Start with the number you cannot see. A mid-market operator loses revenue every week in places no dashboard tracks. A missed call after hours. A lead that sits in a form for three days. A cart that never gets a recovery email. None of it shows up as a line item. All of it adds up.
We call that the Open Loop Tax. It is the cost of work that falls between your tools because no one owns the handoff. This post explains how a closed-loop agency removes that tax, why luup became kratt, and what build/host/run changes for your P&L. We will put a dollar figure on the leak, then show how each worker closes one gap at a time.
What a closed-loop agency is
Most agencies sell you a plan. They hand over a deck, a login, and an invoice. Then they leave. The strategy is sound. The execution is yours. That gap is where the work dies.
A closed-loop agency closes that gap. We design the system, build it, host it, and run it. We stay on the outcome, not the recommendation. The loop runs from "here is your leak" to "here is the money back" without a handoff in the middle. Harvard Business Review has documented for years how strategy fails at the execution seam. We refuse to ship the seam.
The difference shows up in who you call when something breaks. With a handoff agency, you call yourself. The login is yours. The maintenance is yours. With a closed-loop agency, you call us, and we already saw the alert. The loop stays closed because one team owns both halves.
Why luup became kratt
luup named the mechanism. A loop. Useful, but abstract. kratt names the promise.
A kratt is a figure from Estonian folklore. You build it from scrap, hay, old tools, broken things. You give it life, and it hauls wealth back to your house. It does the labor while you sleep. The money comes home.
That is the work. The AI is the kratt: assembled from your existing stack, pointed at a leak, sent out to haul revenue back. The name now matches the job. Build a worker. Point it at the gap. Send the money home. The rebrand was not cosmetic. It changed how we talk to operators, because the myth carries the whole pitch in one image.
The folklore is not decoration. It is the operating model. A kratt only works if someone builds it well and aims it correctly. A badly built kratt burns the house down. The myth warns about the failure mode we see in AI consulting: a tool nobody owns, running unsupervised, doing damage.
So we treat every AI worker the way the folklore treats a kratt. Built from real parts. Aimed at one job. Watched while it runs. That is the difference between a closed-loop agency and a vendor who sells you a chatbot and disappears.
The myth also explains the order of work. You do not summon a kratt and hope. You gather the parts first, then give it one task. We do the same. Groundwork first, one worker next, one leak closed before the second worker gets built.
Build, host, run. Not handoff.
Three words carry the whole model.
- Build. We ship the working system. Not a recommendation, not a Notion doc. A voice agent that answers. A workflow that fires.
- Host. It lives on infrastructure we maintain. You never babysit a Make scenario or a Vapi number at midnight.
- Run. We monitor it, fix it, and improve it while it is live. You point at a recovered number. You do not learn a new stack.
Handoff agencies stop at build, and most stop before that. Gartner research on failed automation projects keeps landing on the same cause: nobody owned the system after launch. Run is the word that fixes it.
Run is also the hardest word to fake. A demo can be built in a day. A system that survives a year of real traffic needs someone watching it. Models drift. APIs change. A phone provider updates a setting and a script breaks. The closed loop means we catch that, not you. That is the part you are paying for.
The Open Loop Tax, with the math
Here is where the abstract becomes a dollar figure. The numbers below are illustrative, modeled on a $4M services business. Yours will differ. The shape will not.
| Open loop | What leaks | Illustrative monthly cost |
|---|---|---|
| After-hours calls | 30 missed calls, 4 would have bought | $8,000 (illustrative) |
| Slow lead follow-up | Leads cool after 3 days in a form | $6,500 (illustrative) |
| No cart recovery | Abandoned checkouts, zero email | $4,200 (illustrative) |
| Off-brand assets | Ad spend on weak creative | $3,000 (illustrative) |
| Manual reporting | 12 founder hours/week, no insight | $5,000 (illustrative) |
| Total tax | Five open loops | $26,700/mo (illustrative) |
Run that across a year. The open loop costs more than six figures in this model. No single number screams. Together they bleed. That is why operators miss it. The leak hides in the gaps, not inside any one tool.
Notice how each row maps to a tool boundary. The phone does not talk to the calendar. The form does not talk to the CRM. The store does not talk to email. Every gap sits where one tool ends and the next begins. The tax is the price of those silent borders. Close the border, and the row goes to zero.
The Closed Loop Audit: our one front door
We have a single entry point. A free AI audit. It maps your business and ranks every revenue leak by impact, scored in illustrative dollars so you see the biggest one first.
