- A food distributor's CIM described world-class quality controls; a press release outside the data room described contract terminations for recurring food quality failures.
- Contract terms presented in a CIM as five-year fixed were, against industry norms, more likely structured as one-plus-four renewable with 60-day termination provisions — a materially different risk profile from the one represented.
- Working capital requirements can be assessed from comparable companies before a balance sheet is ever shared, which means a forecast can be stress-tested pre-IOI rather than in week three.
- Two of the three findings were hypotheses rather than proven facts, and were carried into a CEO conversation as questions — the distinction between an inference and a verified fact is what made them usable.
- The investment team raised all three in a CEO fireside chat; the CEO confirmed all three and offered no satisfactory mitigation, and the team passed.
- The return on a pre-IOI screen is not the memo it produces. It is the diligence process a firm never has to run.
Most accounts of AI in diligence are accounts of work that got done faster. This is an account of work that did not get done at all, which is the more valuable outcome and the harder one to sell.
An investment team ran a food distributor's confidential information memorandum through HuxleyIQ before committing to the process. Three findings came back. All three were material. None of them were in the document.
The press release the CIM did not mention
The CIM was emphatic about quality control. World-class was the register — the kind of language that reads as boilerplate and is therefore skimmed, which is what boilerplate is for.
Outside the data room sat a press release describing contract terminations arising from recurring food quality failures.
That is not a nuance. For a distributor, the ability to hold a contract is the business, and quality failures severe enough to end one are an operational risk that belongs in the first paragraph of a risk section. It was absent. Worse than absent: the document said the opposite, in confident language, about the exact dimension on which the company had demonstrably failed.
Nothing about finding it required sophistication. It required somebody to go looking outside the materials the seller assembled, at the moment the CIM was being read rather than three weeks later — and to notice that the strongest claim in the document was the one worth testing first.
The confident claim is the one to test. A CIM's superlatives are a map of where the seller expects scrutiny.
"Five-year fixed" is a claim about a document
The CIM stated five-year fixed contracts. Read quickly, that is a revenue-durability story: locked customers, predictable base, a multiple that can carry some leverage.
Set against how contracts in the sector are actually written, the stated terms were more consistent with a one-plus-four renewable structure carrying 60-day termination provisions. That is a different asset. A five-year fixed contract is a five-year obligation. One year plus four renewals, terminable on 60 days, is a rolling relationship with a two-month floor — repriceable, exitable, and worth less than what was described.
The team did not know this was true. Nobody could have, from a CIM. What the analysis produced was a specific, checkable hypothesis about the gap between how the contracts were characterised and how contracts of that type are normally structured. That is a question, not a finding, and it was carried forward as one.
Working capital without a balance sheet
No balance sheet had been shared. That would ordinarily end the conversation about working capital until well into diligence.
Working through comparable companies, the analysis concluded that working capital requirements would constrain growth and drive a substantial miss against the forecast the CIM presented. Distribution businesses fund their growth through inventory and receivables; the growth rate in the plan implied a cash requirement that the business, on comparable economics, would struggle to fund.
Again: an inference. Comparables are not the company. But it is an inference precise enough to put to management, and it converts a forecast from something to be accepted pre-IOI into something to be defended.
What the fireside chat was for
The team took all three into their fireside chat with the CEO.
This is the part of the sequence that matters most and gets described least. The output of a pre-IOI screen is not a verdict. Two of these three findings were hypotheses, and a system that had presented them as established fact would have been actively dangerous — it would have invited the team to walk into a management meeting asserting things it could not support. What made them usable was that they arrived labelled: here is what the record shows, here is what the record implies, and here is the difference.
The CEO confirmed all three. Asked how each was mitigated, the answers did not satisfy.
The team passed.
The return is the process you never run
Count what was saved and it is not reading time. It is the several weeks a competent diligence process would have spent arriving at the same three issues by the ordinary route — the quality-control problem surfacing in customer references, the contract structure emerging when the actual agreements were finally produced, the working capital gap appearing when the balance sheet did. All three findings were reachable. They were reachable in week three.
A firm running twenty processes a year can absorb that. A firm running four cannot, because the cost is not only the hours — it is the exclusivity, the deal-team attention, and the accumulating institutional commitment that makes a late no expensive to say.
The team was, in their own account, grateful not to have spent the time. That is the honest measure of a pre-IOI screen: not the memo it produces, but the diligence process a firm never has to run.