Flagship use case

Scope and price new matters from what past matters actually cost.

Every firm sits on years of matter narratives — time entries, billing descriptions, closing memos — that quietly record how the work really unfolded. We use embeddings to read that history, extract the real phases of work inside each matter, and turn it into accurate budgets and defensible pricing for the matters you take on next.

Most firms price new matters from memory, gut feel, and a partner's best guess. The data to do better already exists — buried in free-text narratives that no spreadsheet can read. The problem was never a lack of history; it was that the history was unstructured. Embeddings change that: they let a machine understand what a narrative means, group the work into recognizable phases, and compare a new matter to the ones that genuinely resemble it.

The outcome: a phase-by-phase budget for a new matter, grounded in what comparable matters actually took — with the ranges and confidence to defend it to a client and to your finance committee.
How it works

From free-text narratives to a defensible budget.

Five moves turn unstructured matter history into a scoping and pricing engine.

1

Read the narratives

Time-entry descriptions, matter summaries, and closing memos hold the real story of each matter — what was done, in what order, and how long it took. We bring that text together as the raw material.

Input  Historic matter narratives, time and fee records, matter metadata (type, client, outcome).

2

Embed for meaning, not keywords

We convert every narrative into an embedding — a numerical representation of its meaning — so the system can tell that "reviewed first-round disclosure" and "examined documents produced by opposing counsel" describe the same kind of work, even with no shared words.

Why it matters  Keyword matching misses how differently lawyers write. Meaning-based matching doesn't.

3

Extract the phases of work

Against a phase taxonomy built with your practice groups (e.g. due diligence, drafting, negotiation, closing; or pleadings, discovery, motions, trial), we classify each narrative into the phase it belongs to — reconstructing the real shape of every past matter.

Result  Each historic matter is broken into phases, with the hours, fees, and elapsed time each phase consumed.

4

Build the cost library

We aggregate phase-level cost and effort across the firm's history — not as a single average, but as distributions with the drivers that move them (deal size, jurisdiction, counterparty, complexity). You see the typical case and the tail risk.

Result  A living library: "for this kind of matter, discovery runs X–Y hours, and here's what pushes it higher."

5

Scope & price the next matter

For a new matter, we describe it, embed that description, and retrieve the genuinely comparable historic matters. The engine assembles a phase-by-phase budget from real precedent — with ranges and a confidence signal — instead of a number pulled from the air.

Output  A defensible budget and price, phase by phase, traceable to the matters it's based on.

The project roadmap

From raw matter history to a working pricing engine.

Like all our data work, governance comes first — matter narratives and financials are highly sensitive, so access, conflicts, and confidentiality are designed in before any data is read.

Phase Name Core question it answers Typical duration*
0Discovery & Data AccessWhere does our matter history live, and how do we access it safely?2–3 weeks
1Phase Taxonomy DesignWhat are the real phases of work in our practice areas?2–3 weeks
2Embed & ClassifyHow do we turn narratives into phases at scale?3–4 weeks
3Cost & Effort ModelingWhat does each phase actually cost — and what drives it?3–4 weeks
4Scoping & Pricing EngineHow do we price a new matter from comparable history?3–5 weeks
5Validation & EmbedIs it accurate, and can the firm rely on it?2–4 weeks

*Indicative; scoped per firm and dependent on data quality. Phases overlap in practice.

Phase by phase

What happens in each phase.

0
2–3 weeks
Discovery & Data Access

Objective — Locate the firm's matter history and establish safe, governed access to it.

  • Identify sources: billing narratives, time entries, matter-management records, closing memos, engagement letters.
  • Confirm data quality and coverage — which practice areas have enough history to model.
  • Apply confidentiality, conflicts, and access controls before any narrative is read.
  • Agree what "good" looks like: the matter types and pricing decisions to target first.
Data access map + prioritized matter types + governance sign-off
1
2–3 weeks
Phase Taxonomy Design

Objective — Define the real phases of work for each practice area, with the people who do it.

  • Work with practice-group leaders to name the phases that actually structure their matters.
  • Capture the markers that distinguish phases in how lawyers write narratives.
  • Reconcile differences across practice areas into a coherent, firm-wide taxonomy.
  • Validate the taxonomy against a sample of real matters before scaling.
Agreed, validated phase taxonomy per practice area
2
3–4 weeks
Embed & Classify

Objective — Turn every historic narrative into a phase, at scale.

  • Generate embeddings for all matter narratives and descriptions.
  • Classify each entry into its phase using embeddings plus an LLM, with human review on edge cases.
  • Roll phase classifications up to the matter level, attaching hours, fees, and timelines.
  • Measure classification accuracy against the validated sample from Phase 1.
Full matter history reconstructed into phases
3
3–4 weeks
Cost & Effort Modeling

Objective — Quantify what each phase costs and what moves it.

  • Aggregate hours, fees, and duration per phase across matter types.
  • Build distributions, not just averages — so the typical case and the tail risk are both visible.
  • Identify the drivers (deal size, jurisdiction, counterparty, complexity) that shift each phase.
  • Flag write-offs and overruns to learn where past estimates went wrong.
Phase-level cost library with ranges and drivers
4
3–5 weeks
Scoping & Pricing Engine

Objective — Produce a defensible budget for a new matter from comparable history.

  • Embed a new matter's description and retrieve the genuinely comparable past matters.
  • Assemble a phase-by-phase budget from real precedent, with ranges and a confidence signal.
  • Let the user adjust drivers and see the budget respond.
  • Show the matters each estimate is based on, so it can be explained and defended.
Working scoping & pricing tool, grounded in firm history
5
2–4 weeks
Validation & Embed

Objective — Prove it's accurate and make it part of how the firm scopes work.

  • Back-test the engine against closed matters with known actuals.
  • Integrate into the pitch, engagement-letter, and budgeting workflow.
  • Set up a feedback loop so each new matter improves the next estimate.
  • Train partners and BD to use it and read the confidence signals.
Validated engine + an enabled team + a learning loop
What it delivers

Pricing that wins work — and protects margin.

Accurate budgets

New-matter estimates grounded in what comparable matters actually cost — not a guess that gets written off later.

Defensible pricing

Fixed-fee and AFA proposals you can stand behind, with the precedent to justify them to clients and committees.

Margin protection

See the tail risk before you commit to a fee, so the phases that blow budgets are priced for, not absorbed.

Institutional memory

Knowledge that used to leave with a retiring partner is captured, structured, and reusable across the firm.

This sits directly on the data-readiness foundation. Once matter data is clean, governed, and machine-ready, matter intelligence is what that foundation pays for.
Ready when you are

Turn your matter history into your pricing advantage.

We advise and implement — we build the engine on your data, validate it against your actuals, and stay until your partners are scoping with it.