Ironclad Review

Contracts almost never fail because someone drafted a bad clause. They fail because the draft sat in an inbox for eleven days, went to the wrong reviewer, came back as a Word attachment with tracked changes nobody merged, and was signed against a template that had been superseded.

That is an operations problem wearing legal clothing, and it is why contract lifecycle management exists as a category at all — the value is in removing handoffs, not in improving prose.

Ironclad has covered all six stages for long enough to be a category leader, and the current question is what the AI layer adds on top. This review separates the workflow platform, which is well established, from the agentic review claims, which are newer and need more care.

Which Six Stages Does Ironclad Own?

• Six stages, one system: create, review, sign, store, analyze and fulfill — the whole contract lifecycle rather than one part of it.
• Jurist is the agentic reviewer, described as an AI contract partner purpose-built for legal contract review.
• Ironclad AI runs behind every stage rather than sitting as a separate assistant you have to remember to open.
• 20+ integrations including Salesforce, Microsoft Word and Teams, Slack, DocuSign, Adobe Sign, SAP, Coupa, HubSpot and Box.
• Unusually strong references: Salesforce, Databricks, Reddit, L'Oréal, Shell, Snap, the New York Times and the Associated Press, with Gartner and Forrester leader placements.

Why Cover All Six Stages Rather Than One?

Ironclad structures itself around create, review, sign, store, analyze and fulfill. Point tools exist for each of those, and the argument for one system is not feature breadth — it is that the failures happen between the stages.

A contract created in one tool, reviewed by email, signed in another, and stored in a third loses something at every boundary. Which version was signed. Whether the redline made it in. Where the executed copy went. Anyone who has hunted for a signed PDF across three systems knows the cost.

The stage most often underestimated is fulfil. A signed contract creates obligations — renewal dates, service levels, price escalators, notice periods — and in most organisations those live in someone's calendar until that person leaves. Auto-renewals that nobody caught are a recurring and entirely avoidable expense.

Analyze is the stage that turns the archive into an asset. Once contracts are structured data rather than PDFs, questions become answerable: how many agreements have uncapped liability, which renew in the next quarter, what our average payment terms actually are. Most legal teams cannot answer those today.

That is the real case for a platform over point tools, and it is a stronger one than any individual feature.

The alternative is a point tool aimed at the stage that hurts most. Spellbook lives in drafting and review and asks nothing of the rest of your process.

Contracts rarely fail on the clause. They fail in the eleven days the draft sat in an inbox, and in the executed copy nobody can find.

What Does An Agentic Contract Reviewer Actually Do?

Jurist is described as an agentic AI contract partner purpose-built for legal contract review. The word agentic is doing specific work and worth unpacking.

An assistant answers questions about a contract when asked. An agent works through a review — reading the document, comparing it against your playbook, flagging deviations, proposing redlines and routing what it cannot decide.

For high-volume standard agreements, that is genuinely valuable. Most in-house legal time goes on NDAs, vendor agreements and order forms that are ninety per cent identical to the last one, and the skill is spotting the ten per cent that differs.

The governance question follows immediately, and most legal teams have not answered it. Which agreement types may an agent review unsupervised? What deviation escalates to a human? Who is accountable when an agent approves a clause outside policy? A contract is a binding obligation, and the answer cannot be that the software decided.

The practical position: define the playbook and the escalation rules before switching agentic review on, not after. The technology is ready for this before most legal operating models are.

Integrations Decide Whether Anyone Uses It

Ironclad states more than twenty integrations, including Salesforce, Coupa, Ramp, HubSpot, Slack, Microsoft Word and Teams, Google Drive, Box, Zapier, DocuSign, Adobe Sign and SAP.

That list matters more than any AI capability, for an unglamorous reason: a CLM system only works if the people who are not lawyers actually use it.

Sales does not want to open a legal platform to send an NDA. They want it in Salesforce. Procurement wants it in Coupa. Everyone drafting still wants Microsoft Word, whatever anyone thinks about that. A CLM that requires people to leave the tool they live in gets bypassed, and a bypassed CLM is worse than none because it produces a false record.

