AI Automations
Automate what pays off. Not everything that is possible.
AI has become cheap. Mistakes have not. Before an assistant gets built, the arithmetic comes first: what does the manual step cost today? And what does it cost when the machine gets it wrong?
30 minutesfreevideo callno sales pitch
Prefer writing? contact@balane.tech
Typical situations
- Emails, orders or receipts are moved into another system by hand.
- The same customer questions keep coming in and tie up someone on your team.
- Enquiries have to be read and sorted before anyone can work on them.
- Every week, data is copied out of one system and pasted into another.
- Your team is trying out AI tools, but nobody checks what they produce.
- You want to know where AI pays off in your business and where it doesn’t.
What you get
Do the maths, then build
Frequency times duration times hourly rate. If the number doesn’t carry the build, I advise against it. That way you get a recommendation backed by arithmetic, not by gut feeling.
The failure case is designed first
What happens when the assistant misunderstands? Who notices, how fast, and how do you get back? That answer comes before the first line of code.
A human stays responsible
Every automation gets a point where somebody can intervene. Without it, nothing ships — however good the model is.
Recurring questions and tasks
An assistant answers recurring questions or pre-sorts incoming requests, outside office hours too. What it can’t handle reliably, it passes to a person instead of guessing.
The question is rarely whether something can be automated.
Almost everything can. The question is whether it pays off for you, and who is liable when it goes wrong.
So the start is not a technology choice but a number: how often the step occurs, how long it takes, what it costs. If the return is below the effort, I advise you against it.
Every automation first gets a thought-through failure case and a place where a human can step in. What nobody can supervise does not get built.
Find. Change. Measure.
First it becomes clear where your process gets stuck. Then that exact spot gets changed. At the end your own number shows whether it paid off.
- 01Find
Intro call, 30 minutes
Free, by video, no sales pitch. You describe which work keeps repeating. Afterwards you know whether automation could pay off.
Count the manual steps
Which work repeats, how often, how long does it take? That shows what automation would bring you.
- 02Change
Arithmetic and concept
Return against effort, plus the failure case: what happens when the machine gets it wrong, and who steps in?
Build and connect
The automation gets built and connected to your systems, such as email, CRM or accounting.
Test with real cases
Testing uses real cases from your daily work. Your team learns how to step in when something goes wrong.
- 03Measure
Run and refine
After launch, any place where the automation gets things wrong is tracked and corrected.
Re-measure the metric
The number agreed before the start gets measured again after go-live. If it does not move, things get adjusted.
Frequently asked questions
Typical are recurring tasks: reading incoming emails and documents and moving them into the right system, answering common customer questions, pre-sorting requests, arranging appointments. Which of these pays off for you is shown by the arithmetic up front.
A small automation starts at around €3,000, larger ones cost more. In the intro call you find out what return to expect, so you can decide on solid ground.
First, it gets counted how often a manual step occurs and what it costs. If automating it pays off, it gets built, connected to your systems and tested with real cases. A simple automation can be running in 2–3 weeks, larger ones take 4–8 weeks. If it is still unclear where automation pays off at all, the bottleneck diagnosis helps first.
A few real examples of the manual step, such as forwarded emails or receipts, and an estimate of how often it occurs. Plus access to the systems to be connected, and one person who checks the results during testing.
Usually yes: CRM systems such as HubSpot or Salesforce, email, calendars, accounting, WhatsApp, Slack and many more. The integration follows your system landscape.
It can be built in a GDPR-compliant way: with servers in the EU, privacy considered from the start and transparent processing. If you want, everything runs on your own infrastructure.
With me, from the first calculation to day-to-day operation. I build the automation myself. Where it makes sense, I bring in people I have known for years. I remain your point of contact.
Every automation has a defined failure case and a point where someone on your team steps in. After launch, it gets watched for where it misses, and adjusted. After that you can commission support by effort or on a fixed maintenance contract.
Related articles
What I work with
A selection. What matters is what suits your business.
- OpenAI GPT-4
- Claude AI
- Python
- LangChain
- n8n
- Make/Zapier
- Twilio
- WhatsApp API
- PostgreSQL
- AWS Lambda
The other areas
- Consulting & process optimisationI find out where your processes lose time and money, and change the process.
- Technology & IT consultingWhich systems you need, which you don’t, and how they fit together. Vendor-independent.
- Custom developmentSoftware, web applications and integrations built around how your business works.
Find out in 30 minutes whether this is worth it for you.
You describe where things get stuck. Afterwards you know whether and how I can help, even if the answer is no.
Jonas Höttler
Founder
30 minutesfreevideo callno sales pitch
Prefer writing? contact@balane.tech