Pre-launch · founding cohort forming

Read the contract you were actually sent.

Clauseloom scores an incoming agreement clause by clause against your own playbook — matched, deviating, or missing — and shows the source passage next to the precedent that justifies the call. Nothing is asserted without a citation you can click.

  • Built from your redlines, not a generic corpus
  • Every flag carries a page and paragraph anchor
  • Redline back out to Word, tracked
A contract page with three clauses linked to their precedents An agreement on the left with three highlighted clauses — one matched, one deviating, one missing — each joined by a thread to a card on the right showing the playbook position that justifies the call. 12.1 GOVERNING LAW · MATCHED 8.2 LIABILITY CAP · DEVIATES Their cap only — asymmetric DATA PROCESSING · MISSING YOUR PLAYBOOK § G-02 Governing law Accepted 9 of 9 times § L-04 Liability Cap must be mutual Conceded once, Q2 2024 § D-01 Data terms DPA required where personal data is shared 3 positions cited · 0 uncited assertions
Illustrative view — sample agreement, sample playbook
Shaped around the agreements you see weekly
  • MSAs
  • SaaS order forms
  • DPAs
  • NDAs
  • Statements of work
  • Reseller agreements
  • Vendor paper

The situation

The same twelve contracts, over and over.

Mid-market legal teams do not have a novelty problem. They have a repetition problem — and the cost of that repetition is paid in hourly billing, in stalled deals, and in knowledge that never left one person’s head.

A model that cannot show you where a sentence came from is asking you to sign on its behalf. Clauseloom’s working position
  1. 01

    Outside counsel re-reads boilerplate at partner rates

    The fourth NDA this quarter is materially the third one. Someone still bills for reading it end to end, because nobody can prove which parts are unchanged.

  2. 02

    In-house counsel becomes the bottleneck for sales

    A single reviewer holds the queue. Deals wait behind a first-pass read that a well-instructed process could have done days earlier — and the reviewer knows it.

  3. 03

    “We never accept unlimited liability” lives in one head

    The institutional position is real, consistently applied, and entirely undocumented. When that lawyer is on leave, the position goes with them.

  4. 04

    Generic AI summarisers produce fluent, unauditable paragraphs

    A confident summary is not a review. Without the passage it was drawn from and the precedent that justifies the call, a lawyer has to re-read the whole thing anyway — so nothing was saved.

What Clauseloom does

Your positions, woven into one readable fabric.

A loom takes separate threads and makes cloth. Clauseloom takes your redline history, your negotiated positions and your standard template, and turns them into a playbook that every incoming agreement is measured against — clause by clause, with the working shown.

How Clauseloom turns past redlines into a scored contract Four stages left to right: your redline history feeds a playbook; an incoming contract is compared against that playbook; the result is a clause-by-clause score where each flag cites its source. A return path shows that every position you confirm updates the playbook. INPUT Your redline history Signed sets · tracked changes THE LOOM Your playbook Positions and fallbacks INCOMING Counterparty paper Word or PDF, their draft OUTPUT Scored clauses Matched · cites § Deviates · cites § Missing · cites § Exports back to Word Every position you confirm updates the playbook
The loop is the point: the playbook is not configured once by a consultant, it is written by the decisions your team already made and keeps making.

Load once, not per document

Point Clauseloom at the folder of agreements you have already negotiated. It reads what you struck, what you inserted, and what you accepted without comment — and proposes the positions it thinks you hold.

Three states, not a paragraph

Every clause resolves to matched, deviating, or missing, with a severity your team defined. A reviewer can scan the deviations and ignore the rest with a clear conscience.

Every call carries its citation

The flag links to the exact passage in the incoming document and to the prior negotiation that set your position. If we can’t cite it, we don’t claim it.

Anatomy of a flag

See how a deviation is flagged.

This is the review surface: the counterparty’s clause, the state Clauseloom assigned it, and the two links that let a sceptical lawyer check the work in under ten seconds.

  1. A state, not an opinion

    Matched, deviates, or missing. Severity is drawn from the tier your playbook assigns to that clause family — not from the model’s sense of drama.

  2. The passage, quoted verbatim

    The exact sentence from their paper, with the operative words marked and a page-and-paragraph anchor that opens the source document at that spot.

