When we started building PivotPt Capital’s deal infrastructure, the obvious path was the one every other small PE fund takes: Excel models, shared Dropbox folders, email chains, and a CRM bolted on after the fact. It’s the path of least resistance, and it works until it doesn’t.
We chose a different path. We built the Beco Collective Intelligence Platform: a proprietary, full-stack acquisition and asset management system that handles everything from deal intake to LP distribution, with an AI layer running through the middle. Here’s why, and what it’s taught us.
The core problem with manual boutique hotel underwriting
Boutique hotel deals are information-dense in ways that generic real estate underwriting tools don’t handle well. A single offering memorandum might contain trailing financials in three different formats, occupancy data that needs to be cross-referenced against STR comps, a capitalization table with seller financing assumptions baked in, F&B revenue consolidated with rooms revenue, and a management fee structure that disappears entirely in the expense presentation.
Processing that manually is slow, error-prone, and non-repeatable. Two analysts working the same OM will make different normalization decisions. Those decisions compound through the model. By the time you’re presenting a return projection to an LP, you’ve made dozens of small judgment calls that are documented nowhere and auditable by no one.
We wanted something better: a system where every assumption is explicit, every normalization decision is logged, every model output is traceable back to a source document, and every LP-facing number has a documented methodology behind it.
What the platform actually does
The Beco Collective Intelligence Platform is built on a modern full-stack architecture: Next.js frontend, FastAPI backend, PostgreSQL for transactional data, and BigQuery for market analytics. The Anthropic Claude API handles document intelligence tasks. It’s deployed on Google Cloud Run and integrated with SendGrid for LP outreach.
The core workflow looks like this: a broker OM enters the system as a PDF. The document extraction module parses it, pulls operating statement line items, and maps them against our normalized schema. The underwriting engine, covering 16 modules from market quality scoring to returns waterfall computation, runs against that extracted data and our market intelligence database. Risk flags are surfaced automatically: lease line issues, franchise liquidated damages exposure, Prop 13 reassessment impact, F&B losses buried in consolidated revenue, and occupancy anomalies that suggest the trailing period is unrepresentative.
The output is a structured deal brief with reconstructed NOI, projected NOI under Beco management, IRR and MOIC at the ask price, brand fit score, and a flag register documenting every open issue. That brief goes to our internal investment committee before any external communication happens. LP outreach is gated; nothing goes out without human review and approval.
What proprietary infrastructure signals to LPs
Here’s the less obvious reason we built this: institutional LPs are increasingly asking not just what you invest in, but how you decide. The process question has become as important as the performance question, particularly for emerging managers building a track record.
When we can show an LP that every deal we’ve reviewed has a documented OM Variance Score, a structured flag register, and a traceable assumption set, that’s not just operational discipline. It’s due diligence evidence. It’s proof that the return projections we present are generated by a repeatable process, not reverse-engineered from a target number.
The platform also enables something that Excel-based underwriting can’t: portfolio-level analytics. As we close deals and accumulate actuals, we can benchmark our projections against real performance, identify where our models are systematically optimistic or conservative, and improve them. That feedback loop is the difference between a fund that gets smarter over time and one that makes the same modeling errors across every vintage.
The longer game
We built more infrastructure than Fund I strictly requires. That was intentional. The platform scales to Fund II, Fund III, and eventually to a third-party asset management business where Beco Collective manages properties for other owners using the same system.
We also see optionality in the platform itself. At 30 or more deals under management, the data asset, comprising STR benchmarks, cap rate comps, and OM variance distributions by market, becomes valuable in its own right. We’re not building toward a technology exit. But we’re not ignoring the possibility either.
For now, the platform does what it was built to do: make us faster, more rigorous, and more transparent than a fund our size has any right to be. In a market full of manual processes and optimistic broker numbers, that’s a durable edge.
PivotPt Capital is a boutique hotel acquisition fund. The Beco Collective Intelligence Platform is a proprietary system developed internally and used exclusively for PivotPt Capital deal management and LP operations. For investor materials, visit pivotptcapital.com.
