contract intelligence
Contract Intelligence Software: What the Label Covers and How to Test It
Written by Nisha Varenkar Contract intelligence software turns a pile of signed agreements into data a business can query: who the parties are, what each contract commits you to, which terms carry risk, and what happens next and when. That is the short answer. The longer answer matters more to a buyer, because "contract intelligence" is a marketing label rather than a standard, and two products sold under it can do quite different things. This guide separates the capabilities that vendors bundle under the label, shows which vendors use the term and what they mean by it, gives a protocol for testing a contract intelligence platform on your own documents, and is frank about when the whole category is more than a team needs. ## Five capabilities hiding behind one label Strip away the branding and most products in this space offer some mix of five capabilities. Very few are equally strong at all five. | Capability | What it produces | The question that exposes weak products | |---|---|---| | Extraction | Structured fields: parties, dates, values, governing law, notice periods | How do I correct a wrong field, and does the correction teach anything? | | Risk scoring | A rating or flag per clause or per contract | Risk compared with what: a market standard, or my own playbook? | | Obligation surfacing | A list of things someone must do, by when | Does it separate my obligations from the counterparty's? | | Portfolio analytics | Counts, filters and charts across hundreds or thousands of contracts | Can I build a new question myself, or only use preset dashboards? | | Search | Keyword plus meaning-based search, sometimes question answering | Does every answer link back to the clause it came from? | Extraction is the foundation; the other four sit on top of it. If extraction is shaky, risk scores, obligation lists and dashboards inherit the errors and present them with more confidence than they deserve. That is why accuracy testing, covered below, comes before any dashboard demo. A closer look at how extraction errors are measured is in our piece on contract data extraction accuracy. ## One supply agreement, read five ways Take a single three-year supply agreement with this renewal clause: "This Agreement shall renew automatically for successive one-year periods unless either party gives written notice of non-renewal not less than 90 days before the end of the then-current term." Here is what each capability should do with it. - Extraction records the initial term end (say 30 June), renewal type (automatic, one year), and notice period (90 days, written).
- Risk scoring compares the clause with your playbook. If your policy is that supply contracts must not auto-renew without a price review, the clause is flagged, and the flag should say why.
- Obligation surfacing creates a dated task: decide on renewal and, if leaving, serve written notice no later than 1 April, with an owner named.
- Portfolio analytics counts this contract among all auto-renewing supply agreements with notice windows closing in the next quarter, and totals their annual value.
- Search lets a manager type "which supplier contracts renew on their own?" and see this clause, highlighted, among the results. Notice how each step depends on the one before it. If extraction reads the 90 days as 60, the task date is wrong, the quarter's renewal list is wrong, and the manager acts on a false deadline. The more downstream features a product offers, the more an early error is multiplied, which is the practical argument for testing extraction first. ## Vendors that sell under the contract intelligence name The phrase is used by several large vendors, each with a different centre of gravity. The examples below are drawn from the vendors' own descriptions; they are not a ranking. Icertis brands its product as an AI-native contract intelligence platform and presents it as full contract lifecycle management with AI layered across drafting, obligation tracking, analytics and search. It is aimed at large enterprises, often with procurement and sales contract volumes in the tens of thousands. Evisort was described as an AI-native contract intelligence platform in Workday's 2024 announcement of its agreement to acquire Evisort, which set out plans to bring its document intelligence into Workday's finance and HR products. Buyers already on Workday should ask what is now delivered inside Workday and what remains a separate product. Kira, now part of Litera, is positioned by Litera as purpose-built for high-volume, high-stakes review and due diligence, with pre-built smart fields plus fields a team can train itself. Its natural home is law firm deal review rather than an operating company's renewal calendar. Many CLM vendors also describe their AI features as contract intelligence without making it the product name. So when the label appears in an RFP response, ask which of the five capabilities it refers to. ## Contract intelligence, CLM and plain search are not the same purchase A CLM platform runs the process: intake, drafting, approvals, signature, storage, renewal. Contract intelligence analyses content. Some products do both; some are strong at one and thin at the other. Plain enterprise search finds words. Ask it for every contract with a limitation of liability below one year of fees and it will return every contract containing the word "liability". A contract intelligence tool should understand that the cap is a number, extract it, and compare it. A practical way to decide: if your biggest problem is that contracts take too long to get signed, start with process (CLM or contract automation). If your biggest problem is that nobody knows what the signed contracts say, start with intelligence. ## A 30-contract protocol for testing a contract intelligence platform Vendor accuracy claims are measured on the vendor's own sample. Your contracts are messier. Run this before you buy. Step 1: choose the set. Pick 30 signed contracts that reflect reality: ten on your paper, ten on counterparty paper, five scanned PDFs, five with amendments or side letters. Include at least three in a second language or with non-standard formatting if your portfolio has them. Step 2: build the answer key. Have a lawyer fill a sheet with the correct value for twelve fields per contract (party names, effective date, term, renewal mechanism, notice period, governing law, liability cap, payment term, assignment, termination for convenience, data protection clause present, one obligation of your choice). This takes time, and it is the only honest baseline you will have. Step 3: run the tool