  3. The precedent behind the call

    Which playbook position applies, when it was set, and the negotiation where you last held or conceded it. That link is the difference between a review and a guess.

  4. A fallback you have already approved

    Where your library has one, the accepted alternative language is one click away — with the approval tier that says who can send it without escalating.

Master Services Agreement — counterparty paper, v4.docx Review

Clause map

  • 1–7 Commercial terms
  • 8 Liability
  • 9 Indemnities
  • 10 Data & security
  • 11–14 General
31Clauses read
4Deviations
1Missing
§ 8.2 Limitation of liability Deviates Severity: high

“…in no event shall Supplier’s aggregate liability exceed the fees paid in the twelve (12) months preceding the claim, provided that Customer’s payment and indemnity obligations shall not be subject to any cap.”

The cap is one-directional. Your playbook requires a mutual cap, with carve-outs limited to confidentiality and wilful misconduct.

§ 12.1 Governing law Matched Accepted 9 of 9

“This Agreement shall be governed by the laws of England and Wales…”

Illustrative interface with sample data. No real agreement, counterparty or client is shown.

Getting started

Four steps, and none of them are “re-key your templates”.

  1. STEP 01

    Hand over the history

    Drop in signed agreements, marked-up drafts and your standard template. Clauseloom reads the tracked changes, not just the final text — the edits are where the positions live.

  2. STEP 02

    Confirm the playbook

    You get a draft playbook: one row per position, each with the redlines that suggested it. Keep, edit or reject each one. Nothing enters the playbook without a human confirming it.

  3. STEP 03

    Run the incoming contract

    Upload their paper. You get the clause map, the deviations ordered by severity, and the missing clauses your playbook expects to see — each with its citation.

  4. STEP 04

    Send the redline back

    Accept the fallback language you have already approved, edit what you want, and export a Word document with tracked changes and comments in your house style.

What’s in the workspace

Seven things it does, and does completely.

This is the whole product surface. There is no eighth tab that turns out to be a roadmap item.

  • Playbook from your own redlines

    Positions are inferred from what your team actually negotiated, then confirmed by a human. Every row shows the documents that produced it.

    Ingests .docx tracked changes
  • Clause-by-clause deviation scoring

    Matched, deviating or missing, each with a severity drawn from the tier your playbook assigns. Sort the queue by what would actually stop a signature.

    Severity tiers you define
  • Every flag cites the source passage

    Page, paragraph and the quoted sentence. Click through to the exact spot in the original file. A flag without a citation is a bug, not a feature.

    Anchors survive re-upload
  • Side-by-side against your template

    Their clause and your standard clause in one view, with the differences marked at word level. No tab-switching between two Word windows.

    Word-level diff
  • Negotiation history

    What you conceded, to whom, and when. When a counterparty says “you agreed to this before”, you can see in one line whether that is true.

    Per-position timeline
  • Fallback library with approval tiers

    Pre-approved alternative language, ranked from preferred to last resort, each tagged with who may send it without escalating to the general counsel.

    Tier 1 · 2 · escalate
  • Export a redline back to Word, with tracked changes

    The output is a document your counterparty’s lawyer opens in Word and reads normally: tracked insertions and deletions, comments in your house voice, styles intact. Clauseloom’s job ends where your negotiation resumes — it does not ask anyone outside your team to log into anything.

    A Word document with tracked changes exported from Clauseloom A document page showing a struck-through phrase, an inserted replacement in underline, and a margin comment bubble. TRACKED CHANGES · COMMENT · YOUR STYLES

What we will hold ourselves to

We would rather be measured than described.

Clauseloom has not launched, so we have no customer numbers to show you and we are not going to invent any. What we can show you is exactly what the workspace counts from day one — the three figures a founding team should judge us on.

Contracts reviewed

Volume through the workspace, by contract type and by reviewer. The denominator for everything else — and the first number we will publish once a cohort has run long enough for it to mean something.

Bar chart of contracts reviewed per week Sample data from a test workspace showing weekly review volume rising across eight weeks. Wk 1 Wk 8

Illustrative — sample data from our own test workspace.

Median hours returned per week

Reviewer time before Clauseloom against time after, on the same contract types, measured per team rather than averaged across everyone. If the line does not move, the tool is not working.

Two lines comparing review hours before and after Sample data from a test workspace: a flat upper line for the baseline hours and a declining lower line for hours after adoption. Baseline review hours With Clauseloom

Illustrative — shape of the chart, not a result we have achieved.

