrite a single offering. Marginal deals were dropped for lack of capacity, not lack of merit.

PSV turns commercial real estate knowledge into an operating advantage with an AI brain: well-informed decisions, more efficient AI spend, and a shared intelligence layer that makes the capabilities of your best AI users a firm-wide advantage.
Pacific Software Ventures builds leading AI solutions for CRE firms. It turns commercial real estate knowledge into an operating advantage with an AI brain: a firm's data, institutional knowledge, and judgment put to work as well-informed decisions, more efficient AI spend, and a shared intelligence layer that makes the capabilities of a firm's best AI users a firm-wide advantage. PSV builds a secure, governed Company Brain that gives AI the context to understand how a team actually works, then deploys AI agents that help investors find conviction faster, asset managers operate portfolios more efficiently, and brokers win and execute more business with less manual work, by email or inside Claude, ChatGPT, Microsoft Copilot, and the tools the team already uses. PSV offers four engagements: AI Consulting maps how your firm works and sets the strategy and security posture; Company Brain + Custom AI Agent Builds builds the intelligence layer that gives AI your firm's context and puts it to work; AI Training puts the capability in your team's hands through PSV-led enterprise training and the CRE AI Institute; and AI-Powered Dashboards & Custom Projects covers live dashboards over your own data and one-off builds scoped to your firm. Deployments run in your Claude or Copilot workspace or in your own VPC, agents are scoped to the systems and folders you approve, every number is traced to its source document, and humans approve what becomes firm knowledge. The free AI assessment is an honest read on where you stand and which engagement fits, if any. Built by institutional CRE teams and AI researchers previously at PGIM, Goldman Sachs, and Kochava. Trusted by CRE professionals at Cushman & Wakefield, JLL, Newmark, Marcus & Millichap, Saunders Real Estate, Kimco Realty, Westlake Realty Group, Industry Real Estate Partners, and Stride Commercial.Trusted by CRE professionals at
Partnered with and backed by

The overlooked problem
Most firms make one of these four moves before the context is in place.

What happened
You hired an AI vendor to ship an agent in six weeks.

