What does AI consulting for SMEs in Zurich cost?

There is no list price for AI consulting. Three billing models are common: a project at a fixed price, a monthly retainer for ongoing support, or billing by effort. The initial, no-obligation conversation is free with reputable providers, and the actual proposal follows only once it is clear where your business stands today.

Before getting to the pricing models, it is worth asking the starting question: what exactly should AI take over in your business? Typical first use cases for an SME are sorting and pre-qualifying emails, generating quotes and invoices from templates, a searchable knowledge base for customer service, or summarising meetings and long documents. The more clearly that use case is defined, the more precisely the effort can be estimated.

The three common pricing models

Which model fits depends less on the size of your SME than on how clearly the use case is already defined.

Project at a fixed price

The scope is fixed in advance, for example automating a single workflow or a pilot project that makes the value visible before you invest further. This suits a first step well, because you know beforehand what you get and what it costs. The drawback: anything that comes up during implementation has to be renegotiated.

Monthly retainer

You book a fixed allocation for ongoing support, for instance when several use cases are built one after another or a team regularly needs help choosing and implementing tools. This works well once the first use case is running and more are planned. Watch the notice period and whether unused hours expire.

Billing by effort

You pay for the time actually worked. This fits when the scope is genuinely open at the start, for example during a first process analysis that has not yet produced a clear roadmap. For the very first step it is rarely the cheapest model, because nobody knows in advance where the invoice will land. It tends to make more sense once a rough direction is already set.

What an initial conversation normally costs

Usually nothing. A no-obligation initial conversation exists to understand your starting point: which processes cost time today, what data already exists and where a first use case would realistically pay off. Only after that comes a proposal with a concrete suggestion. The initial conversation for AI consulting in Zurich is built exactly this way: no obligation, no sales pressure, with an early read on where the biggest levers sit.

Be sceptical if a finished package with a price is already on the table during the first call, before anyone has seen your processes. That is a sales pitch, not consulting. An hourly rate with no sense of how many hours a realistic first step actually needs is just as uninformative.

What else influences the price

Beyond the chosen model, four factors noticeably change the effort for an AI project.

• Integration depth. A chatbot that runs standalone on a website is quicker to build than a solution that has to plug into your existing CRM or inventory system. Every additional interface means more coordination and more testing.
• Data volume and quality. Clean, centrally maintained data can be put to use faster than information scattered across several spreadsheets, mailboxes and paper files.
• Languages and locations. Every additional language and every additional location multiplies the work on content, testing and fine-tuning.
• Regulatory requirements. In regulated industries, such as financial services, governance and traceability requirements add effort, but they also save trouble later.

How to recognise a credible proposal

A usable proposal comes after a short analysis, not before it. It names the concrete use case, who carries out the work and how you will measure success within weeks or months. It separates one-off effort, such as setup, from ongoing costs, such as tool licences or support.

• Data quality and system landscape. A proposal that never asks which systems you use today or how clean your data is has missed the biggest cost driver.
• No packaged deal without analysis. Two businesses with the same goal often need very different amounts of groundwork, depending on how many systems are involved.
• Clear separation of consulting and licence costs. Tool subscriptions often keep running after the consulting project ends. A credible proposal states that separately.
• A pilot before the big investment. Anyone who sells a company-wide strategy immediately, instead of proving value with one use case first, is selling themselves rather than your result.

In my experience, the biggest cost driver is rarely the technology itself, but how much groundwork a business has already done. An SME with clean, centrally maintained data often reaches a working use case faster and cheaper than a larger business with sprawling, disconnected systems.

How to frame the budget

The more useful question is not what consulting costs per month, but how much time a use case actually ties up today and what that time is worth. If an employee spends several hours a week manually sorting enquiries, you can work out from that how much effort would justify automating it. That calculation is more honest than any industry benchmark, because it starts from your own situation rather than someone else's number.

Hold that calculation next to the proposal before you commit to a model. A fixed-price project that pays for itself in saved time within a few months is a different decision from a retainer whose value only shows after a year. Either can be right, but only if you have worked it out beforehand, not once the first invoice arrives.

An overview of sensible first use cases is in the article on AI in SMEs, and a structured approach in AI strategy in 5 steps. If, alongside using AI internally, you also want to know whether your own business is found in ChatGPT, Perplexity and other AI search engines, the AI visibility check shows where you stand today.