Deviation catch rate vs. a human baseline

Contracts reviewed twice — once by your team, once by the workspace — and compared clause for clause. We will report the misses as loudly as the catches, because the misses are what a lawyer needs to know.

A dial comparing flags found by the workspace against a human review A partial ring with a segment marked as caught, a smaller segment marked as missed, and a legend beneath. Flagged by both Found only by the lawyer Found only by Clauseloom

Illustrative — the comparison we run, with no score claimed yet.

What you can expect from a pre-launch tool

  • No named customers on this page, because we do not have any to name yet. When we do, they will be named with permission or not at all.
  • The founding cohort is small on purpose: a limited number of teams, onboarded one at a time, with direct access to the people building it.
  • Your documents are yours. We do not train shared models on your agreements, and your playbook is not visible to any other tenant.
  • Clauseloom is a first-pass reviewer, not counsel. It does not give legal advice and it is not a substitute for a lawyer’s judgement.

Pricing

Per seat, annually. No metering per document.

Charging by the document punishes the behaviour we want: running every agreement through, including the ones you assume are fine. So we don’t.

Volume

Talk to us

A contract-volume tier for teams past 500 agreements a year, or with more than one playbook to maintain.


  • Everything in Practice
  • Separate playbooks per entity, region or business line
  • SSO, audit log and retention controls
  • Named contact and quarterly playbook review
Ask for a quote

Playbook onboarding

$4,000 one-time, from

Optional. We build the first playbook with you from your history, then hand it over.


  • We read your archive and draft the positions
  • Two working sessions with your reviewers to confirm them
  • Fallback language drafted for your top ten clause families
  • Teams who prefer to build their own pay nothing extra
Scope an onboarding
  • Annual contracts, invoiced once
  • No per-document charges, ever
  • Founding-cohort pricing is held for the first two years
  • Prices in USD, excluding tax

Questions a lawyer asks first

The sceptical version.

Does this replace a lawyer’s review?

No. Clauseloom does the first pass — the part that is comparison rather than judgement. It tells you which clauses match your positions, which deviate and by how much, and which are absent. Deciding what to do about any of that is legal work, and it stays with you.

The design assumption throughout is that the reviewer does not trust the model. That is why every flag is a link to a passage rather than a paragraph of prose.

What happens to our documents?

They stay in your workspace. We do not train shared models on customer agreements, and no other tenant can see your playbook, your history or your files. Retention is configurable, and deletion means deletion — including from the derived playbook rows that cite the document.

We will sign a DPA and answer a security questionnaire before you upload anything. Ask us for the current one on the contact page.

We don’t have a tidy redline archive. Can we still use it?

Usually, yes. Most teams’ archives are a shared drive of near-duplicates with inconsistent naming, and that is what the ingestion is built for. Twenty to thirty negotiated agreements across your main contract types is enough to draft a first playbook worth arguing with.

If you genuinely have nothing, you can write positions directly instead — it is slower, but the workspace does not require history to function.

Which contract types does it handle?

Commercial agreements on either side of the table: MSAs and their order forms, SaaS and vendor terms, NDAs, DPAs, statements of work, reseller and partner agreements. It is deliberately not a general-purpose document reader — the value comes from knowing the clause families in the contracts you actually see.

What does the model actually do, and where can it be wrong?

It segments the document into clauses, matches each one to the closest position in your playbook, and classifies the relationship. It is most reliable on clause families you have negotiated repeatedly, and least reliable on bespoke commercial schedules that have no analogue in your history.

When it is unsure, it says so and shows you the two candidate positions rather than picking one quietly. Unsure is a state we would rather surface than hide.

Can we start now?

Not quite. Clauseloom is pre-launch and onboarding a small founding cohort one team at a time, so we can sit with each of them properly. If you send us a contract type you review constantly, we will tell you honestly whether you are a good fit for the first group or better served by waiting a couple of releases.

Founding cohort

Bring the contract you are most tired of reading.

Send us one agreement type and the positions you always hold. We will show you what a Clauseloom review of it looks like — flags, citations, precedents and all — and you can tell us where it is wrong.

  • A reply from a person within one business day
  • No document uploads until a DPA is signed
  • No sales sequence, no drip campaign