What came back
You get a vague roadmap, a tool list, and nothing that runs.
Meanwhile
Four things are getting worse underneath it, and none of them fix themselves.
Today's prices are subsidized, and the bill is climbing anyway. Seats, tools, and pilots stack up, with nobody measuring the spend against a return.
The demo worked. On a real OM, a real rent roll, and a real deadline, the output needs more review than the work it replaced.
Every implementation plan waits at the same two gates: the data is not in one place, and IT has not signed off. Neither moves on its own.
Two or three people are getting real leverage. What they have worked out lives in their own chat history, with no way to hand it to the rest of the firm.
All four have the same fix.See what it is
The missing piece
Your data, institutional knowledge, and judgment are spread across every tool, inbox, and head in the firm. The Company Brain ties them into one shared, governed layer of context, so every AI agent answers from your firm instead of from the public internet.
Shared knowledge
The Company Brain
What changes
Five problems every firm hits running AI on its own, and what each one becomes once the Brain is in place.
Access
AI quality depends on the user.
One person's Claude is excellent and another's is mediocre, because their context, prompts, and workflows are different.
Shared AI context across the firm.
Everyone works from the same approved firm knowledge, not just the one power user with the best prompts and a private workflow setup.
Knowledge
Company knowledge is fragmented.
Key context lives across email, CRM, Yardi, Drive, underwriting models, meetings, and people's heads.
Institutional memory that survives people.
Decisions, corrections, relationships, exceptions, and precedent stay with the firm when employees leave or teams change.
Efficiency
Every workflow starts from scratch.
Teams rebuild the same context for underwriting, IC memos, reporting, buyer lists, and the rest, every single time.
Lower token and context waste.
The Brain retrieves only the context a task needs, instead of stuffing whole folders, prompts, and documents into Claude again.
Judgment
AI lacks firm-specific judgment.
It knows CRE broadly. It does not know how this firm underwrites, decides, escalates, or reads an exception.
Better accuracy and consistency.
AI applies the right methodology, precedent, source-of-truth rules, and firm context instead of guessing from generic CRE knowledge.
Learning
There is no shared learning loop.
Manager corrections and past decisions stay trapped in a chat, a model, or a meeting instead of improving the next workflow.
AI that improves over time.
Meaningful human corrections, decisions, and outcomes feed back into the Brain, so the next workflow starts smarter.
Services
Firms arrive anywhere from no AI at all to agents already in production, so the engagement adapts to where yours actually is rather than the other way round. The first three run on the Company Brain; the fourth is scoped to whatever you already have in mind.
Your firm wants AI. There is no policy, no owner, and no agreed starting point.
We audit how the firm runs today, set the AI strategy and security posture with your IT team, and hand you a sequenced roadmap.
Your AI policy is in place, IT is on board, and you know what you want built.
The Company Brain gives AI your firm's context, then the workflows and agents run on top of it, inside your stack and under your review.
Your team already uses AI daily. Now you want them building, not prompting.
PSV-led training on building agents in Claude, and for Brain clients, on building them on the Brain so they inherit your firm's context.
You already know what you want made, and it is not an off-the-shelf agent.
Live dashboards over your own data, internal tools, and the projects that do not fit a standard engagement. Built on the Brain when you have one, standalone when you do not.
Works with your stack
Your firm already runs on these systems. Depending on the platform we integrate directly, work alongside it, or align our agents to the data it holds, always within that provider’s terms of service. You keep the stack you have.
CRE data & analytics
CRM, productivity, and data
Which one is yours
Four engagements, and usually only one is right for a firm at a given moment. Five questions, scored on what you tell us rather than on what costs the most.
Where does AI stand at your firm today?
1 / 5Then the agents
Agents come second, deliberately. Once the Brain holds enough of your context to be useful, our team deploys the agents that run on it, and they live wherever your team already works.
The agent gets a name and an inbox.
Best for
Lean teams, five to ten people, with no IT function and no appetite for a rollout.
Or the same agents run inside the accounts you already license.
Best for
Institutional teams with seats deployed and IT already through its review.
The agent suite
Our team deploys AI agents on the Brain for each desk: investors find conviction faster, asset managers operate portfolios more efficiently, and brokers win and execute more business with less manual work. Every correction you make and every deal an agent sees stays in the firm and shows up in the next one, and you keep the judgment calls.
Acquisitions
Takes what lands in the inbox and hands you a go or no-go, with the model already loaded.
What it knows
The difference between the box you say you buy in and the one your passes actually describe.
Asset management
Runs the morning pulls and the month-end pack you assemble by screenshot today.
What it knows
Which variances your team chases and which it lets ride, from your own commentary rather than a threshold.
Brokerage & BD
Keeps the list alive and hands you the day's dials with the direct numbers already on them.
What it knows
Which owners in your market are actually movable, from your own call history rather than a purchased list.
Capital & IR
Assembles the quarter in your format instead of collecting it person by person.
What it knows
How each LP reads a quarter, and which question comes back every time before anyone asks it.
Operations
Turns a 300-email day into a board with owners and dates, and keeps it moving.
What it knows
Which escalations turn into problems, learned from the ones on your own desk that did.
The rest of the bench
These five are the seats firms fill first. The library holds the rest, by desk and by seat, with what each one is built to know.
Customers
Named, on the record, and still working with us.

“We use PSV to underwrite IOS deals and produce BOVs in 1/10th the time it used to take us.”


“The team is loving PSV AI Strategy and Build Framework.”

“I’ve bragged about you guys.”

