Fund level
Diligence with the fund as the subject: structure, terms, jurisdiction, portfolio construction and existing holdings assessed on look-through.
Fund levelDiligence with the company or asset as the subject
This is diligence at its most familiar altitude. A single company, project or asset is under assessment, the data room is the primary evidence base, and the output is a recommendation the investment committee can interrogate. Reuben AI structures that work so the questions asked are appropriate to the asset class, and every answer carries the document it came from.
The evidence base at company level
Financial statements and management accounts, the cap table and its instruments, customer and supplier contracts, board minutes, technical and product material, regulatory permissions where they apply, and the founder or management team assessment. These are parsed natively rather than summarised, so a figure in a memo can be traced to the line in the statement it came from and the period it covers.
Questions that change with the asset class
A software company, a private credit facility, an infrastructure project and a real asset do not share an evidence set. Workstream templates carry asset-class-specific requirements, so an operating asset is asked about maintenance and offtake while an early-stage company is asked about retention and concentration. This is where breadth of asset class coverage stops being a marketing line and becomes the reason the diligence is usable.
Concentration, dependency and single points of failure
Most losses trace back to a dependency that was visible in the documents and never assembled into a single view. Revenue concentrated in a handful of customers, a supply relationship with no alternative, a licence with a renewal date, key-person risk with no succession. Because the workstreams see each other, a commercial finding and a financial finding that only matter together are surfaced together rather than in separate sections.
What the committee receives
A memo generated from the findings, with section-level confidence reflecting the tiering of the claims beneath it, open questions recorded as open, and every claim linked back to its source. Where evidence does not exist, that is stated. The committee decides with the gaps visible rather than discovering them later.
It becomes the monitoring baseline
The company profile does not close when the deal does. What was underwritten becomes the baseline the position is monitored against, so variance later is measured against the original case rather than against the most recent management update. That is the difference between monitoring and reporting.
Diligence at three levels
The same engine runs at three altitudes. A company or asset can be the subject, a fund can be the subject, or a manager can be the subject. Findings at the lower level roll up, so a position in a fund can be read through to the underlying holdings and back out to the portfolio it sits inside.
Manager level
Diligence with the manager as the subject: track record, team, process consistency, operational controls and governance, evidenced rather than asserted.
Manager levelWhat makes the diligence deep
Diligence is only as good as the context behind the questions it asks. Credit diligence and venture diligence do not ask the same things, a fund in one jurisdiction is not assessed like a fund in another, and a company profile that stops updating at signing stops being useful. Asset class coverage, jurisdiction and vehicle structures, market research and continuously updating company and manager profiles all feed the same diligence record.
Asset class coverage
Native asset classes and sub-asset overlays, so evidence requirements differ by what is being assessed.
Jurisdictions and vehicles
Structures assessed against the conventions of the jurisdiction they are formed in.
Market research
Market structure and comparables attached to the deal rather than held in a separate document.
Living profiles and portfolio view
Company, fund and manager profiles that keep updating, and diligence findings that become the monitoring baseline.
Go deeper on diligence
Each page below covers one part of the diligence layer in detail. Start with the one closest to the decision you are making.
AI due diligence
How extraction, tiering and screening work across the document set.
Deep due diligence
Workstream templates, evidence binding and reference tracking in depth.
Due diligence platform
The platform view: what runs where, and on which investment record.
Due diligence automation
Which parts of diligence are automated, and which stay with people.
PE due diligence software
QofE summaries, debt structure and operational diligence for buyouts.
Best AI due diligence software
A direct comparison against generic assistants and traditional data rooms.
Questions we are asked in diligence demos
Which documents can be processed?
Financial statements, management accounts, cap tables and instrument documents, contracts, board and governance material, technical documentation and regulatory filings. Parsing is native rather than a generic text extraction, so structure is preserved.
Can the same company be diligenced more than once?
Yes. Findings, sources and open questions stay on the investment record, so a later round or a follow-on assessment starts from what was previously underwritten rather than from an empty file.
Does it handle assets that are not operating companies?
Yes. Evidence requirements are configured per asset class, which is how projects, facilities, real assets and funds are assessed on their own terms rather than through a company template.
How does this connect to fund level diligence?
Company level findings are the holdings layer beneath a fund. When a fund is the subject, its existing positions can be read through to this level rather than accepted at the summary the manager provides.