Nurho / automation engineering
See if we're a fit

Your operations, engineered.

The work behind your business shouldn't run on manual labor and rented no-code. Nurho engineers a system that runs it for you, built to scale and yours to own.

See what I'd send your business. Drop your URL below.

One free run per email. Magic-link login, no password, and your data stays private. See privacy.

Rather talk first? See if we're a fit →

42 School engineer · Currently shipping production healthcare software in Finland
The problem

Manual work doesn't scale.
Neither does no-code.

01 / 03

Your team is doing the work software should be doing and every week it costs more.

Repetitive work that doesn't need a human, decisions that follow the same pattern every time, the manual hand-offs that bleed time between every “real” task. The cost isn't in any single slip, it's all of them, compounding into a number nobody puts on the books. Human-priced, machine-repeatable.

02 / 03

No-code tools break at scale, and no one knows why.

Zapier and Make are fine until they aren't. No version control, no real branching, no tests. Bills balloon, Zaps fail silently, and the graveyard of disabled automations grows every quarter.

03 / 03

Whatever you build now, you should still own in five years.

SaaS platforms rent you their version of your operations. Switch tools, or get switched on, and your data, your workflows, and your history walk out with the vendor. The businesses that win long-term own the pieces they run on.

Why Nurho is different

What you're actually buying.

A system on your own data and infrastructure, that you keep.

01

Built on your own data and infrastructure

The system runs on infrastructure you own and learns from your data, not a vendor's. No SaaS platform renting your own operations back to you. Open-source underneath, with the code and credentials in a repository that's yours, GDPR-aware by default. Switch tools or walk away whenever you want, and keep everything.

02

Agentic systems, engineered like software

Not steps wired together in a no-code tool that fails silently. Systems that read your data, decide, and act, built the way real software is: branched, tested on live data, shipped through staging to production. Agentic where that earns its keep, plain automation where that is all it needs. Custom logic no-code can't reach, with the guardrails real software has.

03

Priced against what it pays back

Each build is scoped to a single outcome and priced as a fraction of the value it creates. You see the math, and the projected payback, before any work begins. If the numbers don't work, we say so.

How it works

Find the constraint. Remove it. Repeat.

Every business has one bottleneck limiting growth. Find it, optimize the process, then automate. Optimize then automate, never the other way around.

01

Find the bottleneck

A short diagnostic to find the one constraint costing you the most right now, and to see if working together makes sense. I don't automate everything, only what matters.

30-min call
02

Optimize, then automate

Fix the process first, then build the system that removes the constraint, engineered on open-source infrastructure you own, never rented from a vendor. Never automate a broken process.

Optimize → build
03

You own it

Full handover: the code, the credentials, the documentation, in a repository that's yours. Fire me anytime and keep it all running. The system is yours, not mine.

Full ownership
04

Growth

Removing the right constraint doesn't just fix a problem, it raises the ceiling. Most engagements grow with you, system by system, paced to what you need next. Your call.

Grows with the business
What this looks like in practice

The engineering travels.

It becomes a different internal tool in every business, the same discipline underneath.

INTAKEENRICHREPLYROUTE
system.01 running
01 Leads answered in seconds, not hours.

A lead hits your form. By the time they've closed the tab, they're enriched, scored, routed to the right person, and a personalized first reply is drafted.

Form / email / chat → Enriched + scored → First reply drafted
RESEARCHVERIFYDRAFTSEND
system.02 running
02 Outreach that runs itself, 24/7.

Outreach that researches, personalizes, sends, and follows up, running continuously on infrastructure you own. Your team only ever sees the replies worth their time.

Trigger / cadence → Researched + verified → Personalized + sent
TRIGGERPROCESSSYNCNOTIFY
system.03 running
03 Admin work, quietly removed.

The repeat work no one wants to do (data entry, status syncs, reports, document handling) runs in the background and lands clean on the other side. Your team gets their week back.

Source event → Processed + stored → Systems updated
Where this goes

One bottleneck now.
Your own internal tooling over time.

We don't build one automation and disappear. Each system is a piece of something larger: your own AI-native internal tooling, shaped to how your business actually runs, on your data and your infrastructure. We start with the one constraint costing you the most, prove it pays back, then build the next piece when you're ready. A few systems in, you have software that fits your operations exactly, that no competitor can buy off a shelf, and that you own outright.

On pricing

Priced against
the value created.