Case studies
Every number comes from work we handed a client. Each card opens the full write-up.
Buy side
Owners, operators, and allocators screening more deals with the same team.
Sell side
Investment sales, tenant rep, and advisory teams winning and working more listings.
Prefer to hear it from a client? Ask for a reference.
Your environment
Secure and governed by design. Where the work runs is a deployment decision you make, not a constraint we impose.
by PSVLiveCRE-specific AI training, Claude workflows and Skills, live sessions, and a community of CRE practitioners. Continuously updated as the tools change.
Our members come from teams at
Common Questions
What teams ask before working with us.
PSV helps commercial real estate firms turn their data, institutional knowledge, and judgment into a proprietary operating advantage: well-informed decisions, more efficient AI spend, and a shared intelligence layer that makes what your best AI users can do a firm-wide capability. Three services get you there. AI Consulting maps how your firm works and finds where AI pays off. Company Brain + Custom AI Agent Builds builds the secure, governed intelligence layer that gives AI your firm's context, then deploys the agents that run on it inside the tools your team already uses. AI Training makes your team AI-native, through PSV-led programs and the CRE AI Institute. Most clients start with one and grow into the others.
Those are the surfaces your team works in, and we build on top of them, not against them. Out of the box they start every task without your firm's context: your deals, templates, policies, or past decisions. PSV builds the Company Brain that supplies that context and the workflows that finish the work inside those tools, so output lands in your own model and CRM with every number traced to its source.
The secure, governed intelligence layer underneath our flagship engagement: what your firm knows, how it works, and how it decides, structured so AI can use it. It sits under Claude, ChatGPT, or Copilot, connected through MCP, and every agent that runs on it shares the same context, so what your best AI users have worked out becomes something the whole firm can use. Humans approve what becomes durable firm knowledge, so the Brain reflects the firm, not a model's guess.
That is the whole point. The mandate to use AI usually lands without anyone showing the team how, so the work keeps getting done by hand. We close that gap directly: hands-on training in your real workflows, plus deployed workflows that do the heavy lifting so your people supervise output instead of fighting a tool. No prompt-engineering degree required.
Because seats are not a system. A Copilot license sitting unused is the norm, not the exception, because nothing was built around it. We deploy AI into a specific, high-value process with a clear owner and measurable output. It gets used because it is doing real work, not waiting for someone to think up a prompt. That is also where AI spend gets efficient: the seats and the tokens go to work with a measurable return instead of to experiments that stall.
Hallucination is a design problem, not a reason to quit. Our systems are grounded in your data and source-traced, with validation, confidence thresholds, and approval gates on anything high-stakes. Every answer shows its work and you keep final say. The bar is output you would put your own name on.
No. Commercial real estate is our flagship and where we go deepest, but the strategy, training, and engineering work for any operating business. The CRE work is simply the proof that it holds up under real numbers and real stakes.
Ten minutes of questions about your firm, your workflows, and your current AI usage. You get an honest read on your AI maturity, the highest-value opportunities, and which engagement fits, if any. If the answer is none yet, it says that too. Qualified firms get direct access to the calendar.
You own it. The code, the workflows, the data, and the configuration are yours, and they run on infrastructure you control. No per-seat lock-in, no hostage data. If we ever part ways, you keep everything plus a runbook to operate it without us.
Yes. Encrypted in transit and at rest, access scoped and logged, and we never train shared models on your data. For regulated workloads we deploy into your own cloud or VPC and use providers with zero-retention agreements. We share our full controls documentation and sign your security addenda on request.
Ammunition. We deploy inside your own environment, your VPC or M365 world, behind your SSO, with access scoped and logged. We write the security memo for your IT team, answer their questionnaire, and get on the call with them. You should not have to fight that battle alone, and our job is not done until we pass your audit.
A real fear, and we treat it as step zero. Before any AI touches your numbers, we tie your template out against your last closed deals. If something has quietly not been ticking and tying, you find out first, from us, not from an AI that learned the mistake.
First engagements scope in days, not months. Custom engagements begin at $10,000, and the shape depends on the path: AI Consulting, an Institute seat for training, or a Company Brain build with workflows deployed on top. We share exact numbers on the first call, with no surprises later.
Then we route to whatever is. You are never married to one model. We run a harness that sends each task to the best model for it and swaps as the frontier moves, so you get the upside without the lock-in. The data, the workflows, and the system itself stay yours either way.
Don't take our word for it
Pick your favorite AI and ask what it thinks about PSV. No filter, no spin.
Let’s talk
Take the free AI assessment. Ten minutes, an honest read on where you stand, and which engagement fits, if any.
We take on a limited number of new partners each quarter.
Prefer email? zed@pacificsoftwareventures.com