Infrastructure buyers have heard “AIOps” before: selling with evidence, not claims
The buyer sponsored the last platform that promised correlation and delivered dashboards. Their scepticism is earned. Here is how to meet it: four evidence levels, a claims register, the P1 walkthrough, and the pilot that turns a target into a result.
In Verification Is the New Bottleneck in Enterprise AI I argued that the organizations allowed to do more with AI will be the ones that can prove what their AI did. This piece applies that argument to selling, in the market where it bites hardest. Infrastructure buyers have heard “AIOps” before. They sponsored it. It correlated some alerts, produced a dashboard, and the senior engineer kept rebuilding the picture by hand. Their scepticism is not a mood. It is earned, and the seller who treats it as an objection to be handled will lose to the seller who treats it as the standard to be met.
The standard is evidence. Not more evidence than the buyer asks for — honest evidence, labelled by what it is, offered in a form the buyer can check on their own network. That is how I run positioning for the companies I advise in this market, and it is the fourth component of The Seven C’s Outreach Program™: evidence over adjectives.
- The buyer has sponsored a platform that promised correlation and delivered dashboards. Their default response to a new claim is “show me on my network,” and a young company usually cannot — not yet. The way through is not a bigger claim. It is a labelled one.
- Every proof point has one of four evidence levels: measured on a customer, measured in a pilot, a design target, or a pattern from discovery conversations. The label is always visible. A target quoted as a result is the fastest way to lose a technical buyer for good.
- A claims register, a P1 walkthrough and a paid proof of value are how a company earns the first two levels. Until then, the honest design target wins more deals than the unlabelled miracle, because it is the first thing the buyer can verify.
1. Why the buyer’s scepticism is the right starting point
The previous generation of tools made three promises: correlation, noise reduction, and a single view. The buyer got the single view. What they did not get was the thing they actually needed — the explanation, the proof that a change worked, the evidence that failover would. The gap between the promise and the delivery is now the buyer’s prior, and every new vendor inherits it.
Three consequences for the seller:
- 01Adjectives are read as warnings“AI-powered,” “autonomous,” “self-healing” signal that the vendor has not been in the buyer’s seat. The engineer who reads them reaches for the delete key, and the architect who reads them reaches for the hardest question they know.
- 02The embarrassing question is the buying question“Show me on my network” is not hostility. It is the buyer telling you exactly what would make them buy. A seller who cannot answer it has a product problem or a proof problem, and should know which.
- 03Honesty is a differentiatorIn a category with a reputation for over-claiming, the vendor that labels a target as a target stands out immediately. Buyers remember the one vendor who said “we have not measured that yet.”
2. The proof hierarchy
Every proof point a company uses sits at one of four levels, in a fixed order of strength. The label travels with the claim — in the email, the deck, the reply and the pilot readout.
Named with permission, or described without identification
“A regional health system measured a median time from alert to explanation of [X] minutes across [N] live incidents over ninety days, scored by their own engineers.” The strongest claim there is, and the rarest for a young company.
Anonymized, with the measure and the window
“In a forty-five-day pilot last quarter the product explained [N] of [M] past incidents without being told the cause, as judged against the customer’s post-incident reviews.” Honest, specific, and available to any company that runs real proofs of value.
Stated as a target, never as a result
“Our design target is root cause in under five minutes, measured against a reference architecture.” Perfectly respectable — as long as the word “target” is in the sentence and the reference architecture is named.
“In more than one discovery conversation we heard…”
The weakest level, and still useful: it tells the buyer the vendor has been listening to people like them. It never becomes a number.
A young company has mostly levels three and four. That is fine. The buyer trusts a labelled target more than an unlabelled miracle, because the labelled target is the first thing they can verify — the claim is specific enough to be false, and the vendor has shown they know the difference.
3. Where claims drift, and the register that stops it
Claims do not usually start dishonest. They drift. A design target on the website becomes “customers see” in a deck, becomes “we deliver” in a seller’s reply to an objection, and by the time the architect asks for the source, nobody can find one. The control is boring and effective: a claims register.
| Column | What goes in it |
|---|---|
| Claim | The sentence as it may be used, verbatim |
| Evidence level | 1 measured on a customer; 2 measured in a pilot; 3 design target; 4 pattern |
| Source | The pilot readout, the customer’s written permission, the reference-architecture test, the discovery notes |
| Permission | What may be said, to whom, and whether the customer may be named |
| Expiry | When the claim must be re-verified or retired |
| Approved by | The named person accountable for it |
The rule that gives the register teeth: no message, deck, reply, website page or AI-drafted sentence uses a claim that is not in it. The register is reviewed monthly; expired claims come out; new pilot measures go in. It takes an hour a month and it is the difference between a company that can survive a technical buyer’s due diligence and one that cannot.
