AI & Automation
AI automation for the work your team repeats every day
We use AI where it saves real work: sorting inquiries, preparing drafts, answering questions from your own content. A person reviews anything before it reaches a customer.
- Start
- A small pilot with a fixed scope
- Review
- A person approves before anything goes out
- Cost
- Fixed price to build, model usage billed as used
- Contact
- Lucas Kleipoedszus, from first call to launch
The problem
Where the hours go
The processes worth automating are rarely exciting. They still eat hours every week.
Repetitive manual work
The same steps every day: copying data, renaming files, writing status emails. Mistakes creep in while retyping.
Inquiries typed in twice
A request arrives by form or email and someone copies it by hand into the CRM, a spreadsheet or the ticket system.
A growing content backlog
Product copy, articles and descriptions sit on the list. Nobody has time for the first draft.
The same questions, again
Support and sales answer questions that are already covered on the website, in the manual or in the FAQ.
Our approach
Start with the process, then pick the tool
First we look at which steps cost you time today. Then we decide whether fixed rules are enough or a language model is needed; plenty of automation works without AI. Where AI writes text or suggests decisions, your team approves it. AI is the wrong tool where every answer has to be exactly right and nobody checks it, such as legal advice or binding prices.
What you get
What we build
AI inside your existing website
Summaries, writing help or classification as a feature of the site you already have. API keys stay on the server.
Assistants grounded in your content
An assistant retrieves answers from your pages, documents and FAQ and cites the source. Anything not covered there goes to your team.
AI agents with approval steps
Agents handle several steps on their own, such as looking up data and preparing a draft. Before any action that leaves the building, they wait for approval.
Workflow automation
Fixed flows between forms, inbox, spreadsheets and calendar. Plain rules wherever rules will do.
Inquiries as structured data
Incoming requests are read, split into fields and passed to your CRM, inbox or ticket system. Unclear cases go to a person.
AI in WordPress
Drafts for editors, answers to product questions, internal search across your content. For the platform itself, see our WordPress page.
Logging and cost limits
Traceable logs of AI calls without unnecessary personal data. Usage limits keep model costs within the agreed budget.
Demo
From brief to draft, with approval
Try how a short content brief turns into a structured WordPress draft that is published only after a person reviews it. The demo is a simulation in your browser: no real AI runs and no data is sent.
Simulation · no real AI, no data is sent
- Received
- In progress
- Human review
Heat pump subsidies for existing homes: a guide for owners of single-family homes
Heat pump subsidies for existing homes: subsidy programmes that may apply, steps from application to payout and what the installer takes care of.
- H2Subsidy programmes that may apply
- H2Steps from application to payout
- H2What the installer takes care of
- H2Your next step
Thinking about "Heat pump subsidies for existing homes"? Here, owners of single-family homes find the key answers step by step: subsidy programmes that may apply, steps from application to payout and what the installer takes care of.
- Back up claims on "subsidy programmes that may apply" with a current source
- Back up claims on "steps from application to payout" with a current source
- Back up claims on "what the installer takes care of" with a current source
- Check figures, deadlines and amounts against the original source
- Approve tone and wording for owners of single-family homes
Use cases
Typical use cases
Processes that work well as a pilot. We figure out which one fits you on the first call.
Inquiry intake
Contact form → request, budget and deadline as fields → CRM entry plus a reply draft in your inbox.
Support assistant
Answers common questions from your manual and FAQ, cites the source and hands anything unclear to your team.
Editorial drafts
A short brief becomes a structured draft in WordPress, which an editor reviews and publishes.
Product questions in your shop
Customers ask about size, material or compatibility. The answer is based on your product data.
Internal search
Your team finds policies, templates and past answers by asking a question in plain language.
Document triage
Incoming PDFs and emails are recognised by type, named and attached to the right case.
Process
Four steps to a working system
Discover
We map the process as it runs today: who does what, which data moves where, and where errors happen.
Design
We decide what runs automatically, where AI helps, where a person approves and which data the system may see.
Build
We build a pilot using real examples from your day-to-day work. You test it before it goes live.
Launch & Improve
Go-live with logging and a cost overview. Then we extend what has proven useful.
Pricing
Fixed price to build, usage-based model costs
There are no package prices for AI and automation work; processes and systems differ too much. We quote the build at a fixed price. On top of that, the model provider bills usage (tokens). We estimate those running costs in the quote based on the volume you expect.
- Free 15-minute intro call
- Goals and scope agreed in writing
- Fixed-price quote including an estimate of running model costs
- Work starts only after your written approval
Proof
What we can show
We have no client AI projects to point to yet. What we can show is automation running in production, plus platforms and WordPress sites we built. That is why we start with a small, fixed-scope pilot.
-
Our own customer pipeline
Our website packages run on an automation in production: Stripe checkout, an onboarding form with image upload, every order logged as an issue in GitHub, a status page with follow-up questions, and confirmation emails. If filing fails, the order reaches us by email. The pipeline runs on fixed rules, without AI.
Read the case study: Our own customer pipeline (in German) -
Pfotenweb
A platform for animal-welfare groups: club website, animal profiles, an inbox for adoption inquiries and contracts in one place.
Read the case study: Pfotenweb (in German) -
FiveRP
A WordPress portal with a FiveM server list, guides and resources for players and server owners.
Read the case study: FiveRP (in German)
Questions
Questions about AI and automation
Does every automation need AI?
No. Many processes are automated more reliably and more cheaply with fixed rules. We use AI when text has to be understood or written, for example free-form inquiries.
How do you handle data protection?
API keys live on the server only. The model receives only the data the task needs. You sign a data processing agreement with the model provider, and where the provider offers processing in the EU, we use it. Whether a provider is acceptable for your case is for you and your data protection officer to decide; we do not give legal advice.
Can the AI get things wrong?
Yes. Language models can state wrong things convincingly. That is why a person reviews by default before anything reaches a customer, and assistants rely on your approved content and cite their source.
Which AI model do you use?
That depends on the task, your data protection requirements and cost. We connect the model so it can be swapped later.
What does it cost to run?
Model providers bill by usage, usually by the amount of text processed (tokens). The total depends on how often the system runs and how much text it handles. We estimate it in the quote and set usage limits.
Will it work with our existing software?
If your CRM, ticket system or shop offers an API or an import, usually yes. We check this in the first step, before you get a quote.
Where should we start?
With a process that happens often and whose result is easy to check. That becomes a pilot with a fixed scope. If it works, we build on it.
Which process costs you the most time?
Tell us briefly what happens by hand today. You get an honest view on whether AI, simple automation or neither is worth it.
- A personal reply from Lucas by email
- A free 15-minute call about your process
- A fixed-price quote with estimated model costs
- Work starts after your written approval