The transfer test is simple: does the system run at a new portfolio company without the founder present. Two live deployments on one runtime answer that question. A hundred slides of modelled projections do not.
What PE Buyers Should Ask Before Believing Any AI Vendor Pitch
Every AI vendor walking into a PE diligence meeting arrives with the same kit: a polished deck, a projected ROI curve, and a demo environment that has never touched a real customer. The transfer test cuts through all of it. The question is not whether the system looks impressive in a conference room. The question is whether it runs at a new portfolio company without the founder in the room. Two live deployments on one runtime answer that question. A hundred slides of modelled projections do not.
Why the AI Vendor Pitch Has a Structural Problem
More than three billion dollars has been deployed into AI roll-ups, firms that acquire service businesses and then apply AI to their operations. General Catalyst has allocated roughly 1.5 billion dollars from its Creation Strategy to incubate AI-native companies and acquire fragmented service businesses. Thrive Capital launched a dedicated vehicle and brought OpenAI in as an equity partner, with OpenAI embedding research and engineering teams directly inside Thrive-owned companies. Long Lake took Amex Global Business Travel private for 6.3 billion dollars. The capital is serious. The intent is real.
But every one of those players is capital-first. They buy the business, then scramble to build the AI. The vendor pitch that lands in their diligence process is almost always a projection built on top of a pilot, not a mechanism verified across multiple live environments. That distinction is the entire ballgame for a PE buyer who needs to underwrite the system, not the story.
According to McKinsey's State of AI research, only about one-third of organizations have moved beyond pilot phases to enterprise-wide AI scaling. The gap between experimentation and verified production is not a technology problem. It is a governance and transfer problem. And it is exactly the gap that a well-structured diligence process should expose.
The Transfer Test as a Diligence Framework
The transfer test is simple to state and hard to fake. Does the system run at a new portfolio company without the founder present? Not with the vendor's implementation team on-site for six months. Not with a dedicated integration engineer holding the hand of every workflow. Without the founder. Without the original team. On its own.
Most AI vendor pitches fail this test before the first question is asked, because the system being pitched has never been deployed twice. It has been deployed once, carefully, with the vendor's best people managing every edge case. The deck then extrapolates that single deployment into a portfolio-wide projection. That is not a mechanism. That is a model.
The questions that expose this gap are not technical. They are operational:
A vendor with a real system answers every one of these questions in under two minutes. A vendor with a pitch deck pivots to the roadmap.
Modelled Projections Versus Verified Mechanisms
The distinction between a modelled projection and a verified mechanism is the most important concept in AI vendor diligence, and it is almost never named explicitly in a pitch meeting.
A modelled projection says: based on our pilot data, we estimate that deploying this system across your portfolio will reduce dispatch labor costs by 30 percent. A verified mechanism says: here is the dispatch agent running live on a twenty-truck fleet, here is the windshield time before and after, here is the overtime reduction number, and here is the second fleet where the same agent is running on the same runtime with the same results.
ServiceTitan, Jobber, and Housecall Pro all produce data that supports modelled projections. They record what happened. They surface dashboards. They give an operator the inputs needed to build a projection. But the projection still requires a human to act on the data, and the human is the variable that a PE buyer cannot underwrite at scale. The system that runs the work itself, not the system that records it, is the only system that survives the transfer test.
The WeLaunch orchestration brain is the architecture that makes the transfer test passable. The horizontal runtime, the big brain and fast brain router, the MCP connectors, and the shared agent state layer are portable by design. When Dex runs dispatch on a twenty-truck facility fleet and Molly runs checkout on the same deployment, those agents are not bespoke builds. They run on the same brain that runs the dunning, renewal, and winback lifecycle for a pest control business with 64,000 customers. One runtime. Multiple live deployments. That is the receipt. See the orchestration brain running in production before accepting any projection as a substitute.
What the Audit Trail Tells You That the Demo Never Will
Governance is what makes autonomy safe to underwrite. A PE buyer who cannot audit the system cannot price the risk. The audit trail is not a compliance checkbox. It is the mechanism that converts an autonomous system into an insurable asset.
