A customer tells you they could rebuild your product themselves in a weekend. Maybe someone on their team already did a piece of it. The renewal conversation, which was going fine, is now a pricing conversation.
Before anything else: some downward pressure is real and has nothing to do with you. Budgets tighten. Categories mature. Buyers get more sophisticated about what things should cost. I am not going to pretend that every price objection is a positioning failure, because that is a comfortable story and it is not true.
But the specific pressure that arrived in the last two years has a specific mechanism, and it is worth being precise about what actually moved.
Nothing Happened to Your Product
Walk through how it usually goes.
A customer is using your product. It is working. There is no service failure, no missed outcome, no complaint sitting in a queue. Then somebody on their team needs a report your tool does not produce the way they want it, so they open a model, connect their billing system, and get the report in four minutes.
Your product did not change that day. Their estimate of what it would cost to live without it did.
Value perception does not respond to what your product does. It responds to what the customer believes the alternative costs.
That is the entire event. And it explains why the pressure can arrive on accounts that are healthy by every measure you track, which is the part that confuses people. You go looking for a satisfaction problem and there isn't one.
Two Different Threats That Get Talked About as One
The first is customer self-replacement. They do not build a competitor to sell. They build the version of your thing that they specifically need, which is a much smaller job than the one you did, because they only have to satisfy one customer and they already know exactly what that customer wants.
The second is being disintermediated from above. If your product is a thin layer over a model, the company that makes the model can absorb what you do whenever the category gets big enough or horizontal enough to be worth the effort. That is not a heavy lift for them. It is a feature release.
I paid about ten dollars a month for an AI presentation tool for the better part of a year. I stopped when the model I already used could produce the same thing directly. That was not a competitor beating them. It was the floor rising underneath them.
The Collapse That Did Not Happen
There was a stretch, not long ago, when this got talked about as the end of commercial software. Companies were said to be worth dramatically less because anyone could rebuild them over a weekend.
That turned out to be mostly noise, for a reason that should be obvious and was not obvious at the time. You can build a customer relationship application in a weekend. You cannot build a thirty-year-old customer relationship business in a weekend: the ecosystem, the integrations everyone already built against it, the uptime, the disaster recovery, the compliance posture, the entire apparatus nobody thinks about because it has never once failed them.
So the collapse did not arrive. But it left something behind, and the residue is the real story.
It made a lot of founders sit down and ask whether they had the defensibility they assumed they had. And it handed every buyer a lever.
The Discount as Protection Money
Customers used it, and some of them used it explicitly. The shape of the conversation is: I have the tools to replace this, and a discount would make the question go away.
That is not a pricing negotiation. It is a bet on your confidence, and most vendors pay it, because they are not sure the customer is wrong.
Here is where the vendor's own knowledge works against them. If you build software, you know how fast software can be built now. You have watched it happen on your own screen. You also know every competitor in your category, including the ones with two users, one of whom is related to the founder.
Your customer knows none of that. They know the two or three real options, and they are not moving to something unproven, because nobody's career improved by championing the vendor nobody had heard of.
So the vendor negotiates against a threat model far more sophisticated and far more frightening than the one the buyer actually holds.
The Artifact Was Never the Business
A founder I know had their source code taken by someone who had worked on it, who then tried to hold it for payment on the theory that possession of the code meant possession of the company.
It did not, and it never does. Ask what that person could actually do with it. Could they build the business? They would need the judgment about what to build next, the relationships that produce the first customers, the understanding of why the product is shaped the way it is, the willingness to keep going when it is not working. The code is a record of decisions. It is not the capacity to make them.
The advice I gave, at that stage, was to rewrite it. Anyone who has lost a draft knows why. The second version is tighter, because you no longer write the parts that were never load-bearing, and you write it fast and slightly angry, which turns out to be a good state for shipping.
Hold that next to the pricing conversation. If your entire business could be transferred by copying a directory, then the customer threatening to rebuild you is right, and no discount is going to fix that. If it could not, then they are wrong, and the only question left is why they believe otherwise.
The Answer, Said Out Loud
The response to a customer who says they could build this themselves is not a discount and it is not a defense of your feature set.
It is this: if the only thing you think you are buying here is access to software, then we have done a terrible job of positioning what we actually do, and we would deserve to lose your business.
Say it and mean it. Because either it is true, in which case the discount was never going to save the account and you have finally learned something, or it is not true, in which case you have just moved the conversation to the ground you should have been standing on the entire time.
What is on that ground: the workflows you encoded because you have watched hundreds of companies do this badly. The defaults that are correct because you learned what wrong looks like. The judgment about what not to build. What you know from the rest of your customer base that no individual customer can see. None of that copies over when someone recreates your interface.
Do Not Sell the Keyboard
There is a version of this failure that is entirely self-inflicted, and it comes from confusing the interface with the value.
I am writing this from a voice memo. If my keyboard stopped working I would keep going, slightly less comfortably. The keyboard is genuinely useful and it is not what produces the writing. A company selling keyboards on the premise that value lives in the act of typing has a serious problem coming.
Plenty of software is sold that way: the value proposition is that people log in and use it. When usage is the value proposition, anything that produces a similar output through a different surface reads as a complete substitute. Increasingly your product will be consumed by something other than your interface, by an agent or a workflow the customer assembled themselves, and if you have taught them that logging in is the point then you have taught them that a different login is an even trade.
Wrapping a Model Is Fine. Wrapping Only a Model Is Not.
Being built on a model is not the problem. The question is what else is inside the wrapper.
A wrapper around a model alone gets absorbed. A wrapper around decades of accumulated expertise, in a domain where the customer knows they lack that expertise, does not, because the customer can tell the difference between the raw output and the output that knows what it is doing.
The positioning that works is not defensive about the model. It is the opposite: the model is remarkable, and the model alone will not get you there. Vendors who hedge about how much AI is inside their product are answering a question nobody asked while leaving the real one unaddressed.
Why This Blocks Expansion
Here is why this belongs in a body of work about expansion revenue rather than in a marketing conversation.
You cannot ask a customer for more money while they are actively arguing that you are worth less. It does not matter how good the next thing you would sell them is. The inventory, the milestones, the orchestration, all of it stalls, because every one of those conversations opens on a foundation the customer is currently disputing.
It works exactly like delivery-failure churn. When you have not produced the outcome you promised, expansion is not slow, it is precluded, and the work is delivery. This is the same structure with a different cause. When the customer does not believe what you are worth, expansion is precluded, and the work is positioning.
Fix that first. Then the rest of the machinery has something to stand on.
Next Steps
The useful question is not whether AI can replace you. Sometimes it can. The useful question is whether you lost the account to a capability or to a perception, because those have completely different remedies and only one of them is about your product.
Perception is controllable. It is a thing you build, deliberately, over time, in every conversation where you explain what the customer is actually paying for. Most vendors have simply never done that work, because for fifteen years they did not have to.
So look for the controllables. They are usually where the money is.
Once the ground is solid, the question becomes what your existing customers should already be buying and are not. That is latent revenue, and finding roughly what yours is takes about ninety seconds.
Lincoln Murphy formally named and popularized Customer Success starting in 2010 and has spent 15 years connecting it to expansion revenue and commercial outcomes. Read The Premise.