- A deal team evaluating an automation-driven value proposition recognised early that judging it required more understanding of mortgage lending workflows and economics than the team had.
- HuxleyIQ produced a mortgage origination process map — application through closing, typical 30 to 60 day timelines, and the points where manual labour concentrates — and integrated it into the investment analysis rather than leaving it in a chat thread.
- Platform features were mapped against the origination stages they touch, so claimed time savings, error reduction and cost elimination could be located in a specific step rather than accepted in aggregate.
- HuxleyIQ flagged that the inferred benefits — cost-per-loan reduction, cycle-time compression, error-rate improvement — derived from vendor claims and not from independent verification.
- That flag became a diligence workstream, which is the correct outcome: the distinction between a platform assertion and a verified fact is the thing a deal team is paid to maintain.
- Domain context belongs in the shared analysis every reader sees, not in one analyst's private research, or the investment committee evaluates the thesis on unequal footing.
Every deal opens with a knowledge gap. A team that has never underwritten a business in a particular industry has to learn enough about it to judge whether the thesis is sound, and it has to do that inside the same window in which it is also reading the materials, modelling the outcome and forming a view.
In practice the gap usually closes by deference. Somebody on the team has adjacent experience, or the banker explains the industry, or the founder does. The explanation that wins is the one delivered most fluently, and it is almost always delivered by a party with an interest in the outcome.
The team named the gap
A team was evaluating a platform whose value proposition rested on automation. The claim was the usual shape: the software removes manual work, and the removal of manual work is worth money.
Whether that claim was true depended entirely on facts about mortgage lending the team did not yet have. Where does the manual work actually sit? What does it cost? Is the slow part of the process slow because a person is doing it, or because a third party has not responded yet — because automating the second thing saves nothing.
The team recognised this and said so, which is the part that most deals skip. Naming a domain gap early is an act of discipline; it is much easier to proceed and let the gap quietly resolve itself into whatever the seller said.
Context built into the analysis, not around it
Using the integrated chatbot, the team asked HuxleyIQ to close it. Two supplements came back, and both were written into the investment analysis rather than returned as a side conversation.
The first was industry context and process mapping: the mortgage origination lifecycle from application through closing, the typical 30 to 60 day timeline, and where in that sequence manual labour concentrates.
The second mapped the platform's features — automated data aggregation, intelligent document processing, real-time verification — against the origination stages each one touches, and set out where the claimed time savings, error reduction and cost elimination would arise.
The mapping is what makes the value proposition falsifiable. A claim about cost per loan, taken in aggregate, can only be believed or disbelieved. The same claim attached to a named step in a named process can be checked: is that step actually where the hours go, and does the feature actually remove them?
That the supplements went into the analysis rather than into one analyst's notes is not an administrative detail. Domain context that lives in a private thread produces an investment committee where one person has it and everyone else is nodding.
The flag that became a workstream
The most useful output was a caveat.
HuxleyIQ flagged that the inferred benefits — cost-per-loan reduction, cycle-time compression, error-rate improvement — derived from the vendor's own claims and not from independent verification. The mapping showed where the savings would appear if the claims held. It did not establish that they held.
Consider the alternative. A system that had produced the same process map, the same feature mapping and the same benefit figures without that line would have laundered a vendor's marketing into the firm's own investment analysis, wearing the firm's formatting and carrying the authority of internal work. It would have looked more finished. It would have been worse than useless, because the team would have stopped asking.
Instead the flag became a workstream: those three metrics were carried forward as things to be validated, not as things established.
The distinction between what a platform asserts and what has been verified is not a technicality. It is most of the job.
Learning the industry is diligence
It is tempting to file this under research and treat it as preparatory — something done before the real work starts.
It is the real work. An investment thesis is a claim that a particular mechanism will produce a particular result in a particular industry, and it cannot be evaluated by anyone who does not understand the mechanism. A team that does not close its domain gap does not thereby avoid forming a view; it forms one anyway, on borrowed understanding, usually borrowed from a seller.
What changes when the gap is closed early and in writing is not that the team knows more. It is that the thesis gets argued on shared footing, by people who can now tell the difference between a feature and a benefit — and who know which of the benefits nobody has checked yet.