There is a rep on your floor who prices deals in a tool nobody bought. They built it themselves, a spreadsheet at first and then something closer to an app, because the quoting screen in the CRM was configured for a deal they do not sell. It is not rebellion. It is the last mile of a job the system never finished, and the rep finished it alone. The company already runs on a handful of tools like this one. You know which CRM you pay for. You do not know which tools actually win your deals. The rep is an asset. The unmapped build is a risk. This piece is about naming that pattern while the rep is still in the building.
The rep who stopped waiting
The CRM has told reps for years how selling is supposed to look: which stages exist, what counts as a deal worth reporting, who owns the next step. The model fits the average sale of an average company. It has never fit the specific sale this rep makes, with its own margins, approvals, paperwork, and a rhythm the system cannot see.
For years the choice was compliance or exit: enter the data, take the discount the screen allows, or leave the number off and argue about it in the forecast call. The rep who is handy found a third path. They built a layer on top of the CRM, or around it, that prices and tracks the way they actually sell. The CRM still gets its update, and the rep's real pipeline is tracked too. Two versions of the truth now exist, and only one of them is the one deals get won on.
What they build instead
The builds are plain, and they follow the work. A quoting tool that prices the way the rep actually sells, not the way the system was configured. A script writer that turns the rep's best calls into repeatable talk tracks. A research assistant that reads the laws and regulations governing their deals, so nobody pages through them at the deadline. Small helpers that take the filing and follow-up the CRM makes drudgery and do it in the background.
None of it looks like software to the people who approve budgets. It looks like a spreadsheet, a saved prompt, a folder of drafts. But every one of them has logic, inputs, outputs, and rules, and the rules decide what a customer is quoted and when a deal is real. That is software. The rep built it without a budget line, doing alone what a vendor would sell as a long implementation.
This instinct is not new. People have built their own work software since 1985, when the spreadsheet put a real programming environment on an office desk. Your commission sheet, your price list, your capacity tracker: the company already runs on end-user builds. AI lowered the cost of the next one, and the rep was the first person in the building to notice.
Why nobody ever said yes
The rep never asked permission. The vendor answer to a missing feature is a roadmap, or a checkbox buried two tiers up the pricing page. The IT answer is a ticket with no owner and no date. Neither answer is malice. Both are capacity. One rep's selling problem is a rounding error on a product roadmap and a disruption in an IT queue, and the rep knows the request will never surface.
They built quietly. The tool worked, the company kept running, and nobody noticed, which is exactly what the rep needed. The CRM is not the villain in this story. It was built to fit the average deal, and the average deal is a fiction no individual rep ever closes. The gap between the system's idea of selling and the rep's actual selling is the point. Someone was always going to fill it.
The market already repriced for this
The market has started pricing the gap. Klarna dropped Salesforce's flagship CRM product in late 2024 and replaced it with a homegrown AI system. Forbes put the early February session loss at roughly $300 billion in software market value. TechCrunch tracked the fallout: nearly a trillion dollars erased from software and services stocks within weeks, and forward earnings multiples down from about 39 times to about 21. One Zoho chief executive called AI "the pin that is popping this inflated balloon."
None of that is a verdict on the CRM. It is a verdict on the price of software when the price of building software collapses. In April, a Retool survey covered by VentureBeat found 78% of teams planning to build custom tools by the end of 2026. Most of those builds will not cross a procurement desk. They will happen at desks, the way the quoting tool did. The price of building collapsed. The price of saying precisely what you want did not.
The asset and the unmapped risk
Hold both halves at once. The rep who builds is an asset. They encode the real rules of your business, the exceptions no vendor ever learned, into working tools, on your payroll, for free. A consultant would charge six figures to extract half of what that rep knows and would still get it wrong, because the knowledge lives in the rep, not in any document they could hand over.
The unmapped build is a risk of the same size. The quoting tool holds customer data in a place nobody audits. Its output has never been reviewed by anyone who owns the business rule. Its logic lives in one head, and if the rep takes a job across town, the tool stops being maintained and nobody knows where to start. None of this is a reason to stop the rep. The building is the point. The risk is that nobody mapped the estate, named the builder, or set rules around it, and AI widened the blast radius, because the next ten reps can build the next ten tools just as easily.
What to do
For the sales leader or owner, the job is small.
- Map it. Ask the reps, not the CRM administrator: what do you quote from, and who built it? Put every answer on one page: what the tool does, what data it touches, who maintains it, who reviews it.
- Name it. Call the builds what they are: internal tools. Give them an owner, a folder, and a line in the conversation. Say the sentence out loud: keep building, and tell us what you are building.
- Fund the time. An hour a week and a real connection to the data is the whole budget. The rep supplies the energy. You supply the access and the reviewer, and you stop making them ask.
- Set the two rules before anything else ships. Rule one: where the data lives. The CRM stays the system of record, and the tool reads it instead of holding a second private copy. Rule two: who reviews the output. A named person who owns the business rule signs off before a quote reaches a customer. Put both rules on one page and post it where the reps can see it.
If you are the rep who built the tool, keep building, and tell someone. Hiding made sense when the only answers were a roadmap and a ticket with no owner. The page above is the answer that makes hiding unnecessary.
If you suspect a rep is building around the CRM, or you are that rep, [email protected] is the fastest conversation worth having.