TL;DR
Usually not the reading. Waiting, deciding twice, chasing the counterparty and re-reading what was agreed consume it, and only one is fixed by reading faster.
Usually not because the reading is slow. A commercial agreement takes a competent reviewer under an hour to read properly. If your cycle time is eleven days, the reading is not the eleven days.
Four things consume it — waiting, deciding twice, chasing the counterparty, and re-reading what was already agreed. Only one of them is fixed by reviewing faster.
So what is actually taking the time?
Split a cycle into elapsed time and working time and the gap is the answer.
Working time is someone reading, marking and deciding. Elapsed time is everything between: the contract sitting in an inbox, waiting for a second opinion, waiting for the counterparty, waiting for a signature.
In most teams the working time is hours and the elapsed time is days. A system that halves the reading changes hours. It does not touch the days.
That is the whole diagnosis, and it is why a contract review purchase so often produces a faster review and an unchanged cycle time — which is a very expensive way to be disappointed.
What are the four causes?
Named, because each has a different fix:
| # | Cause | What it looks like |
|---|---|---|
| 1 | Queue | It sat for four days before anyone opened it. The reviewer was not slow; they were busy |
| 2 | Deciding twice | No agreed position, so every deviation becomes a fresh conversation with someone senior |
| 3 | Counterparty round-trips | Three exchanges where one would do, because each pass raises issues that could have been raised together |
| 4 | Re-reading | Nobody can say what was agreed last time with this supplier, so the whole thing is read from scratch |
Cause 2 is the one that masquerades as complexity. It feels like the contract is hard. It is usually that nobody has decided the company's position, so it gets decided again, per contract, by whoever is available — which is also why two reviewers disagree.
Cause 4 is the one that compounds. Every review that ends without leaving a record guarantees the next one starts cold.
How do you find out which one is yours?
Time the stages separately for one month. Not the total — the stages:
- Received → opened. This is queue.
- Opened → first markup returned. This is working time. It is usually the smallest number.
- Markup → internal agreement. This is deciding twice.
- Each counterparty round-trip. Count them.
The largest number is your problem. Most teams are surprised, and the surprise is the point — the stage everyone complains about is rarely the stage with the days in it.
This costs a spreadsheet and a month. No software, no vendor, no project. It is the cheapest diagnostic in this whole subject and almost nobody runs it before buying.
Which of the four does AI actually fix?
Directly, one. Indirectly, two. Not the fourth without a different build.
- Working time — yes, substantially. Reviewing against a playbook turns an hour of reading into minutes of checking, and redlining turns findings into proposed edits.
- Deciding twice — indirectly, and this is the bigger win. A system forces the positions to be written down. The act of building the playbook is what removes the second decision, not the software that reads it.
- Round-trips — indirectly. Raising every issue in one consistent pass beats discovering them across three.
- Queue — no. If a contract waits four days for attention, a faster review produces the same wait followed by a faster review. Nothing about AI addresses capacity or priority.
- Re-reading — only if the output is retained as structured obligations rather than a marked-up file.
So the honest claim is narrower than the market's, and it is still worth having: AI compresses the part that was already the smallest, and forces the decision that was actually costing you the days.
Does adding reviewers fix it?
Only if the largest number is queue — and even then, less than you would expect.
Adding a reviewer raises capacity, which shortens the wait. It also adds a source of inconsistency, because the new person will apply the positions differently unless the positions are written down. You can shorten cause 1 and worsen cause 2 in the same hire.
There is a second-order effect worth knowing before you make the case for headcount. More reviewers means more variation in what gets accepted, which means more counterparties learn that outcomes depend on who picked the contract up. That feeds cause 3 — round-trips increase when the other side thinks a different answer is available.
The sequence that works is the reverse of the instinct: write the positions first, then add capacity against a standard. A second reviewer applying an agreed playbook adds throughput. A second reviewer applying their own judgement adds throughput and variance.
What should you do first?
In this order, and only the third involves buying anything:
- Measure the stages for a month. You cannot fix the largest number until you know which one it is.
- Write the positions down. Cause 2 is the most common and the fix is a checklist of positions, not a system.
- Then consider a system — and size it against the stage that is actually costing you, not the one that is most annoying.
If steps 1 and 2 fix it, that is a good outcome and it cost you nothing. We would rather tell you that than sell you the third step against the wrong problem.
About Cognilium Cognilium builds AI optimization apps for Microsoft Dynamics 365 — companion apps that optimize the pricing, inventory, warehouse and planning decisions your ERP manages but can't optimize. Dynamics is your system of record. Cognilium is your system of intelligence. https://cognilium.ai · https://www.linkedin.com/company/37180269/
Legal AI Ops. We transform legal workflows with agentic AI, copilots, agentic workflows and decision intelligence — built into core workflows rather than beside them, to raise productivity and cut operational overhead. Contract Review Copilot is the contract-review app in that family. It ships as Paralegent AI, in production today. How we build Legal AI Ops — custom AI capabilities on top of legal work, against your playbook and your Dynamics 365.
Bring one month of stage timings to a 15-minute call and we will tell you which of the four you have — including when the answer is that you do not need us.
Sources
No external source is cited. Every claim is our own diagnostic position, labelled as such.
Sources and fact-check
| # | § | Claim | Tier | Source | Verdict |
|---|---|---|---|---|---|
| 1 | 1 | A commercial agreement takes under an hour to read properly | T2 — ours, a working estimate stated as ours. Not a benchmark, not measured, not attached to a customer | Internal definition | PASS |
| 2 | 1 | Working time is hours, elapsed time is days | T2 — ours, deliberately unquantified beyond the order of magnitude | Internal definition | PASS |
| 3 | 2 | The four causes | T2 — ours | Internal definition | PASS |
| 4 | 2 | "Deciding twice" usually means no position was agreed | T2 — ours, the article's sharpest claim, stated as ours | Internal definition | PASS |
| 5 | 3 | The four stages to time | T2 — ours, a method | Internal definition | PASS |
| 6 | 4 | What AI fixes and does not | T2 — capability + ours. Explicitly narrower than the market's claim, and says AI does nothing for a queue | Profile §1 | PASS |
| 6b | 4a | Adding reviewers can shorten queue and worsen consistency; variance feeds round-trips | T2 — ours, an argument stated as ours | Internal definition | PASS |
| 7 | 5 | The three-step order, two of which involve no purchase | T2 — ours | Internal definition | PASS |
Tier summary: 0 × T1, 8 × T2 — 0 × T4.
Deliberately unquantified throughout. "Eleven days" in the opening is an illustrative example, not a claim about anyone; "under an hour" is labelled as our estimate in row 1. There is no measured cycle time, no percentage, no customer, and no before/after — which is exactly what an article on this subject is usually built from.
Disclosure: the CTA explicitly includes "when the answer is that you do not need us."
