Ask five agencies what an AI automation project costs and you get five versions of "it depends". That is partly honest, because scope really does drive price, and partly convenient, because a market without published prices favours the seller. So here are the actual numbers: survey data, published price lists and consistent ranges from across Germany, Austria and Switzerland, as of July 2026.
What providers charge per day
The baseline is well documented. The Freelancer-Kompass survey (over 3,200 freelancers in the region) puts the average IT freelancer at roughly 95 to 104 euros per hour, with data and AI specialists at the upper end. Malt's rate barometer shows data experts averaging around 750 euros per day and strategy consultants around 1,000.
Agencies price above that. Published comparisons cluster specialized AI agencies at 1,200 to 2,500 euros per day, mid-size consultancies at 2,000 to 3,500, and the big consulting brands at 2,500 to 5,000. Switzerland runs roughly a third higher in francs.
The useful takeaway is not the day rate itself. It is that day rates are the wrong way to buy this work. A project billed by the day has no natural end. What you want is a fixed price for a defined result, and the ranges below are what those results trade for.
What projects cost, by type
A single automated workflow (one process, one or two systems connected, for example inquiry intake into your CRM with a drafted reply): published ranges sit at 1,500 to 5,000 euros. This is the honest entry point, not the four-figure "AI transformation" some ads promise.
A multi-step automation across two or three systems, with an approval step and error handling: 4,000 to 12,000 euros in most published price lists, up to 25,000 when interfaces get difficult.
A knowledge assistant on your own documents (the thing most SMEs actually want): tiered fixed prices exist here. One published list charges 5,600 to 11,200 euros up to 500 documents and 11,200 to 22,400 euros up to 5,000 documents. Broader market ranges for a solid mid-market build run 15,000 to 40,000 euros. Enterprise versions with pilots and rollout programs reach six figures, which an SME does not need.
Document and invoice extraction: no reliable standalone market rate exists, because most vendors quote case by case. Priced as a mid-complexity automation it lands in the 4,000 to 12,000 euro band. The economics are easy to check on your own numbers: manual processing costs several euros per document in labor, automated processing costs cents.
A phone or voice agent: the software itself has become cheap, roughly 10 to 30 cents per minute across platforms, so a small business handling 300 minutes a month pays under 50 euros in usage. The project cost is in connecting it to your calendar, your CRM and your escalation rules, which prices like a medium automation.
What the first step costs
Discovery has real market prices too: process audits and AI readiness assessments are published at 1,500 to 7,000 euros in Germany and Austria, and 2,500 to 15,000 francs in Switzerland. Germany's BAFA program subsidizes qualifying consulting at 50 to 80 percent, which is why several providers price exactly at the funding ceiling of 3,500 euros.
Two things matter more than the audit price. First, whether it produces a decision you can act on, meaning a shortlist with expected effort and payback per use case, not a slide deck about AI trends. Second, whether the fee is credited if you proceed to implementation. If it is not, you are paying twice for the same thinking.
The costs almost nobody budgets
This is where most published comparisons go quiet, and where projects go over budget:
- Model usage. A typical SME assistant handling 1,000 to 5,000 requests a month costs roughly 50 to 200 euros per month in AI model fees. A document-heavy knowledge assistant with hosting can reach 250 to 1,400 euros per month all-in.
- Maintenance. Published rules of thumb put upkeep at 10 to 25 percent of the build cost per year. Systems drift: your products change, your documents change, the models change.
- Data preparation. For knowledge assistants, cleaning and structuring your documents routinely eats 30 to 50 percent of the total budget. If a quote for a document assistant contains no line for this, the quote is incomplete.
- Channel fees. Small but real: WhatsApp business conversations, telephony minutes, integration platform subscriptions of 10 to 50 euros per month.
A fair mental model: whatever the build costs, plan 15 to 20 percent of that per year to keep it good.
The four things that actually drive your price
- Number of systems touched. Each additional integration (ERP, CRM, mailbox, calendar) adds interface work. One system is cheap. Four is not.
- Data quality. Clean, structured documents make a knowledge assistant a small project. A shared drive of scanned PDFs from 2011 makes it a large one.
- Consequence of errors. A system that drafts replies for a human to approve is much cheaper to build responsibly than one that acts on its own. For most SMEs the draft-plus-approval design is also simply the right one.
- Hosting constraints. EU cloud is the default and adds little. Strict on-premise requirements add real cost and are worth it only when your data genuinely demands it.
How to pay less
Buy fixed scope, not day rates, so overruns are the provider's problem instead of yours. Start with one process instead of a platform, since the 1,500 to 5,000 euro single workflow teaches you more about your organization's readiness than any strategy engagement. And check funding before you sign anything: several programs in Austria and Germany cover 30 to 60 percent of exactly this kind of project, and we keep a verified list of what is actually still open in our funding overview.
For transparency, since this whole article argues for published prices: our own audit starts at 2,500 euros and is credited in full against the implementation, implementations are fixed-fee and named in the proposal, and ongoing care is monthly and cancellable anytime. Those numbers sit deliberately at the lower, fixed-price end of the ranges above, and now you know exactly where they sit in the market.