The Stump, the Mirror, and the System

Why AI Without a Revenue Operating System Is Just Expensive Guessing

John Braze, Cory Bray

A Boy, a Stump, and a Question That Should Keep CEOs Up at Night

In February 2026, Michael Burry (the investor who shorted the 2008 housing market) published an essay that deserves attention from every executive writing checks for AI tooling. He did not talk about models or computing. He talked about a deaf child from 1880.

The story, originally presented at the Smithsonian Institution, concerned Melville Ballard, a deaf-mute teacher at the Columbia Institute in Washington, D.C. Before Ballard ever learned a single written word, he was already reasoning about the origin of the universe. At eight or nine years old, riding through the countryside with his father, he spotted a tree stump and silently asked himself whether the first man could have risen from it. Then he reasoned the idea away: a stump is the remnant of a tree, and trees grow from seeds, so that could not explain where humanity began. He moved on to the sun, the moon, the stars.

No language. No textbooks. No teacher. Just a child reasoning with pure thought.

Burry’s provocation was direct: today’s Large Language Models put language before the capacity for reason. They predict the next word with extraordinary fluency, but there is no Melville Ballard inside them. No silent reasoner behind the words. Just an increasingly sophisticated mirror.

“By putting language first, before the capacity for true reason, we are not building intelligence; we are building an increasingly sophisticated mirror.”

Michael Burry, Cassandra Unchained, February 2026

It is a striking metaphor. And for those of us who spend our days building revenue systems (pipeline architecture, stage gates, coaching sellers, debugging forecasts), it is also a devastatingly accurate description of what happens when companies deploy AI into a go-to-market motion that has no operating backbone.

The Mirror Problem in B2B Revenue

Burry was talking about AI research. But the pattern he described (fluency without understanding, language without reason) is precisely what we see every week inside scaling B2B companies.

Think about the last AI pitch you sat through. Personalized outbound at scale. Automated call summaries. AI-generated forecasts. Intelligent lead scoring. Every one of those capabilities sounds transformative. And every one of them assumes something that most companies have not yet built: a system underneath that defines what “good” looks like.

Without that system, AI does exactly what LLMs do. It produces fluent output from a broken foundation. The emails sound eloquent. The sequences seem sophisticated. The forecast looks confident. And the results are mediocre, because no one has defined who the right buyer is, what a qualified deal actually requires, how a rep should run a discovery call, or what happens when a deal stalls at proof of value.

Here is the thing that most AI vendors will never tell you: tools amplify whatever system you already run. If the system is sound, AI becomes a genuine force multiplier. If the system is broken, AI multiplies the chaos. Faster.

That is the mirror problem in B2B revenue. And it plays out on two sides of the house simultaneously: in RevOps (where the data and processes live) and in Sales Enablement (where the coaching and capability live). When either side is weak, AI has nothing solid to build on. When both sides are disconnected, AI amplifies fragmentation.

Reason First: What Burry’s Framework Means for Your Revenue System

Burry’s framework gives us a clean lens: reason must come before language. The capacity for structured thought must exist before language can unlock understanding from it.

Translated into revenue operations and sales enablement, this principle produces a simple rule. The operating system (your processes, your definitions, your coaching infrastructure) must exist before AI tools can produce value from the data they process.

Let’s walk through what this looks like on both sides.

The RevOps Side: Building the Capacity for Reason

RevOps is the operating backbone that keeps Marketing, Sales, Product, and Customer Success running on the same truth. When RevOps is established, you can clearly see the pipeline, trust the numbers, and adjust confidently. When it is built “for show”, you get pretty dashboards, stale records, and explanations that fall apart under scrutiny.

The RevOps backbone includes a clean data model with defined lifecycle stages. It includes enforced exit criteria at every stage gate, so deals cannot drift forward on gut feel. It includes SLAs that govern response time, routing, and handoffs. It includes hygiene rules that never bend: every deal must have a next step with a date, close dates must be updated weekly, and loss reasons must be recorded the moment a deal dies.

These are not administrative exercises. They are the reasoning capacity of your revenue system. Without them, your CRM is just a graveyard of good intentions. With them, every number tells a story, every pipeline review produces an action, and every forecast has evidence behind it.

Now consider what happens when AI sits on top of this backbone. Anomaly detection can flag stale next steps and stage creep before forecast meetings, allowing managers to coach sooner rather than react later. AI can monitor SLA compliance in real time and surface routing failures before leads go cold. Pipeline risk models can identify deals likely to slip based on actual stage behavior, not just close dates.

That is AI amplifying a system that works. The reason existed first. The language (the AI’s output) became useful because it had something real to describe.

The Enablement Side: Building the Language That Scales

If RevOps is the reasoning capacity of your revenue system, Sales Enablement is the language layer. It is how the system’s intelligence gets expressed in the field, in real conversations with real buyers.

