Question 1

How does the weekly cancellation actually work?

Accepted Answer

End any week, with one message. No notice period, no exit interview, no fine print. We invoice weekly, so the most you're ever committed to is the current week.

Question 2

What if I'm not blown away by the work?

Accepted Answer

It's in your contract: tell us, and we refund that week. No questions, no invoices to dispute, no calls to escalate. The only rule: refunds apply to the most recent week.

Question 3

Why don't you track hours?

Accepted Answer

Because hours are the wrong metric. If we're optimizing for hours billed, we're not optimizing for your outcome. The deal is simpler: every week, we earn the next one. If we don't, you don't pay. We're free to spend zero hours or sixty. What matters is whether you're blown away.

Question 4

What does 'expectations detached from reality' mean?

Accepted Answer

We work with operators, not lottery winners. If a request would require breaking physics, the law, or a third party's systems, we say so, and if we can't align, we walk. The guarantee is mutual: you can fire us any week; we can also fire ourselves.

Question 5

Are you an AI agency?

Accepted Answer

Yes, and an honest one. We're a senior product team in Austria that builds AI agents and AI products end to end. Unlike pure-play AI agencies that ship one feature and leave, we own the whole build: architecture, evals, billing, observability. We've shipped this for enterprise AI and SaaS clients. And we'll tell you when AI is the wrong tool, even if it shrinks the engagement.

Question 6

What is agentic SaaS, and can you build it?

Accepted Answer

Agentic SaaS is a product where AI agents do the core work, planning and acting across tools, not a chatbot bolted onto a dashboard. Yes, we build it: the agent loop, the tool integrations, and the unglamorous production scaffolding (auth, billing, evals, guardrails, observability) that decides whether it survives real users.

Question 7

Can you automate our workflows with AI?

Accepted Answer

Yes. AI workflow automation is our most common agent build: triage, internal research, ops pipelines, and tasks that run unattended on a schedule. We ground every workflow in retrieval and evals so you can measure when the model is wrong, instead of finding out from a customer. We also tell you which steps are better left to a rules engine.

Question 8

Do you do AI agent development from Austria, and do you work remotely?

Accepted Answer

We're based in Tirol, Austria, and work remote-first with clients across the DACH region and internationally. Time zone overlap is wide, and we ship in your repo and your cloud (AWS, GCP, Azure, or self-hosted), so where we sit rarely matters to the build.

Question 9

Do you fine-tune your own models or only integrate existing APIs?

Accepted Answer

Both, depending on what's right. For 90% of business use cases, well-prompted frontier models (OpenAI, Anthropic, open-weights like Llama) outperform a custom fine-tune at a fraction of the cost. We reach for fine-tuning only when the task is narrow, the data is proprietary, and the cost math works. We'll tell you which case you're in honestly.

Question 10

How do you handle hallucinations and reliability in production?

Accepted Answer

Three layers: structured outputs with JSON schema validation, retrieval-augmented generation grounding the model in your sources, and evaluation harnesses that score real responses against expected behavior on every deploy. We don't ship AI features without a way to measure when they're wrong.

Question 11

Where does our proprietary data live, and do you train on it?

Accepted Answer

Your data sits where you tell it to, and the products we build for you run under your own license with the AI provider, so the data-privacy terms are whatever you've signed. If you have an enterprise contract with OpenAI, Anthropic, Azure, etc., your data is contractually excluded from training. If you're on a default tier, review the provider's terms before you wire production data through it. For sensitive cases, we deploy open-weights models in your own cloud (AWS Bedrock, GCP Vertex, self-hosted) so the question never arises. We never use your data to train anything for anyone else.

Question 12

Typical timeline from idea to production AI feature?

Accepted Answer

Prototype: a week. Production-ready with evals, guardrails, and observability: 4–8 weeks. The slow part isn't the AI, it's everything around it: auth, billing, rate limiting, content moderation, audit logs. We've shipped enough to know where the time actually goes.

Question 13

Which AI frameworks and libraries do you use in production?

Accepted Answer

Depends on the build. For RAG and agents: LangChain, LangGraph, LlamaIndex, and the Vercel AI SDK on the frontend. For self-hosted inference: vLLM, Ollama, llama.cpp, Hugging Face Transformers. For evaluation: Braintrust, Phoenix, OpenAI evals. For observability: LangSmith, Helicone, Langfuse. We pick boring, proven tools over the hype cycle, the AI stack shifts every six weeks so we choose what we can rip out cleanly.

Question 14

What does building an AI feature actually cost?

Accepted Answer

Prompt-engineering integration on top of an existing app: €5-15k. A RAG system over your own docs with evals and a real UI: €15-40k. A multi-step agent with tools, memory, and guardrails: €40-100k+. Runtime API costs are separate and depend on model and token volume. We budget the API spend into the proposal so you don't get a six-figure surprise from OpenAI in month two.

Question 15

Will we get locked into one model or vendor?

Accepted Answer

Not if we build it right. We keep the business logic separate from the model behind a routing layer, so swapping GPT for Claude, Gemini, or an open-weight model like Llama is a config change, not a rewrite. We ship in your repo and your cloud, and for sensitive or cost-sensitive workloads we run open-weight models you host yourself. You own the code and the infrastructure. The lock-in risk is real, and we architect against it from day one.

Question 16

When should we NOT add AI to our product?

Accepted Answer

When the same job is doable with a SQL query, a rules engine, or a form. When you need sub-200ms latency. When 100% deterministic output is non-negotiable (legal contracts, financial postings, medical dosing). When there's no feedback loop to catch when the model is wrong. We'll tell you 'don't do this' if the use case doesn't justify it, even if it shrinks the engagement.

Question 17

Can you build AI agents that take real actions, or just chatbots?

Accepted Answer

Real agents. We build AI agents that call tools (function calling, MCP), execute multi-step plans (LangGraph state machines), read and write to your databases and APIs, and run unattended on schedules. Shipped examples: invoice-triage bots, internal-research agents, content-ops pipelines, automated QA harnesses. Chatbots are the boring case. Agents that move work forward are where the leverage lives.

Question 18

We want to use AI internally, not ship an AI product. Can you help?

Accepted Answer

That's a different service: AI Enablement. This page is about building AI products for your customers. If your goal is to take work off your own team (automating internal processes, running workshops, setting up tooling on your own infrastructure), start there instead.

AI Agents & AI Products That Survive Production

Why most AI projects never ship

What We Do

From use case to production

Discovery

Architecture

Grounding

Evals

Guardrails & observability

Ship & hand over

The production stack we build on

What it costs

What principles guide our AI work?

Proof, not promises

What clients say

FAQs

More in AI & Frontier Tech

Get to know us