Ivo

Every in-house legal team has the same bottleneck and it is not interesting work. Vendor agreements, NDAs, order forms and data processing addenda arrive faster than anyone can read them, and most are ninety per cent identical to the last one.

The skill is spotting the ten per cent that differs, and specifically the parts that differ from what your company will accept — which is not the same as what a textbook considers standard. Your company may care intensely about liability caps and not at all about governing law, or the reverse.

That distinction is where most contract AI falls down and where Ivo's design starts. It reviews against your playbook rather than a generic one, and it does it in the document where lawyers already work.

What Does Ivo Do To A Contract?

• It applies redlines in Word or Google Docs, in the document itself, rather than producing a report you then have to act on.
• It reviews against your playbook, your historical agreements and the deal context — not a generic standard.
• Clause benchmarking by governing law, contract type, party role and industry, which is context a junior lawyer would not have.
• Four products: Review for redlining, Assistant for research and comparison, Intelligence for querying executed contracts, and Collaborate in early access.
• SOC 2 and ISO 27001 certified, GDPR and CCPA compliant, with a stated commitment never to train AI models on customer data.

Why Does Reviewing In Word Matter So Much?

Ivo applies redlines directly in Microsoft Word and Google Docs. That sounds like an integration detail and it is the reason lawyers will use it.

Contract negotiation happens in Word with tracked changes. It has for thirty years, counterparties expect it, and no legal technology has changed that — because the counterparty's lawyer at another company is not going to log into your platform.

Tools that produce a separate review report create work rather than removing it: someone reads the report, decides which findings to accept, and then makes the changes in Word by hand. The analysis was automated and the labour was not.

Applying redlines in the document itself removes that translation step entirely. What comes back is a marked-up contract you can review, adjust and send — the same artefact you would have produced yourself, faster.

The thing to test is fidelity rather than existence. Does it preserve the counterparty's existing tracked changes and comments? Does formatting survive? A tool that produces clean redlines on a clean document and mangles a third-round negotiation draft has solved the easy half.

Working inside Word is the difference between a tool lawyers adopt and one they trial. Platforms like Ironclad ask the organisation to move its process; this asks nobody to move anything.

The counterparty's lawyer is never going to log into your platform. That single fact is why contract AI has to work inside Word.

Your Playbook Beats A Generic Standard

Ivo evaluates contracts against playbooks, historical agreements and deal context, and can convert your templates and repository contracts into working playbooks.

That last capability is the practical unlock. Most legal teams do not have a written playbook — they have institutional knowledge about what the company accepts, held by two or three people and applied inconsistently by everyone else.

Deriving a playbook from agreements you have already signed turns that tacit knowledge into something explicit and checkable. It also surfaces an uncomfortable and useful fact: what you actually agreed to is often not what you believe your position is.

Reviewing against historical agreements rather than a standard is the correct approach for a second reason. Your negotiating position depends on who you are dealing with — a large customer extracts terms a small vendor does not — and a generic reviewer flagging every deviation from an ideal produces noise rather than judgement.

Clause Benchmarking Is The Genuinely New Capability

Ivo benchmarks clauses by governing law, contract type, party role and industry. This is the feature that gives a reviewer something they could not easily get otherwise.

The question a lawyer asks constantly and can rarely answer with data is whether a term is market. Is a twelve-month liability cap normal for a software agreement under English law where we are the customer? Experienced practitioners develop intuition about this; nobody has statistics.

Segmenting by all four dimensions is what makes the answer meaningful. Market terms differ sharply between jurisdictions, between contract types, between being the supplier and the customer, and between industries. A benchmark that ignores those is an average of incomparable things.

The caveat to hold is that any benchmark reflects the population it was computed from, and that population is not disclosed. Whether it is drawn from public filings, customer agreements or licensed data changes what the number means — and a benchmark drawn mostly from large-company agreements will mislead a small one.

Ask about the basis directly. It does not make the feature less useful; it changes how much weight a specific number should carry in a negotiation.

Benchmarking is also where it diverges from drafting assistants. Spellbook helps you write the clause; benchmarking tells you whether the clause you were sent is unusual.