And here is where most companies get it wrong: they treat enablement as training. A quarterly offsite. A content library that nobody opens. A new slide deck when the product updates. That is not enablement. That defies reality.

Real enablement is a system of continuous capability building. It starts before a rep’s first day and never stops. It looks like this:

A clear, teachable sales process with defined stages, exit criteria, and a Mutual Action Plan template that structures every deal. Standardized talk tracks and an objection bank that maps the most common resistance to specific proof-of-value assets. A library of recorded calls labeled by stage, so new reps can study what “good” sounds like at every point in the buyer’s journey. A coaching rhythm built on deliberate practice: annotating discovery notes, running structured role-plays with real buyer personas, and letting senior reps model live calls weekly.

The goal is elegant: make success teachable. If your best rep wins because they are naturally gifted, you have a personality, not a system. If your newest rep can follow the process and produce consistent results within 90 days, you have enablement.

Enablement always beats charisma. That sentence might be the most important idea in our book. It is certainly the most resisted. Every founder wants to believe that hiring the right person will solve the problem. But big resumes cannot compensate for broken foundations. Performance stabilizes only when onboarding starts on day one, not day thirty, and managers inspect behavior weekly, not quarterly.

Now consider what happens when AI sits on top of this enablement infrastructure. AI can analyze discovery calls and surface which questions a rep consistently misses, then recommend the specific coaching module that addresses the gap. It can match a stalled deal’s objection pattern to the highest-converting proof-of-value asset in your library. It can generate message variants for testing, then hand them to reps who edit for tone and context because they have been trained to know what good sounds like.

That is AI amplifying capability, not replacing it. The enablement system taught the rep how to think. AI helps them move faster within a framework they already understand.

The Interrelationship: Why RevOps and Enablement Must Be One System

Here is the insight that the market keeps missing.

Most companies treat RevOps and Sales Enablement as separate functions. RevOps owns the data, the CRM, the dashboards, and the workflow logic. Enablement owns the training, content, coaching, and playbooks. They often report to different leaders, use different tools, and operate on different cadences.

This separation is itself a form of the mirror problem. Each function produces fluent output in its own domain. RevOps generates clean dashboards. Enablement produces polished training decks. But without a shared operating backbone connecting them, the outputs never compound.

The dashboard shows that deals are stalling at the Proof stage. Enablement does not use that data to target coaching. The training covers objection handling for a new competitor. RevOps does not track whether objection frequency changes after the training deploys. The CRM captures loss reasons. Nobody feeds those reasons back into the enablement content to update the proof-of-value library.

Two systems, both producing good work in isolation, both failing to create the compounding effect that drives real revenue performance.

When you unify RevOps and Sales Enablement into a single revenue operating system, something fundamental changes:

RevOps provides the truth. Clean data, enforced stage definitions, SLAs, and pipeline visibility. This is the “reason” layer: the structured capacity that makes everything else possible.

Sales Enablement provides the capability. Playbooks, coaching, proof-of-value libraries, and deliberate practice rhythms. This is the “language” layer: the means by which the reasoning capacity gets expressed in the field.

AI multiplies both together. When RevOps and Enablement share a single system, AI has access to both the ground truth and the institutional knowledge it needs to operate effectively. It can flag a deal stalling at Proof and recommend the specific asset that advanced similar deals last quarter. It can detect that a rep’s discovery calls consistently miss a key question and surface the relevant coaching module. It can identify that stage creep correlates with a specific competitor and trigger the objection bank with competitive proof points.

This is Burry’s framework applied to revenue: reason (RevOps) combined with language (Enablement) produces understanding (consistent, scalable, AI-accelerated performance). Remove either side, and you are back to the mirror.

The Unified System: How Each Layer Enables AI

• • •

The Real Danger: Automating Your Way to Mediocrity

There is a pattern we see across companies as they scale from $3M to $50M in revenue. The CEO feels the urgency to grow. The board is asking for efficiency metrics. The market is saturated with AI vendors promising to “10x your pipeline.” The temptation is real: buy the tool, deploy it fast, and hope the technology papers over the cracks.

Let us be honest about what happens next.

The AI outbound tool generates thousands of personalized emails. Reply rates stay flat because the ICP was never validated, so the messages reach people who were never going to buy. The AI call summarizer produces beautiful notes after every meeting. Nobody reads them because there is no defined next step, no MAP, and no coaching rhythm that turns call insights into behavior change. The AI forecast model projects revenue with confidence. The numbers are wrong because stage definitions are not enforced, so deals sit in “Proof of Value” for months without anyone noticing.

None of these are AI failure. They are system failures that AI made visible faster. That is the irony: AI is often the best diagnostic tool a company can deploy, precisely because it exposes the gaps that everyone has been working around.

Burry’s Smithsonian story offers a test we can borrow and apply directly. The original formulation: an entity does not possess the capacity for understanding until reason is demonstrated in the absence of language.