The Word integration specifically is the one to examine closely. Contract negotiation happens in Word with tracked changes, and it will keep happening there for years. How cleanly a platform round-trips a Word document — preserving comments, redlines and formatting through external counterparty edits — determines whether lawyers accept it.

Test that specifically during evaluation, with a real counterparty redline rather than an internal one. It is where CLM implementations most often lose their users.

Integration depth is also what separates a platform purchase from an assistant purchase. A firm-facing tool like Harvey AI is bought by lawyers; a CLM is bought by legal operations and used every day by sales.

• Test the Microsoft Word round-trip with a genuine counterparty redline, not a clean internal edit.
• Confirm sales can start and track a contract entirely from within Salesforce or HubSpot.
• Check what data flows back to the CRM, since that is what makes adoption stick.
• Establish how executed documents reach your document management system.
• Ask which integrations are native and which run through Zapier, because the difference is reliability.

How To Read The Numbers On The Site

Ironclad publishes efficiency claims — contracting time reduction, adoption within 71 days, contracting cost reduction, template consolidation — and in the version we retrieved several of those percentages did not render, appearing without their figures.

That is a presentation problem rather than a substantive one, but it leaves a reader unable to evaluate the claims at all, which is worth noting for anyone reading the site rather than a deck.

The 71-day adoption figure is the one worth taking seriously if it holds, because implementation timeline is where CLM projects genuinely fail. Contract platforms have a reputation for eighteen-month deployments that finish after the sponsor has left, and a vendor putting a specific number on adoption is making a checkable commitment.

The far stronger evidence here is elsewhere. Salesforce, Databricks, Reddit, Shell, L'Oréal, the New York Times and the Associated Press are named as customers, alongside Gartner Magic Quadrant and Forrester Wave leader placements and a stated annual recurring revenue above 200 million dollars.

For enterprise software, that combination — demanding named customers, independent analyst placement and revenue scale — is considerably better evidence than any efficiency percentage, and it is what a procurement committee will actually weigh.

Vendor numbers deserve the same scepticism everywhere. Analysis platforms such as Luminance publish comparable figures, and none of them are independently audited.

Evaluating A CLM Without A Two-Year Project

CLM evaluations go wrong by focusing on features in a demo rather than on adoption in reality. This sequence tests the things that actually decide the outcome.

The bar is step two. If your lawyers reject the Word experience, nothing else about the platform matters.

1. Pick your highest-volume agreement type — usually NDAs or vendor forms — and map its current path end to end, including every person who touches it.
2. Test the Microsoft Word round-trip with a real counterparty redline, checking that comments, tracked changes and formatting survive intact.
3. Have someone from sales, not legal, start a contract from within your CRM. If that is awkward, adoption will fail regardless of the legal features.
4. Run Jurist against ten agreements you have already reviewed, and compare what it flagged with what your team flagged. Count both the misses and the false alarms.
5. Write the escalation policy before enabling agentic review: which agreement types, which deviations, who is accountable.
6. Ask for reference calls with two named customers of similar size and structure, which at this scale you should get.

Procuring Ironclad

There is no published pricing; the site directs to a pricing page and the route in is a sales conversation. That is standard for enterprise CLM.

In this category cost normally scales with contract volume, seat count, which modules you take and the length of the term. Implementation and configuration are usually a separate line, and they are frequently the larger one in year one.

The most important budget item is not on any invoice: your own people's time. A CLM deployment requires someone to define workflows, encode the playbook, migrate existing agreements and drive adoption across departments that did not ask for it. Projects fail when that role is nobody's job.

The security and compliance documentation lives at a dedicated security site referenced from the footer, alongside a status page and a support portal. Request the full pack early — for a system holding every contract your company has signed, that review is not a formality.

We do not publish amounts. Budget for implementation and internal ownership alongside the licence, since those determine whether the licence produces anything.

Ironclad: Stages, Integrations And Scale