Ciceri reads every document in the data room overnight. Categorized, extracted, every answer pinned to the passage it came from. Spend the time on judgment, not filing.
13 categories tuned to corporate & M&A diligence: Material Contracts; Company, Structure, Corporate & Regulatory; Finance & Strategy; Labor & Employee Matters (including benefit plans); Taxes; Intellectual Property; IT & Software; Data Protection (privacy & security); Real & Personal Property; Legal; Insurance; Marketing; Other. Rename them, restructure them, or swap in your firm's own. The categories are yours.
Diligence on a real deal looks like this. Thousands of documents land in a data room over a weekend. A few hundred on a small carve-out; ten thousand on a take-private. By Monday, three lawyers on the deal team have a checklist, a coffee pot, and ten days to signing.
They will spend most of that runway doing what is, in practice, a first pass: opening every file, deciding what it is, where it goes, what to flag, what to skip.
It is necessary work. It does not require a J.D., and it should not be billed at the rate of one.
Pick any buyer-side deal. Same Tuesday morning, same data-room drop. The week that follows looks nothing alike.
Upload the data room. Ciceri does the first pass: it sorts every document into your taxonomy, lifts the fields you care about, and stamps every output with a citation back to the source. Your team picks up at the part that requires a lawyer.
Ciceri reads each file end-to-end and classifies it against your category tree. Ambiguous documents surface in a review queue with the suggested category and why. Set a category by hand and Ciceri won't overwrite it. Not on the next run, not ever.
Define a field once, whether that's counterparty, governing law, term, or indemnity cap, and Ciceri lifts it from every document of that type. Export to Excel. Hand it to the deal team. Drop it into the disclosure schedule.
Every extracted value carries its source: where it sits in the document, and the exact passage behind it. Read the quote beside the answer, or open the document and check it in context. Nothing has to be taken on faith.
A document view that shows the answer and the proof beside it. A request list that knows what's checked off and what's still outstanding, by category, by counterparty, by partner.
Categorization is the easy half. The work that actually moves a deal is what's below: questions answered with citations, comparison across the whole deal, an audit trail you can defend.
Pick a clause: change-of-control trigger, IP carve-out, anti-dilution adjustment. Ciceri checks every contract, lists the ones that have one, and shows the extracted consequence right there. Search the full text too, when you need it. Every contract gets checked; only the ones that need a lawyer land on your desk.
Type the question the way you'd say it out loud. The search agent turns it into a filter you can see, plain chips you can read and edit, and runs it across the deal. No black box. Every row is a real document; every value is cited. One click saves it as a view your whole team shares.
Permissions live on the deal, not the firm: View, Edit, Approve, Export, Delete, Access, each granted per member. A buyer-side partner can't see, search, or get alerts on the seller-side deal in the same firm's account. Ciceri refuses to remove the last admin, so the deal can't lock itself out. Every grant on the record.
Indemnity caps from every contract. Governing-law clauses by counterparty. Change-of-control consequences by trigger. Every extracted value, comparable across the whole deal in one filterable, sortable view, whatever the document type.
Change-of-control consequence on every MSA. Assignment restriction on every lease. Anti-dilution adjustment on every stock purchase. Every field your firm tracks, defined once and run on every deal. Every team starts from the same place, and the disclosure schedule stops depending on who staffed the deal.
Every approval, every override, every export is recorded with the person, the timestamp, and the value before and after. The trail builds itself as you work; nobody reconstructs it the night before a deposition.
Lawyers are trained to exercise judgment.
They are not trained, and should not be paid, to be a clerk.
The technology to fix this exists. The product, until now, has not.
Confidentiality. Defensibility. The right of a lawyer to override the model. These are the parts we spent the most time on. They were settled before the first feature shipped, and settled on the side of the partner who has to defend the work.
No document, no search result, no notification ever crosses from one firm's account into another. We test that boundary on every change; nothing that fails goes live.
Every AI-extracted term links to the document and passage it came from, where you can read it and override it. When the model isn't sure (a rough scan, an odd margin note), the value goes to a human to set rather than being guessed.
A lawyer sets the final answer. Ciceri proposes; counsel disposes. Once you've reviewed a document, the system honors that decision, even if the model updates later.
Encrypted at rest and in transit using current industry standards. Never used to train models, ours or our vendors'. Your documents don't leave your account, and they don't make anyone else's product smarter.
Access is per-deal, not per-firm. A partner on the buyer side cannot see, search, or be alerted on the seller-side deal living in the same firm's account. Ethical walls that hold by permission, not by accident.
If you leave, you get your documents and your extractions back in a format you can use. Your work product goes with you.
Worth being clear about the perimeter. None of these are on the roadmap; none are coming. They're not in scope, by design.
Or markup, or redline. Ciceri reads. Your team writes.
The model proposes; the lawyer decides. Every output is reviewable and overridable on the way to the export.
We change where they start the morning. The first pass is done by sunrise; the lawyer's day begins at the analysis, not the index.
Models are wrong sometimes. Sources are scanned sometimes. The workflow knows it, flags it, and asks a human. No silent guesses.
We charge for AI work. Categorizing a document, extracting a field, or running a query costs one Diligence Unit (DU), at one flat rate we share when we onboard you. No seats, no platform fee, no minimums.
No volume tiers. No per-seat fees. The price your two-attorney boutique sees is the price your 800-attorney firm sees.
Try before you buy. Run a small deal end-to-end before you commit. Walk away during the trial and owe nothing. No procurement gauntlet.
Pre-paid balance, transparent decrement. We never bill you for something you've already used.
Every deal carries a usage report: one line per AI call, tied to a specific document. Flat per-unit pricing keeps ABA 93-379-style pass-through math simple for your billing team. No volume tiers, no end-of-year true-ups, nothing to disentangle. The bill you send the client is the bill we sent you.
Six of them. If yours isn't here, write to us and we'll add it.
We started Ciceri because we'd done diligence ourselves and watched the same absurdity repeat on every deal: a roomful of trained lawyers, billed at full rate, spending their first week doing the work of a clerk.
The technology to fix this has existed for a while. The product hasn't. Nobody had built one around the facts of the job: a lawyer defends every answer, confidentiality is non-negotiable, and the model is sometimes wrong, so the workflow has to know it.
So we built it.
If the first pass is eating your team's week, we'd like to show you what we've built. Twenty minutes. We'll do the talking.