How Westlake Realty Group cut underwriting to 20 minutes, then took AI into asset management

Westlake Realty Group underwrites across five asset classes, and each offering took six to eight hours of senior time before anyone would put a number in front of a decision maker: rekeying the rent roll, rebuilding the model, pulling comparables, and then checking the whole thing again because a single mistyped figure changes the answer.

The binding constraint was capacity, not judgment. Marginal deals were dropped because there were not enough hours in the week to underwrite them, which meant the deals that got a real look were the ones that happened to arrive on a quiet week rather than the ones that deserved it.

Asset management had the same shape. Each year every property budget is rolled forward by hand from leases, contracts, and actuals, and the review that matters, whether an assumption still holds, waits behind the assembly.

PSV built the underwriting analyst against Westlake's own eight-tab institutional model rather than a generic template. The output arrives in the format the firm already argues over, so nothing has to be re-modeled before it can be reviewed.

Deal materials go in, the analyst parses them, populates the model, pulls comparables, and applies the firm's return thresholds. Every figure it writes carries a pointer back to the source document. When a reviewer questions a number, the answer is a page and a cell reference rather than a rebuild.

One engine covers all five asset classes the firm buys. There is no separate tool, template, or process to maintain per property type.

The underwriting work became the template for the operations side of the firm. PSV built Westlake a Company Brain over its own document library, so an assistant asked for next year's budget on a property starts from the firm's leases, contracts, and ledger detail instead of a blank workbook.

Each draft budget arrives with reviewer notes explaining the reasoning behind its assumptions, and with a list of what it could not confirm, such as a contract it cannot find. The reviewer's first read is of the open questions rather than the arithmetic.

The Brain works inside the firm's own controls. A user only gets what their document permissions already allow, it is set up around the firm's sign-off levels, and a correction made on one budget carries forward to the next.

Per-deal review dropped from six to eight hours to about 20 minutes, roughly 18 to 24 times faster on the senior time each deal costs.

Because every number is traceable, review became checking rather than rebuilding. The senior hours that used to go into assembling a model now go into the deals worth arguing about.

Capacity stopped acting as the filter. Offerings that would have been dropped for lack of hours get underwritten to the same standard as everything else.

In asset management the work is in rollout, with the firm's own team reviewing each batch of budgets before the next one starts. It is deliberately published without a figure. We would rather show what is being built than print a number the rollout has not earned yet.

Every deal lands in the firm's own eight-tab institutional model with comps pulled and every number traced to its source. Senior review takes about 20 minutes. The same approach now runs in asset management, where the firm's Company Brain drafts each property's annual budget from the firm's own records for the operations team to review.

Per-deal review dropped from six to eight hours to about 20 minutes, roughly 18 to 24 times faster on the senior time each deal costs.

Westlake's senior team reviews every model, and its operations team reviews each batch of draft budgets before the next one starts.

Underwriting figures are from the published PSV CRE case-study brief. The asset-management work is in rollout and carries no figure yet.