No obligation. No 40-page report you never read. You get a ranked list. The top item is where we would start. If you move forward, we build, host, and run the AI that closes it. The audit is the front door because the leak should choose the project, not the salesperson.
The audit takes about 30 minutes of your time. We look at your stack, your call logs, your form fills, and your follow-up timing. Then we rank. Most operators are surprised by which leak lands on top. It is rarely the one they were worried about. That is the point of measuring before building.
You can run it here, or read more about how we work on the about page. The case studies show the pattern on real operator problems.
Groundwork first: one connected brain
You cannot point a kratt at a leak if it cannot see the business. So we lay groundwork before we build workers. We connect your data into one brain the AI can think with: CRM, calendar, store, billing, support.
Most stacks are islands. Salesforce does not talk to Shopify. HubSpot does not see Klaviyo events. We wire them through n8n and Make so every worker shares one source of truth. Groundwork is unglamorous. It is also why the workers function at all.
Skip this step and you get the AI horror stories you have read about. A bot that books a slot already taken. An agent that quotes a price the system changed last week. The brain is the fix. One source of truth means every worker sees the same business. Build the brain once, and every worker after that gets cheaper and safer to ship.
The 4-System Stack
Once the brain is connected, the workers plug in. We build from four systems. Each closes a class of open loop.
- Voice Agents. 24/7 inbound and outbound, built on Vapi, ElevenLabs, Twilio, and Cal.com. The agent answers the 9pm call and books the slot. See voice agents.
- Automation. Operational and reporting workflows that route everything and return founder hours. See automation for agencies.
- Ad Factory and Brand DNA. On-brand assets at volume, so ad spend stops funding weak creative.
- Website and Vibe Code. Launch-ready creative shipped fast, not redesigned every 18 months.
You do not buy all four. The audit decides which leak is biggest. We build that worker first. Then we measure. Only when the first number moves do we point the next worker at the next leak. One closed loop funds the one after it.
A worked example: the voice worker
Take the first row of the table. 30 missed calls a month, 4 of them buyers, $8,000 in illustrative lost revenue. That is one open loop.
The kratt for that loop is a voice agent. We build it on Vapi for orchestration, ElevenLabs for the voice, and Twilio for the line. It answers every call, qualifies, and books straight into Cal.com. It writes the lead back to the CRM through the connected brain.
We host it. We watch the transcripts. We tune the script when a pattern breaks. You see the calendar fill. One loop closed, one number recovered, zero new logins for your team.
Walk the math forward. If the agent saves even 2 of those 4 buyers a month, that is half the row recovered in this model. The agent runs at a flat monthly cost, not a salary. It does not call in sick and it does not skip the 9pm call. The gap that taxed you every night is now a booking instead.
The Brand DNA Sprint
The Ad Factory needs fuel. That fuel is your Brand DNA, codified once. We run a short sprint to capture voice, palette, claims, and proof into a single system. We build it in Figma so every asset after that is faster and cheaper.
Brand DNA is an asset on the balance sheet. Codify it once. Every channel gets cheaper forever. That is the second open loop closed: ad spend stops funding off-brand creative because the system cannot produce it.
The Recovery Guarantee
A closed-loop agency owns the outcome, so we put it in writing. Your revenue stops leaking, or we work free until it does. No lock-in.
That guarantee only works because we host and run the system. A handoff agency cannot promise it, because after the handoff they control nothing. We control the worker the whole time it is live. So we can stand behind the number. OpenAI and McKinsey both keep noting the same gap between AI pilots and AI in production. The guarantee is how we force ourselves across it.
The no-lock-in clause matters as much as the free-work clause. If we cannot move the number, you should be able to walk. A guarantee that traps you is not a guarantee. Ours is built so the incentive sits on our side: we get paid when the leak closes, and we keep working when it does not.
Who runs the loop: 2 people, clear roles
kratt is two operators. Oskar builds and ships the systems. Karl-Kristjan runs the audit and owns the outcome. No account managers, no relay race, no fourth person who never touched the code.
Small is the point. The folklore kratt has one master. Our workers do too. When something breaks at 2am, the person who built it is the person who fixes it. That is what owning the outcome means in practice. We cap onboarding at 4 slots a month so it stays true.
Where to start
The open loop is already taxing you. The table above is one model of where. Yours is sitting in your CRM right now, ranked by nobody.
So rank it. The audit names the biggest leak in illustrative dollars, and the top item becomes the first worker we build, host, and run. Read more on the blog if you want the longer thinking first. When you are ready, run the free AI audit at our audit page and we will tell you exactly where your revenue is leaking.