01 / Outcome-scoped

Every engagement is priced as a fraction of the value it creates. One number, one scope, no surprises.

02 / Known before we start

If the math doesn't work, we won't pretend it does. You'll know the cost and the projected payback before any work begins.

03 / The math, on the call

Real numbers depend on which bottleneck we're removing, so they live on the call, not in a pricing tier. Book the call and we'll do the math together.

See if we're a fit

Engagements typically start in the low thousands.

FAQ

Real questions. Real answers.

Every engagement has a one-time build fee plus an optional ongoing retainer if you want me running it for you. Both are sized to the value the system creates, not to hours or seats. You'll see the numbers, and the projected payback, before any work begins. If the math doesn't work, we won't pretend it does.
Simple workflows ship in a day, complex ones in about a week, including testing. Speed comes from having a proper engineering pipeline behind the work: closed testing, staging, production, the same CI/CD discipline used to ship real software. You see tests pass on real data before anything goes live. I diagnose the bottleneck on the call before scoping, so the exact timeline depends on what we're removing together, not on a fixed schedule.
No. You explain what's painful and where the money is leaking. I handle the engineering side and translate everything important back into business language. The deliverable is a working system and clean documentation, so you don't need to know how to read the code to own it.
Most prospects already know exactly what's eating their week, they just haven't framed it as a bottleneck, or they assumed it wasn't the kind of thing software could handle. The first call is mostly translation: you describe the friction in your business, I show you which parts a system can absorb. You don't need a spec. You need an honest description of what's slow, expensive, or repeating.
Engineering judgment travels. It lives in HOW I work, not WHICH industry. I build production software for a living, including a GDPR-compliant scheduling product in Finnish healthcare, so the engineering rigor is real and active, not theoretical. The discipline of finding a bottleneck, optimizing the process, and engineering the system around it is the same whether you run a marketing agency, a real-estate operation, or a Nordic SMB. Whether your specific operations are a fit is what we figure out on the call, both ways.
Both work, until they don't. No version control, no real branching, no tests, no custom logic. Bills balloon as you grow. And you own none of it: cancel the subscription and the workflows disappear. Nurho builds on open-source orchestration tooling, deployed on infrastructure you own, so what you pay for is yours, not rented.
Yes, it's one of the most common first engagements. I audit what's actually running, rebuild what's worth keeping on your own infrastructure, delete what shouldn't exist, and version-control the result. You walk away with a system, not a subscription.
A VA runs 40 hours a week, churns, makes mistakes, needs management, and costs the same every month forever. A well-built automation runs 24/7, never quits, and gets cheaper per execution as you grow. The ROI math is direct, and we'll do it together on the call.
A few things. The system is monitored continuously: if something fails, an alert hits Slack, Telegram, or wherever your team already lives, with the failure context and the affected workflow. You see it before your customers do. On higher-tier builds, an agent triages the failure automatically: diagnoses the cause, opens a PR with a fix, and pings you for review, so the system can repair itself before you even check your phone. Either way, the infrastructure is on your account, the runbooks are in your repository, and production-grade systems are designed to fail safely, not silently.
Everything runs on infrastructure you own. The code, the credentials, the documentation are in a repository you control. Any developer with relevant experience can pick it up tomorrow. You're not buying access to me, you're buying a system you keep.
Yes. The US, UK and Europe are all covered. Most work is async, with meaningful overlap with every client's working day for calls and iteration.
GDPR-aware by default. Everything self-hosted on infrastructure you control, no third-party data brokers, no opaque sub-processors. EU-resident hosting on request. Reasonable security primitives (audit logs, PII redaction, scoped credentials) are part of the build, not an upsell.
Systems that do more than follow a fixed script. They read your data, decide what to do, act, then check their work. A simple example: an outreach system that researches a lead, drafts a personalized message, and only surfaces the ones worth your time. A more involved one: a system that notices a failure, diagnoses the cause, and opens a fix for your review. Where a plain automation is enough, that is what you get. Agentic is used where the judgment earns it, not as a label.
No. Your data stays on infrastructure you own and is never used to train a shared model. Where a step calls an AI model, it sends only what that step needs, on your accounts, to providers configured for no-retention where that option exists. Your operations and your data do not become someone else's training set.
Book a call

Book a call.

Thirty minutes. I find the bottleneck that's costing you the most right now, and you walk away with an answer either way. No slide deck, no discovery theatre.

See if we're a fit or email hello@nurho.net