Five layers, from raw data to decisions.
Each layer builds on the one below it. Your data becomes facts, the facts become a connected model of your business, and a rules engine applies your firm's logic the same way every time, with the evidence attached.
Follow one number up the stack: a Q3 capital account statement shows a management fee of $1.24M.
Every case ends in one of four outcomes, with the evidence attached.
ExampleFlagged for review: $1.24M charged against $1.20M allowed by the LPA, a $40K difference, with both pages attached.
- ✓Resolved by the rulesThe rules settle it, and the result is recorded with its sources.
- !Flagged for reviewA person sees the exception, with the evidence from both sides.
- ✉Request draftedA follow-up to an investor, client or counterparty, ready to approve.
- ↑Escalated to an expertJudgment calls go to the right person. Their decision becomes a rule.
Every expert decision updates the rules
Tens of thousands of deterministic rules, learned from your documents and procedures. Every result links to its evidence, and the rules keep improving as your experts approve or correct them.
ExampleFee = 1.5% of commitments (LPA §8.1), or $1.20M. Investor C pays no fee on co-invest (side letter §4).
Links facts and concepts across every source, so “MGMT fee” in one file and “management fee” in another mean the same thing.
Example“MGMT fee” in the general ledger and “management fee” on the statement are one concept: Fund III, Investor C, Q3.
Pulls the facts out of every document and record, each one tied to the page it came from.
ExampleManagement fee: $1.24M · capital account statement, p. 4
PDFs, spreadsheets, emails, transactions and legacy systems. Structured and unstructured, often a decade or more of it.
ExampleQ3 capital account statement (PDF), the general ledger, the LPA and every side letter.
Start from what you already have.
There is no data-preparation project before you start. Segmental reads the documents, spreadsheets, emails and system records your team already works from, including years of history.
Facts, each with its source.
Extraction is the plumbing, not the product. What matters is that every fact keeps a link to the exact page and field it came from, so anything built on top can be checked.
One meaning for every concept.
Segmental resolves the same entity, term or number across every source into one connected model of your business. When sources disagree, it follows the precedence your firm sets instead of averaging them into a plausible answer.
Rules, not guesses.
This is what sets Segmental apart. Instead of asking a model to guess, Segmental learns deterministic rules from your documents, procedures and past decisions, and applies them the same way every time. When documents change or an expert corrects a result, the rules update themselves. First rules are ready to review within days of starting.
Every case ends in a decision you can check.
Whatever the workflow, the rules settle what they can and send the rest to a person, with the evidence attached. What an expert decides becomes a rule for the next case.
What sits on top depends on who you are.
Each team sees only the agents and products for its own work, all on the same foundation.
Three reasons firms trust the output
Open to a side-by-side test against other vendors, on your own documents.
See the rules engine on your own documents.
Bring a handful of your documents and procedures. First rules are ready to review within days.