About Throughline
Who builds this
Throughline is a small team of builders inside Keller Group. We build working tools for the paperwork that leaks money. Ten of them run in production today, inside the companies we come from.
Who answers
The team
There are two or three of us, depending on what's being built. Whoever answers your first email is also one of the builders.
Between us we built ten production systems and the platform they run on — that is the whole company. When you write to Throughline, one of the builders answers. There is no layer of account people between you and the person who did the work.
A shop this small, asking to sit inside your ERP, is a fair thing to be wary of. Weigh it against the work instead of taking it on trust. Every system we ship has a name, a job it does, and a company that depends on it. The full list is at /systems/.
Where this comes from
Origin
None of these systems started as a product. Each came out of a chore inside a working company, with real money attached to getting it right.
Keller Group's companies needed tools that didn't exist yet: a hub to route the day's work, a weekly KPI scorecard, an agenda for the leadership meeting, a CRM that remembers every call, a monitor on the ad accounts. We built those five, in that order: Primary, Scorecard, Cadence, Pensieve, Anomaly Detector. Ten production systems run now, built by a small team, inside working companies.
Some of the ten are plain. Scorecard collects weekly numbers. Cadence builds the agenda for one leadership meeting. It is not clever software. The meeting has run on it every week for months, which is the only test that counts. Four more do marketing work for the agency side of the family. The mistakes landed on us first, inside these systems and the companies that had to live with them. You are not the experiment.
The flagship is the rebate-and-margin tool we built inside Kelsan, a multi-state distributor in our own group. It runs inside their Epicor P21 ERP every morning. The tool uses AI to read the rebate terms in the contracts and check every line item against them, day after day. It has recovered over $100,000 in margin the manual process was missing.
Kelsan runs the tool inside its Epicor P21 every morning, and the number is real. Autoblogger works for clients outside the family today. We'll name them when they say we can.
The full write-up, numbers included, is at /k-1/.
Why 'Throughline'
A throughline is the thing that holds while everything else changes. That is what a tool has to be to earn a spot on this page — the same check, running day after day, whoever is at work that morning. The Kelsan build answers to that standard.
The longevity question
Keller Group
Worrying about the size of the team is fair. The answer is the family of operating companies behind it, and the systems running there today.
Throughline is part of Keller Group, the family of operating companies it grew out of. "We're not going anywhere" is easy to say, so here is the version you can check: the systems are in use inside those companies now. If Throughline never sold another build, Scorecard still owes them next week's numbers, and Anomaly Detector still has to watch the ad accounts on Monday morning.
What it costs
The cost
Every engagement starts with the AI Capability Audit. Fee, duration, deliverables, and what you keep — on one page.
The whole engagement runs from that one page: the fee, the fixed duration, the deliverables, and what you keep either way. It's all at /audit/.
How we work
Your people
What happens to your people is the right question. Here's the answer we build to, and the loop that enforces it.
Nothing changes in your business without a person's sign-off. The people who know the numbers best are the ones with the least time to chase them — so the tool does the chasing, and your people keep the judgment. We don't build dashboards to watch your own staff. And when we hand a tool over, we hand it to your team. Documentation and training are part of the package, built so the person at the desk can run it, not one specialist.
Campaign Manager, one of the systems that runs our own companies, checks its own recommendation for holes before it asks a person to approve anything, and the result is measured against a prediction written down in advance. Every system we build for a client inherits that shape: the tool recommends, a person approves, and we measure the outcome against the prediction.
Your IT company or MSP gets a veto before anything touches your systems, and we would rather they use it. The security one-pager is written for them: where the data lives, how the model cost works, and what we never touch. It's at /security/.
Part of Keller Group
The next step
You've seen who builds these and where they run. The rest is terms, and the terms fit on one page.
Fixed fee ($7,500)fixed durationa build-or-don't-build verdict you keep either way