In revenue terms, your go-to-market motion does not possess the capacity for AI-accelerated performance until it can produce consistent results without AI. If the process only works when a specific founder runs the call, or when a specific rep improvises the demo, or when a specific manager overrides the forecast, you do not have a system. You have a personality. And personalities do not scale.

What to Do About It

If you are a CEO or revenue leader at a B2B company in the $3M to $50M range, here is the sequence that works for you. It is not complicated. It does require discipline.

Step 1: Diagnose Before You Deploy

Before adding any AI tooling, answer these questions honestly. Can you explain why deals close and why they stall? Do your stage definitions have enforced exit criteria? Is your sales process documented well enough that a new rep can follow it without the founder in the room? Do Marketing, Sales, and Finance report the same numbers? Does your enablement system produce consistent ramp times, or does every new hire figure it out differently? If the answer to any of these is no, the system needs work before AI can help.

Step 2: Build the Backbone (Both Sides)

Install the RevOps infrastructure: a clean data model, lifecycle definitions with required fields, SLAs that prevent leaks, and hygiene rules that never bend. Simultaneously, build the enablement infrastructure: a documented sales process with exit criteria at every stage, an enablement pack that mirrors the customer journey, an objection bank tied to proof-of-value assets, and a coaching cadence that inspects behavior weekly. This can be done in 30 days. It does not require a massive investment in technology. It requires a commitment to treating RevOps and Enablement as a single integrated system, not two separate departments.

Step 3: Assign the Right Level of AI Autonomy

Start with Assist: let AI draft, summarize, and suggest next steps. Move to Suggest: let AI score, rank, and recommend with a human reviewing every output. Only after Assist and Suggest have proven reliable across two full sales cycles should you move to Decide, where AI takes automated actions. Most companies jump straight to Decide. That is language before reason. It does not work.

Step 4: Measure AI Like Any Other GTM Motion

Pick 100 records per use case. Score accuracy, relevance, tone, and actionability on a 1 to 5 scale. Keep only if the median score is at or above 4 for two consecutive weeks and a downstream metric improves. Add a red team pass: try to break it with edge cases and outdated data. AI that cannot survive scrutiny does not deserve autonomy.

Step 5: Run the Weekly Cadence

This is where everything compounds. Monday pipeline review surfaces the data. Midweek coaching addresses the gaps revealed by the data. Friday sync feeds learnings back into enablement content and RevOps definitions. AI insights that do not get reviewed, acted on, and fed back into the system are not insights. They are decorations. The cadence is where RevOps and Enablement converge, where AI output becomes organizational learning, and where the mirror finally becomes a window.

The Silence Before the Words

Melville Ballard did not need language to begin reasoning. He needed language to bring his reasoning to the world: to share it, to build on it, to teach with it.

Your revenue system is the same.

AI is not the reasoning. AI is the language layer that brings your system’s intelligence to the world at speed and scale. But the reasoning (the stage gates, the exit criteria, the SLAs, the coached behaviors, the objection bank, the proof-of-value library, the weekly cadence) must exist first.

RevOps provides the structure. Enablement provides the skill. Together, they create the operating system that AI needs to actually be intelligent. Apart, they produce two halves of a mirror, each reflecting activity without producing understanding.

Companies that build the system first will compound in value. Companies that skip it will spend millions on tools that show activity but don't produce results.

The question is not whether you should use AI in your go-to-market. You should. The question is whether your revenue system is ready to make AI intelligent, or whether you are just giving it more words to predict.

Build the system. Then let AI accelerate it. That is the sequence. That is the strategy. That is what we mean when we say: Go To Market Like You Mean It.

Ready to build the system before deploying the tools?

Start with a Revenue System Diagnostic: a structured assessment that shows exactly where your revenue operating system stands today and what to fix first. It is how we help CEOs stop guessing and start building.

mirmeridian.com/revenue-system-diagnostic

About the Authors

John Braze is CEO of Mir Meridian, a B2B go-to-market advisory firm, and co-author of Go to Market Like You Mean It: The Tactical Field Guide for Scaling SaaS Revenue. Mir Meridian helps B2B companies scaling from $3M to $50M build a systematic go-to-market infrastructure through RevenueOS™, a unified platform combining RevOps and Sales Enablement.

Cory Bray is Co-Founder of ClozeLoop and CoachCRM, and co-author of eight books on B2B sales performance, including The Sales Enablement Playbook, Triangle Selling, Sales Development, and Sales Playbooks. A Wharton graduate with a background spanning sales, finance, and operations, Cory has worked with over 500 companies worldwide to build repeatable sales processes, structured enablement systems, and coaching frameworks that drive measurable revenue growth.

References

Michael Burry, “History Rhymes: Large Language Models Off to a Bad Start?” Cassandra Unchained, February 28, 2026.