The fast brain in WeLaunch's architecture suppresses double contact. Agents share state so no customer receives two calls about the same invoice. Every decision is logged. The hard 20 percent of edge cases, the ones that require human judgment, escalate to a named human with a defined SLA. That is not a limitation of the system. That is the design. Autonomy without a defined escalation path is not a product. It is a liability.
When a vendor cannot show you the audit trail for the last thirty days of agent decisions, the diligence answer is not "we need more information." The diligence answer is that the system is not ready to be underwritten. The Facility19 control tower is the live proof point that this architecture exists outside a pitch deck. Eight agents plus one brain running a twenty-truck fleet, with every dispatch decision, compliance flag, and overtime trigger logged and auditable.
Retainer Revenue Versus Product Revenue as an Investor Pitch
One more question that almost never gets asked in a vendor pitch meeting: is the revenue this system generates retainer revenue or product revenue?
Retainer revenue means the vendor charges a monthly fee to keep the system running. Product revenue means the system generates measurable, attributable output, recovered invoices, reduced windshield time, closed churn gaps, that can be tied directly to EBITDA. The difference matters at exit. A system that generates product revenue compounds. Every serviced job makes the next one cheaper to win because the review data and the route data are reused to find the next customer on the same street. That density is not a feature. It is the exit multiple.
A roughly 10x model ROI on a 64,000-customer lifecycle is not a projection. It is a sized and automated mechanism. See how the home services lifecycle is structured before accepting a vendor's modelled ROI as equivalent. And when evaluating legal or adjacent verticals, the ten-agent legal deployment shows what on-premise-ready orchestration looks like when it is already live.
WeLaunch is not a fund that needs AI. It is the AI that funds need. The brain was built first. It is live in production. The capital conversation comes after the receipts, not before.
One brain. Every portfolio company.
Frequently Asked Questions
What is the transfer test and why does it matter in AI vendor diligence?
The transfer test asks whether an AI system runs at a new portfolio company without the original implementation team present. It matters because most AI vendor pitches are built on a single, carefully managed deployment. A system that cannot transfer to a second environment without rebuilding is not a scalable asset; it is a bespoke project dressed as a product.
How do I tell the difference between a modelled projection and a verified mechanism?
A modelled projection extrapolates from pilot data to estimate future performance. A verified mechanism shows the same system running live in at least two separate environments, on the same runtime, with documented before-and-after operational numbers. Ask the vendor to name the second live deployment and show the audit log. If they cannot, the projection is the only thing on the table.
What questions should a PE operating partner ask an AI vendor before a portfolio deployment?
Ask for the second live production deployment by name, the runtime architecture that connects all deployments, the audit trail for the last thirty days of agent decisions, the escalation path for edge cases the system cannot handle autonomously, and the internal ownership structure after the vendor's launch team exits. A vendor with a real system answers all five in under two minutes.
Why does the audit trail matter more than the demo for underwriting an AI system?
A demo shows the system performing under controlled conditions. An audit trail shows how the system behaved across thousands of real decisions, including the edge cases, the escalations, and the suppressed double-contact events. Governance is what converts an autonomous system into an insurable, underwritable asset. Without the audit trail, the risk cannot be priced.
How does WeLaunch's orchestration brain differ from field service software like ServiceTitan or Jobber?
ServiceTitan and Jobber record what happened and surface dashboards for a human to act on. WeLaunch's orchestration brain runs the work itself: dispatch, checkout, compliance, dunning, and renewal execute autonomously through named agents on a shared runtime. The distinction is not a feature comparison. It is the difference between a system that informs decisions and a system that makes them.
What does density mean in the context of an AI-native back office, and why does it affect exit multiples?
Density means that every completed job generates data, route history, customer reviews, and payment patterns, that makes the next job cheaper to acquire and service. The WeLaunch loop runs from lead through invoice and back to lead, compounding that density with each cycle. At a twelve-times exit multiple, a recovered churn dollar and a reduced cost-per-acquisition are not operational wins. They are balance sheet events.
Ready to See the Brain Before the Pitch
If you are evaluating AI vendors for a portfolio company or preparing for a diligence conversation, the right starting point is a live system, not a slide deck. See one brain running across portfolio companies and bring the transfer test questions with you. The receipts are already there.