• Ask what population the clause benchmarks are computed from, since that determines what they mean.
• Check benchmarks in your specific governing law rather than accepting a global figure.
• Confirm the party-role distinction is applied, because supplier and customer positions are not comparable.
• Use benchmarks to inform judgement rather than to settle a negotiation, which they cannot do.
• Test whether the playbook derived from your own contracts matches what you believe your position is.

The Data Commitment Is Stated Clearly

Ivo states SOC 2 certification, ISO 27001 certification, GDPR and CCPA compliance, and that it never trains AI models on customer data.

That last sentence is the one that matters most and the one legal teams ask about first. Contracts are among the most commercially sensitive documents a company holds — pricing, terms, customer identities, negotiating positions — and a tool that learned from them would be learning from every customer's confidential material.

Stating it plainly rather than burying it in terms is the right behaviour, and it is not universal in this category. The follow-up for procurement is to confirm the commitment appears in the contract rather than only on the website, because a marketing statement is not an enforceable obligation.

The two certifications cover different things and both are useful: SOC 2 attests to operational controls, ISO 27001 to a maintained information security management system. A Trust Center is referenced from the footer, which is where the reports would live.

The customer list makes the security posture more credible than the certifications alone do. Meta, Uber, IBM, Intel, Coinbase and Atlassian all run vendor security reviews that are considerably more demanding than a typical questionnaire, and passing them repeatedly is evidence a badge is not.

Data commitments are worth reading side by side across vendors. Luminance and Harvey AI both publish them, and the differences matter for privileged material.

Testing Contract AI Against Reviews You Already Did

The only honest way to evaluate contract review is against agreements where you know what the right answer was, because a plausible-sounding redline on an unfamiliar contract tells you nothing.

The bar is step three. A reviewer that misses something your team caught is a reviewer you cannot leave unsupervised on that agreement type.

1. Take ten agreements your team reviewed in the last six months, keeping your own markup separately.
2. Let Ivo derive a playbook from your executed contracts, then read that playbook. It will tell you what you actually agree to, which is frequently a surprise.
3. Run the ten and compare its redlines with yours. Count the misses and the false flags separately — they have different consequences.
4. Test on a third-round negotiation draft with counterparty tracked changes and comments, not a clean document. Fidelity on messy documents is the real test.
5. Check a clause benchmark against something you know the market position on, in your governing law, and see whether it matches your experience.
6. Ask for reference calls before committing — with customers of this profile, you should get them.

The Details Ivo Keeps Behind Sales

No pricing appears anywhere on the pages we read, and access runs through a sales process. That is standard for legal technology sold to in-house teams.

More notable is the absence of any accuracy figure. Contract review has a measurable notion of correctness — did it catch the deviations a lawyer would — and no vendor in this category publishes it, Ivo included. The reason is presumably that the number depends entirely on the agreement type and the playbook, which is fair, and it still means the only evaluation is your own.

No headquarters location appears on the pages we read either, though a named legal entity, a registered trademark and a Trust Center are all present.

The Collaborate product is listed as early access, which is honest labelling and worth reading as such — early access means it is not yet something to build a process around.

We do not publish amounts. Request pricing alongside the SOC 2 and ISO reports, and ask specifically whether the no-training commitment is contractual rather than a website statement.

Ivo: The Product Facts

Which Legal Teams Ivo Fits

An excellent fit when

• Your bottleneck is volume of routine agreements rather than complex bespoke work.
• Your lawyers negotiate in Word and will not adopt anything that moves them out of it.
• You have institutional knowledge about acceptable terms that has never been written down.
• You want to know whether a term is market in your jurisdiction, not in general.
• Confidentiality is the blocking concern, and an explicit no-training commitment resolves it.

Not worth the setup when

• You need workflow, approvals, signature and obligation management, which is a full CLM rather than a reviewer.
• Your contracts are highly bespoke, where playbook comparison has little to compare against.
• You want published accuracy data before deciding, since none exists in this category.
• You need a self-serve tool you can try today, as access runs through sales.
• Your volume is low enough that review is not actually your constraint.