4. Evidence in the message
What this looks like in a first email is one labelled sentence where the adjectives used to be. Compare two versions of the same proof line, for a product that verifies change windows:
“Our customers eliminate failed changes.”
A claim with no level, no source and no number. The architect reads it as a warning; the leader reads it as marketing; nobody can check it.
“In a pilot last quarter, one real change window was compared before and after and the variance against intent was recorded in the ticket — measured by the customer’s change manager.”
A level-two claim with a method and an owner. It invites the buyer to run the same measure on their own window.
The second version is longer and it converts better, because it does the buyer’s due diligence for them in advance. It also sets up the only ask that matters in this market, which is not “a demo” but “bring us your last P1.”
5. The P1 walkthrough: evidence before the pilot
Between discovery and a pilot sits the most effective conversion step I know for this market. The buyer brings their last major incident or their last failed change. They tell the story — first alert, who did what, how long until the explanation, what it cost. Then the product works the same class of problem on a comparable environment, in thirty minutes, with the commands and outputs behind every conclusion visible on the screen. It is not a demo. It is the buyer’s own story retold by the product, and the Critic gets to ask the access question in a setting where nothing is at stake.
Two rules make it work. Show the evidence, not the interface: the buyer should leave able to describe the evidence standard, not the screens. And say what it would not have caught. The admission is what earns the pilot, because it proves the vendor knows where its own edges are. The replay on the buyer’s own network — their past incidents, the product not told the cause — happens inside the proof of value, not before it, and it is where a level-three claim becomes a level-two one.
6. Earning the first two levels
The proof hierarchy is not a hierarchy a company climbs by writing better copy. It climbs by running proofs of value with written success criteria, measured by the buyer’s own engineers, and by asking the customers who saw the measures for permission to describe them. Each pilot readout is a level-two claim waiting to be registered. Each customer who signs after one is a level-one claim waiting for permission. The company that runs six honest pilots in a quarter ends the quarter with a register full of things it can say, and the company that runs six free trials ends it with a dashboard.
That is the argument of Proof of Value That Converts, and it is why the program’s closing instrument is a paid, time-boxed proof with criteria the buyer writes, the Economic Buyer in the readout, and the misses reported first.
7. The posture that wins the technical buyer
- Lead with the event, not the category. The first line names something the buyer lived through.
- Label every claim. Measured, pilot-measured, design target or pattern — visibly.
- Keep a claims register and let nothing outside it reach a buyer.
- Offer the P1 walkthrough before anyone says the word “demo.”
- Say what it would not have caught. The admission earns the pilot.
- Run paid proofs of value with the buyer’s criteria, and register the results.
- Never disparage the incumbent. “Keep what you run; we reason over it” is the only comparison the program makes.
None of this is a workaround for a weak product. It is the discipline that lets a strong product be believed by the one audience that has learned not to believe anything until it runs on their network.
The Seven C’s Outreach Program™
Evidence over adjectives is one of the program’s three conventions. The proof hierarchy, the claims register, the P1 walkthrough and the proof-of-value kit come with it.
See the program »Verification Is the New Bottleneck in Enterprise AI
The organizations that can prove what their AI did are the ones allowed to do more of it. The essay this piece applies to selling.
Read the essay »Questions this piece answers
How do you sell AIOps or network automation to buyers who have been burned before?
Lead with the event they lived through, not the category; label every claim by its evidence level; offer to work their own last incident in thirty minutes with the evidence visible; and propose a paid, time-boxed proof of value with success criteria they write. The buyer trusts the vendor that volunteers its misses.
What is the difference between a design target and a measured result?
A design target is what the product was built to achieve, measured against a reference architecture under the vendor’s control. A measured result is what the product did on a customer’s own network, scored by the customer’s engineers over a defined window. The first is a promise; the second is evidence. Both can be stated — but never the first in the language of the second.
What is a claims register?
A maintained list of every proof point a company uses in selling, with its evidence level, its source, the permission attached to it and its expiry. No message, deck or reply uses a claim that is not in the register. It is the single control that stops a design target from drifting into a result somewhere between the website and a sales call.
Part of the outreach series for companies selling into AIOps and network operations: Signal-Based Outbound for Network Operations, Selling to the NOC, Close the Loop, Don’t Break Up, this piece, Proof of Value That Converts and Operator Edition No. 05: The Outbound Agent Mandate.
Value hypothesis and positioning
One of the five stages of the Revenue Foundation. Every piece for this stage, in order, is on the Field Guide.
Open »The Seven C’s Outreach Program™
The proof hierarchy and the claims register are part of one motion — signals, the committee, the message spine, the deal-force check and the proof of value — built with you.
See the program »New essays and operator tools, as they publish.
A few times a month, by email. No list is sold or shared; one click to leave.