The difference between document-first and fact-first analysis is the difference between reviewing evidence and knowing what it proves
When you review documents, you’re not really interested in the documents themselves. You’re trying to establish facts inside them: what happened, when it happened, whether other sources agree, and whether the evidence holds together. That’s true in litigation, insurance claims, investigations, journalism, law enforcement, or any case where decisions depend on evidence.
Most AI document tools still operate at the document level. They retrieve, summarise or tabulate text, then rely on the lawyer to infer what the underlying facts are, whether they are complete, and whether they are consistent across the record.
That doesn’t work in complex cases, where facts rarely sit in one place. Context is often established many pages before a key statement appears, and the same event is described across emails, attachments, transcripts, and third party records. Analysed in isolation, those fragments never quite connect. You may have all the relevant material, but not a clear view of what the evidence actually supports, which is why you still end up reconstructing the facts manually using spreadsheets, Word chronologies, and assumptions.
Starting from facts, not documents
Wexler starts from the premise that facts, not documents, should be the primary object of analysis.
Imagine defending a multi party fraud claim alleging misstatements of asset valuations over a three year period, supported by a production of 80,000 documents. Relevant material sits across board packs, valuation committee minutes, emails between multiple individuals, late produced messages, third party reports using inconsistent terminology, privileged material, foreign language documents, and spreadsheets where the decisive figure is buried deep inside cell BX47 of a tab labeled "Backup (old).”
According to the claimants, your client knew. According to your client, reliance was placed on external advisers. According to those advisers, concerns were raised. No single document resolves the issue, and the record is not organised around these facts that will decide the case.
By extracting relevant facts as discrete, source linked units and assembling them into a live chronology, Wexler shows exactly where accounts diverge. It becomes clear who says what, when, and which documents support or undermine each position, surfacing the inconsistencies that ultimately determine outcome.
Here’s how that facts-first approach works under the hood.
1. Document classification preserves context from the outset
Different documents behave differently, and need to be handled accordingly. A witness transcript might rely on context set many pages earlier. A contract could depend on defined terms. A spreadsheet contains the key financial fact in tab E147. An email chain unfolds over time.
Wexler identifies what each document is and routes it through a processing path designed for that type. Context is preserved from the start, so key statements are interpreted as they were intended, not in isolation.
2. Per-document extraction captures every fact at source, at scale
Once documents are classified, facts are extracted from each one independently. Events, dates, people, actions, and assertions are identified as they appear in the source.
Because this step runs in parallel across large document sets, thousands of documents can be analysed at once. Work that would take weeks of manual review can be done in hours, without sampling or shortcuts, giving you early visibility into the full evidential picture.
3. Cross-document consolidation connects related evidence into a single, coherent factual record
This is where factual analysis usually fails, and where Wexler focuses most of its effort.
If one document says Adam Smith flew to Paris, another confirms the booking, and a third shows the boarding pass, Wexler recognises these as multiple sources for the same event, not separate facts. Those references are consolidated into a single factual record, with every supporting source retained.
This means events are represented once, accurately, and backed by all available evidence, rather than duplicated, fragmented, or missed entirely.
4. Contextualisation highlights what matters most without losing the full record
Not every fact matters equally.
Once facts are extracted and consolidated, they’re classified by type and significance, then situated within the context that matters for your case. That might be the issues in dispute or the coverage questions in a claim.
What structured facts make possible
Instead of working with document fragments, you work from a complete factual record. Relevant events, dates, and assertions are identified once and kept connected as the evidence evolves.
That makes inconsistencies visible early. If one source places a meeting on Tuesday and another on Thursday, or figures don’t reconcile across reports, you see it directly in the record. These are the issues that decide outcomes, and they’re the ones document-first analysis routinely misses.
That upfront investment in facts pays off everywhere downstream, from chronologies and examination prep to claim assessment and advice. If your work depends on understanding what actually happened across large volumes of evidence, starting with facts is the only approach that holds up under pressure.