cold. Load the 30 contracts without any vendor tuning. Record every field the tool got right, wrong, or missed. Step 4: check risk scoring for consistency. Run five contracts twice, a day apart. A risk flag that changes between runs on an unchanged document tells you how much to trust it. Step 5: ask real portfolio questions. Three examples: which contracts renew in the next 90 days with notice still possible; which customer contracts restrict assignment on a change of control; which suppliers may raise prices next quarter. Compare the answers with your answer key. Step 6: score. | Measure | How to record it | |---|---| | Field errors | Wrong plus missed fields out of 360 (30 contracts x 12 fields) | | Correction effort | Minutes to fix the errors in the tool | | Scan handling | Errors on the five scanned contracts versus the rest | | Consistency | Number of risk flags that changed between two runs | | Traceability | Share of answers linked to the source clause (count them) | | Question answering | Correct results on your three portfolio questions | This test also tells you whether a reviewer still has to read every contract, which is the real cost question. For a wider view of where AI review goes wrong, see is AI contract review accurate. ## Questions portfolio analytics should answer on day one Dashboards look impressive in a demo and go unused in month three unless they answer questions someone already asks. Write your questions first, then check each tool against them. | Business question | Fields the tool must have extracted | Who asks it | |---|---|---| | What do we lose if this supplier fails? | Liability cap, exclusions, termination rights, step-in rights | Procurement, risk | | Which customer contracts block a sale of the business? | Assignment, change of control, consent requirements | Legal, finance in M&A | | Where are we exposed to a new data law? | Data protection clause, processor role, transfer terms | Privacy, legal | | What revenue renews next quarter? | Term end, renewal type, notice period, contract value | Sales, finance | | Which contracts let the other side raise prices? | Escalation clause, index, cap, notice of increase | Finance | If a vendor cannot answer three of your five questions on your test set, the dashboards will not save it. ## Where the category is overkill Contract intelligence earns its cost when the portfolio is too large or too varied to read. It is often unnecessary when: - Your contracts are mostly your own templates with few negotiated changes. The data is already known; a good template library and a register are enough.
- The whole portfolio could be read and summarised by one paralegal in a few weeks. A one-off review project may be cheaper than an annual licence.
- Nobody owns the output. Extracted data with no renewal owner, no risk owner and no process to act on flags is an expensive spreadsheet.
- Your real bottleneck is negotiation speed on new deals. Analysis of signed contracts will not shorten a redline cycle. ## Costs that sit outside the licence Budget for the work around the software, not only the subscription: - Collection. Finding the signed versions, with amendments, across email, drives and filing cabinets.
- Scanning and OCR. Old paper contracts need clean scans; poor scans produce poor fields.
- Taxonomy. Agreeing on field definitions, such as what counts as a renewal date, before extraction.
- Verification. Someone has to check fields that drive money or deadlines, at least at first.
- Upkeep. New contracts must flow in by default, or the data decays within a year. ## Adira and contract intelligence: what it covers today Adira's Document Intelligence Dashboard is contract intelligence at the level of a single document. Open a contract and it shows each position as favourable, neutral or unfavourable to your company, red flags, vulnerabilities with severity, key dates, obligations, monetary terms, missing clauses and Indian compliance items, with a single risk-posture meter. Because Adira models the contract as a structured clause tree, you can act on a finding directly: Clause-Level Editing applies a plain-English instruction across the document and shows the diff. Its Company Persona tunes the analysis to your own preferred terms rather than a generic market view. The limit is portfolio scale. Adira's single pane across all contracts, the Company Command Centre, is on the roadmap rather than live, so a team whose main need today is analytics across many thousands of legacy contracts should consider other tools built for that first. Adira's own measured result across the 18 law firms using it is that contract review and due diligence fell from 10 to 12 days to under 2 hours with only 2 to 3 rechecks needed; that is Adira's figure, and the 30-contract protocol above is the way to check it on your documents during the 7-day trial at adiralaw.com. For how flagged obligations feed day-to-day work, see what is obligation management.
Frequently asked questions
- What is contract intelligence?
- It is software that reads signed contracts and turns them into structured, searchable data: parties, dates, values, obligations and risk flags, plus analytics and search across the portfolio. It is a marketing label rather than a formal standard, so check which capabilities a vendor means.
- Is contract intelligence the same as CLM?
- No. CLM runs the contract process from request to renewal. Contract intelligence analyses the content of contracts. Some products combine both, but many are much stronger at one than the other.
- How many contracts do we need before contract intelligence is worth it?
- There is no fixed number. It pays off when the portfolio is too large or varied to read in a reasonable time and someone will act on the output. If one paralegal could review everything in a few weeks, a one-off project may be cheaper.
- How do we test extraction accuracy fairly?
- Build an answer key from your own signed contracts, run the tool without vendor tuning, and count wrong and missed fields per contract. Include scans, amendments and counterparty paper so the test reflects your real portfolio.
- Does Adira offer portfolio-wide analytics?
- Adira analyses each document in depth in its Document Intelligence Dashboard. Its portfolio-wide view, the Company Command Centre, is on the roadmap, so teams that need large-scale analytics across legacy contracts today should evaluate tools built for that.
Sources
See how Adira drafts in your voice and reads contracts from your side.
Explore the showroomWorking through a contract like this? Weave is Adira’s free tool to read, mark up, and connect any contract in your browser — no account needed.
Try Weave — free