This is The Agentic Commercial Model, Season 2: the build. Issue 02 set out the loop. Issue 03 built the instrument. This is what happens when every counterparty in that loop is an algorithm.
Picture your next renewal, twelve months out.
Your customer's finance function has an agent watching spend. It reads the invoices, pulls the consumption data, and assembles the question every procurement process now starts with. What are we paying for this, and what did we get.
The answer to the first half is precise to the penny. The answer to the second half is whatever evidence you have made available. If you have supplied nothing machine-readable, their agent answers using spend alone, and spend without proof always looks like waste.
Nobody in that chain is hostile. No human has yet formed an opinion. The case against you was assembled automatically, from data you provided, in a format you never thought to shape.
That is the machine most companies are not even aware of, and it is one of four.
Four machines, one loop
The mistake in most writing on this subject is treating it as a marketing problem. Get cited in ChatGPT, win the shortlist, job done. That is one machine, at one stage, and winning it while losing the other three is a short career.
There is a machine at every stage of the loop.
At land, the discovery machine builds the shortlist before you know the buyer exists. At realise, the customer's own audit machine decides continuously whether you are worth what they pay. At expand, the orchestration machine routes work between capabilities and either finds you or does not. At advocate, the citation machine turns your proven outcomes into public evidence, which feeds straight back into discovery.
Four machines. One loop. And, as it turns out, one substrate underneath all of them.
Land: the discovery machine
Season 1 covered this ground, so I will be brief and stick to what has changed.
The numbers have got worse for anyone hoping their SEO investment transfers. An analysis of 250 million AI search results found traditional SEO metrics explain between four and seven percent of citation variance. Fifteen years of discipline, explaining almost nothing about the moment that now decides consideration.
Worse, presence decays. Between forty and sixty percent of cited domains change every month, with Google AI Overviews at 59.3% drift and ChatGPT at 54.1%. This is not a project you complete. It is a monthly cadence you run forever, which is why it belongs inside the loop rather than beside it.
The build is a query bank. Thirty to fifty questions your buyers actually ask a model, drawn from sales conversations rather than a keyword tool, run monthly, tracked per engine and never pooled into an average. The board metric is share of answer, the percentage of that bank where you appear. It belongs next to pipeline, because it now determines pipeline.
Realise: the customer's audit machine
This is the machine nobody is preparing for, and it is the one that will quietly kill more renewals than anything in the discovery layer.
Your customer is instrumenting themselves. FinOps tooling, procurement agents, internal copilots with access to spend and usage data. The question of whether a vendor is worth it used to be asked once a year by a human who had a relationship with you. It is now asked continuously by a system that does not.
That system reads consumption because consumption is what you gave it. Unless you publish outcomes in a form it can consume, your value case is invisible to the entity now forming the value judgement.
So the instrument from Issue 03 is not only your internal measurement. It is an output. Verified outcomes, delivered to the customer as structured, machine-readable evidence, on the same cadence their agent is watching. Not a slide in a quarterly review that no algorithm will ever read.
Companies that get this right will win renewals before anyone schedules a conversation. Companies that do not will discover their value case was argued in their absence, by a machine, using only the half of the ledger that made them look expensive.
Expand: the orchestration machine
As agents route work between capabilities, discoverability stops being a marketing property and becomes a technical one.
If the customer's orchestration layer cannot find your product, read what it does, or invoke it, you are not in the expansion path. It does not matter how good the product is or how happy the human champion is. The work gets routed to whatever the machine can see and call.
This is a genuine strategic exposure for companies whose product is excellent and whose interfaces assume a human. Expansion in an agentic estate flows through the orchestrator, and the orchestrator reads capability, not brochures.
Advocate: the citation machine
The loop closes here, and this is where the four machines turn out to be one system.
Every verified outcome from the realise stage is a proof asset. Published, specific, attributable, machine-readable, it becomes exactly the material the discovery machine cites when the next buyer asks who they should consider. Your delivered value at one customer becomes your eligibility at the next.
That is the compounding the old funnel never had. The line consumed budget to fill its top. The loop generates its own.
One substrate
Look at what all four machines want and it is the same thing.
Verifiable, published, machine-readable evidence of value delivered.
The discovery machine cites it. The audit machine needs it to justify your renewal. The orchestration machine routes on it. The citation machine publishes it. Which means you cannot win any of them without the instrument, and you cannot win any of them with marketing alone.
This is why the season is sequenced the way it is. Operating model, org, instrument, then this. Winning the machine is not a discipline you bolt on. It is what the loop looks like when every counterparty is an algorithm.
Consistency is the product now
Here is the part that matters more than any of the mechanics.
Doing this across all four stages produces something no individual tactic delivers. Consistency. The same verified evidence, the same claims, the same numbers, appearing at discovery, at audit, at routing and in the public record. A machine that reads you at four different points and finds the same answer four times is a machine that trusts you.
And trust is now the scarce commodity, for a reason the industry keeps skirting around. Agentic systems are probabilistic as much as they are deterministic. They estimate, they infer, they get things wrong in ways a database never did. Every enterprise deploying them knows this, which is why the buying decision has quietly shifted from capability to confidence.
You cannot make a probabilistic system deterministic. However you can be the vendor whose evidence is consistent everywhere it is checked, and in a market of confident-sounding uncertainty, that consistency is the differentiator. Not the demo. Not the benchmark. Whether what you claim holds up wherever a machine looks.
The human is the constant
Which brings me to the only fixed point in all of this.
Every machine described above is probabilistic. The discovery engine drifts, the audit agent infers, the orchestrator routes on incomplete information. Build a commercial model that depends on all four behaving predictably and you have built on sand.
The human in the loop is the constant. The person who defines what a verified outcome means. The person who decides when the evidence is good enough to publish. The person who takes the call when the machine's answer is wrong, and the person the customer wants when something matters.
That is not sentimentality about jobs. It is architecture. In a system where every automated component is probabilistic, the deterministic element is judgement, and judgement is still human. Every serious agentic deployment keeps a human in the loop for exactly this reason, and the ones that removed the human are the case studies now used as warnings.
So the machines get the evidence, consistently, at every stage. The human owns the judgement, permanently, at every stage. And the entire structure exists to serve the thing this newsletter has been building towards since the first issue.
Not the deal. Not the quarter. The forever customer, who is now evaluated by machines at every point in the relationship, and who will stay with whoever those machines consistently find worth staying with.
Win the machines. Keep the human. Build for the customer who never leaves.
Next issue: Pricing by Design. The hybrid architecture in full, and how to share efficiency gains with your customer on purpose.
First published in the Agentic Commercial Model newsletter on LinkedIn, August 27, 2026. Read the original on LinkedIn.