The under-discussed CLM purchase trigger isn't ROI. It's the institutional knowledge that walks out the door when your most tenured contract person leaves.

Every company has a Diane.

You've met her. She is the senior paralegal who is the firm's institutional memory for contracts. The contract administrator at the construction company who set up the filing system everyone now uses. The longest-tenured person on the legal-ops team, the one whose name shows up on every old document, who knows where every signed PDF lives because she filed every one of them herself.

She is the CLM the company does not know it is running.

The conversations we have usually open with something presentable: we are growing, our process is too manual, our renewals are slipping, our sales team needs faster turnaround. Then 10 minutes in, the buyer says it almost in passing.

"My contract administrator has been in the business for a long time. She's the one who has all the knowledge on how to do these contracts."

That is the actual reason. Everything else is the rationalization. Someone who has carried a meaningful share of the company's contract knowledge in their head is leaving, and the system that depended on them is about to find out exactly how much it depended on them.

What Diane knows that nobody else does

Diane is a judgment system. Filing is the 10 percent of her job that gets documented. The other 90 percent, the part that makes her irreplaceable, is everything she has learned about how this company actually contracts, which is something that no template, no playbook, and no onboarding deck has ever fully captured.

A sample, drawn from what we hear in real conversations:

• She knows that the master services agreement with the largest reseller has a price-cap rider that was negotiated years ago and still lives in someone's archived Outlook folder. The clock on it runs on the international subsidiary's fiscal calendar, not the parent company's. Two of the three people who knew about it have already left.

• She knows that the standard NDA template is, in practice, four templates. Sales uses one. The implementation team uses another. The original, the one legal actually approves, gets pulled out only when counsel is in the room.

• She knows that one specific counterparty always negotiates the indemnification clause. Always. Concede gracefully on round one and you keep the deal closing on schedule. Push back and you lose two weeks.

• She knows the filing convention. Not the official one. The one she actually uses, which evolved over years and works because she enforces it. There is a system somewhere, and there is the way Diane does it, and the way Diane does it is the system that works.

• She knows, when an executive walks up and says "do we have anything signed with this vendor before," exactly which folder to open and which year to scroll back to. The search took her 12 seconds. It would take a successor 12 hours.

None of this is in your CLM. None of this is in your shared drive. None of this is in the policy manual you wrote two years ago and nobody reads.

It is in Diane.

What breaks when Diane leaves

Sudden departure. She leaves with two weeks' notice. The team divides her work into "filing" and "questions about old contracts," and discovers within a quarter that the second category is the entire job. The records-custodian role gets handed, by default, to whoever happens to be senior in legal that week. "Well, she's gone, so now I am considered the records custodian".

Slow handoff. Diane gives a longer period notice. The company agrees to "train her replacement." What this looks like in practice is a junior paralegal shadowing Diane for a few weeks of contract execution, picking up a small fraction of what is actually in her head, and leaving with a three-page document that captures none of the unwritten rules. Diane finishes. The replacement is competent. The replacement is not Diane. Things start to slip. Small things first, then the kind of thing that costs the company a multi-year contracted protection no one else remembered existed.

The platform purchase. This is the version that brings vendors into the conversation. The CFO, or the CEO, or the new GC takes a hard look at the situation and says some version of "we can't lose this much tribal knowledge" This is the moment the CLM evaluation actually starts.

Why this is happening more right now

Two things are colliding.

First, the person who carried the knowledge is gone, and the company finds out, usually within the next quarter, exactly how much of the workflow lived in one person's head.

The second is structural. Companies that grew through the last decade on email and SharePoint and a single dedicated person are running into the limits of that approach. "There was no home where all of our contracts for our partners actually live".

Diane's departure is the forcing function. Everything else (the platform evaluation, the workflow redesign, the AI conversation) is downstream of it.

What the evaluator does differently

If you are in a CLM evaluation right now and you can see a Diane situation in your future, here is the thing nobody will tell you in a vendor demo: the cost of the platform is not the variable to negotiate. The capture window is.

One: buy on the leading edge of the departure, not the trailing one. Most CLM purchases happen in the wrong order. The notice arrives. The team scrambles. The platform gets bought reactively, after the institutional knowledge has already started leaking.

Two: run a structured 90-day knowledge extraction. Not "shadowing." Not "training." A defined protocol with specific artifacts, a specific cadence, and a specific owner.

What this could look like:

• Daily, 20 to 30 minutes. Diane works a real contract (not a synthetic one) with the successor and an ops sponsor in the room. After the work is done, Diane narrates the why of every non-obvious decision she made into a running decision log. Not the action; the reasoning. Why this clause and not that one. Why concede here and push there. Why this counterparty gets the extended-payment term and that one does not.

• Weekly, 60 minutes. The successor and the ops sponsor review the week's decision log together. Anything Diane did that the successor cannot yet defend on their own becomes a flagged item for the next week's coverage. Anything that recurs three times becomes a candidate for an explicit rule the platform should encode: a workflow rule, a clause-library entry, a counterparty playbook note.

• Four standing artifacts, all updated weekly. A decision log (what she did and why). A counterparty playbook (per major counterparty, the patterns and concessions). A template inventory (every variant of every template that is actually used in practice, with the cases each one applies to). A filing-convention map (the rules Diane uses, written down honestly, including the workarounds).

• Two checkpoints. Day 45: the successor runs a contract end-to-end while Diane observes silently and only intervenes for genuine errors. Day 80: the successor runs a week of contracts solo. Diane is on call but does not initiate. The artifacts are the bridge.

The successor is in the room throughout, but replicating Diane's actions is not the goal. The goal is to extract the rules behind the actions so the platform (and the next person, and the person after that) can apply them.

Three: encode the unwritten rules, even when they look stupid in writing. This is the step every company tries to skip and every successor regrets. The "we always concede on indemnification with this counterparty" rule looks weak when you write it down. Write it down anyway. The "the standard NDA is actually four templates" rule looks embarrassing. Document it anyway.

A short closing

Diane is going to leave eventually. Maybe in two months, maybe in five years. You can wait until she's gone to find out what she knew, or you can use the runway you have to get her to teach the building. The platform is the container.

Buy the platform that gives Diane a place to put what she knows.

That is the platform that will still be working when Diane is not.