Software for exactly one sales team

Ask a sales team what they wish the CRM did and you will not get a feature request. You will get a description of their company.

They want lead scoring that reflects the two industries where their win rate is actually good, rather than the generic fit model. The quote calculator should know about the regional discount the VP approved in March. And somewhere in there is the territory split that has to account for the account manager who covers three states plus one very large customer in a fourth.

None of this is hard. All of it is specific. And specificity is precisely what software vendors cannot sell you, because a feature that serves one company is a feature they cannot amortize.

So sales teams have always lived in the gap: a CRM that is 80% right, and a spreadsheet holding the other 20% together, rebuilt every quarter and emailed round again.

What changed

The reason nobody built the missing 20% was not that it was difficult. It was that it was not worth a developer’s month.

Getting an engineer to build a commission calculator for eleven reps means writing a ticket and then waiting a quarter to explain your compensation plan to somebody who has never carried a bag. The tool would be worth maybe forty hours a year of saved time. The math never worked.

The math works now. Describing your comp plan to an assistant takes about as long as describing it to a person, and the assistant produces the calculator while you are still talking.

That is a real change in what is buildable, and it lands hardest exactly where software was always too generic: the parts of your process that are yours.

What people are building

The published examples are unglamorous and that is the point.

A marketing agency wired up business card capture at conferences. Photograph the card, an assistant reads it, the contact lands in the CRM with a follow-up sequence already attached to it. They put the saving at around five hours a week.

A B2B lead generation firm runs call transcripts through an agent that compares them against what good calls look like and posts the notable moments into Slack. Not call scoring as a product category, just this team’s version of it.

Another agency built lead scoring on their own behavioral signals rather than the vendor’s model.

Notice these are not attempts to replace the CRM. They are the missing 20%, finally built.

Where it stops

Here is the part the demos skip.

Your commission calculator works. It is correct, you tested it against last quarter, and it lives in a folder on your laptop.

Now put it in front of eleven reps.

They are not going to install anything. They are not going to run a command. Half of them will be doing this on a phone between meetings. What they need is a link, and what a link needs is a server, a domain, a certificate, and a decision about who is allowed to see the numbers. That last one is not optional, because a comp calculator shows people money.

This is where most of these tools die. Not from a bug. From the absence of a place to live.

The advice you get is not for you

If you go looking for help with this, you will find engineers answering the question for other engineers.

There is a thread on Reddit from July where someone asks a version of exactly this question: they built agents themselves, they cannot leave their computer running, and they want to know where to put them. The advice is confident and completely inaccessible to anyone outside the profession.

dude just get a cheap vps, you’re overthinking this. digital ocean droplet or a hetzner box is like…

That is the friendliest answer in the thread. It assumes you know what a VPS is, that you can SSH into it, and that a cron job is a thing you can write.

The most useful comment in the thread is also the most discouraging one:

For two agents posting to Instagram, the provider is almost never the bottleneck. The silent failure modes and operational reliability are.

The same commenter’s advice is to add structured logging and a dead man’s switch before worrying about scale. That is correct. It is also a full skill set that has nothing to do with sales, and nobody in a revenue org is going to acquire it to ship a quote calculator.

Someone else in the thread describes a script that slowly ate memory and failed quietly, which is the exact failure mode that turns a useful internal tool into a tool nobody trusts. Silent wrongness is worse than an outage. An outage gets reported. A commission number that is quietly wrong sends one rep back to their own spreadsheet and you never hear about it.

What good looks like from a sales seat

Strip out the infrastructure vocabulary and the requirements are short.

It has to be a link. Not an install, not a download, and it has to work on a phone in a parking lot.

Sign-in has to be your work email, and it has to be there by default. A tool showing commission data cannot be protected by the fact that the URL is hard to guess.

Nobody should have to babysit it. If it breaks, the person who built it should be able to point the same assistant at the error and get it fixed, without learning what a log stream is.

And it should cost roughly nothing while nobody is using it, because a quarterly planning tool is genuinely idle for eleven weeks out of thirteen.

None of that is exotic. It is just a different set of defaults than the clouds were built with, because those clouds were built for software that serves a million people rather than software that serves eleven.

The honest close

Sploot is being built for this gap: hand over the folder your assistant wrote, get back a link your team opens with their work email. It is in private beta, so there is nothing for you to go try today, and I would rather say that plainly than imply otherwise.

What is true regardless is the shift underneath. The reason your team’s specific tools never got built was economic, not technical, and that reason is gone. The constraint that remains is a boring one about where software lives, and boring constraints tend to be the ones that actually get fixed.