Buyer's guide · French-speaking Switzerland
How do you choose the right AI agency in French-speaking Switzerland?
On 4 October 2026 we again counted the providers that show up when you search for an AI agency in French-speaking Switzerland, query by query, for our own market intelligence. There are 25 of them, up from about ten in June. Each claims to be the right choice. This guide sorts them into families and gives the prices they publish, along with the questions that set them apart.
In brief
Choosing an AI provider, in four sentences.
Three kinds of provider sell AI in French-speaking Switzerland: generalist AI agencies, ERP/CRM integrators and freelancers. Public prices run from CHF 490 (a first audit) to CHF 12,800 (supervised agent). The choice depends on whether the AI acts inside your tools or beside them. Djtal publishes its rates and runs over 25 agents in production, starting with its own.
The market
Three kinds of provider, three different trades.
The term ‘AI agency’ covers businesses that have little in common. A provider's background decides what it will do well and what it will underestimate.
| Type | What it does well | Its structural limit | Choose it when |
|---|---|---|---|
| The generalist AI agency | Builds chatbots, content and one-off automations; starts fast and communicates well | Works alongside your management tools, so your ERP or CRM stays a black box to it | You have a marketing need, a brochure website or a first chatbot with no access to sensitive data |
| The ERP/CRM integrator | Knows your data, your access rights and your processes from the inside, and is accountable for keeping the system consistent | Its AI often stops at the vendor's native features, with no custom agent | The project is anchored in the management tool: invoicing, customer records, stock |
| The freelancer | Offers a low entry cost, agility and a direct relationship | Continuity suffers (one person means holidays, illness and busy periods), and so does breadth of skills | The scope is narrow and non-critical, and the budget is tight |
Since summer 2026 the boundaries have been shifting. Several generalist agencies now advertise agents connected to your tools, and integrators are adding their vendor's AI. The question to ask stays the same: who on the team has already written data to a live ERP?
Prices
What the market publishes in French-speaking Switzerland, and what we charge.
These ranges were recorded on 4 October 2026 from prices published by providers in French-speaking Switzerland. Only a minority of providers show their prices, so we quote what can be checked. All amounts exclude VAT.
| Service | Published market prices | Djtal's published rates |
|---|---|---|
| AI audit and scoping | from CHF 490 (short assessment) · free as a 30–60 minute pre-sales session | CHF 800 (4 hours) · CHF 4,800 (24 hours, three costed use cases, 90-day plan) · CHF 8,000 (40 hours) |
| AI agent | CHF 2,500–4,900 depending on channels and lead time (as of 4 October 2026) | CHF 12,800 per agent in production (scoping, guardrails, set-up) · supervision from CHF 190/month |
| Website | subscriptions from about CHF 50/month (platforms) to CHF 20,000 or more (traditional agency) | CHF 4,900, CHF 8,900 or from CHF 14,900 depending on scope, GEO (visibility in AI answers) included · maintenance from CHF 190/month |
| Training | from the discovery workshop to a certified programme | half-day in-house session CHF 1,490 · training needs analysis CHF 600 |
An agent at CHF 2,500 and an agent at CHF 12,800 share a name and cover two different things. The first answers on one channel. The second acts inside your tools, with written guardrails, an action log and monthly supervision. Compare scope before you compare amounts: our guide to the AI audit walks you through this for the first step.
Before you sign
The seven questions that set providers apart.
- 1.
“Which agent is running in production, for whom, and since when?”
Ask for dates and names. A demonstration can be put together in an afternoon; six months in production cannot.
- 2.
“Are your prices public?”
A provider that publishes its rates accepts being compared. One that saves them for the quote keeps control of how your budget is being set.
- 3.
“Will the AI act inside our tools, or beside them?”
An assistant that drafts can be corrected; an agent that writes to your ERP commits your data. The second needs knowledge of the tool, scoped access rights and an action log.
- 4.
“Who supervises after go-live, and at what rate?”
An agent needs looking after, because models evolve, new cases appear and drift has to be corrected. If the contract has no ‘supervision’ line, you will find out what it costs at the first incident.
- 5.
“How will the agent keep up with our changes?”
This is about ongoing maintenance. AI models change several times a year, the law changes (the Swiss Federal Act on Data Protection, the EU AI Act) and your company grows. Ask what the contract covers after commissioning: adapting to new models, keeping compliance up to date and extending the scope as your business develops.
- 6.
“How dependent are we on the AI provider?”
Anthropic, OpenAI, Google, Zoho or any other: your agent relies on a third-party model. Ask what happens if that model changes, goes down or disappears. Are your instructions and know-how kept in an open, reusable format? Does your data stay in your own tools? What would migrating to another provider cost? A well-built agent outlives its model.
- 7.
“And what does AI do in your own company?”
Ask to see their own processes running. Within a minute, the answer separates those who operate from those who recommend.
Where we stand
Djtal, the integrator that became an agent operator.
We come from the second family, with around fifty ERP and CRM deployments since 2017, and we began specialising in AI two years ago. It is now our core business: designing and running supervised AI agents that act inside the management tools you already use, whichever vendor makes them. We have more than 25 agents in production to date. Djtal has been a member of Anthropic's Claude Partner Network since 2026.
Our answers to the seven questions above take one line each. Nine in-house AI agents have run our sales, accounting and communications since May 2026, with versioned logs, and on 22 September 2026 our client Econox published the first verified review of our work, 5.0/5 on Clutch. Our rates are published on this site, and our agents write to our own CRM and accounting system. Supervision starts at CHF 190 a month. As for ongoing maintenance, on 2 September 2026 our agents switched AI model overnight, in production. As for dependence on a provider, our agents live in open text files (Markdown, a format every AI provider reads) version-controlled by their owner, and your data stays in your management tools, so if a provider fails, everything can be reused with another model. We'll show you the rest in a meeting, on screen.
Your starting point decides the first step. The AI strategy audit sets priorities, in-house AI training equips your teams, website design gets you found, and AI agents let you delegate whole processes.
Frequently asked questions
Choosing an AI agency: your questions.
How much does an AI agency cost in French-speaking Switzerland?
Public prices recorded on 4 October 2026 in French-speaking Switzerland: a first audit from CHF 490, a simple AI agent from CHF 2,500, visibility subscriptions at around CHF 600 a month. At Djtal, an audit costs from CHF 800 (4 hours) to CHF 8,000 (40 hours, across several departments), an AI agent in production costs CHF 12,800 and supervision starts at CHF 190 a month. A gap in price almost always reflects a gap in scope, so ask what the amount covers, hour by hour.
What is the difference between an AI agency and an ERP/CRM integrator?
An AI agency starts from the use case (a chatbot, content, a one-off automation) and works alongside your tools. An integrator starts from your management system (CRM, ERP, accounting) and knows it from the inside: access rights, data, processes. For an AI agent that has to act in your tools (raise an invoice, update a customer record), knowing the tool matters more than creativity, because a faulty entry in an ERP costs more than an average piece of text.
Should we choose a local AI agency or work with one remotely?
The useful criterion is accountability, far more than the postcode. A Swiss company answers to Swiss law (the Swiss Federal Act on Data Protection, FADP) and travels when the project calls for it: scoping workshops, team training, go-live support. Day-to-day delivery works very well remotely. Be wary above all of providers with no Swiss legal entity when your customer data is involved.
How do you check that an AI agency really delivers?
Ask for something dated and verifiable: which agent runs in production, for whom, since when, with what measured result? A verified customer review on a third-party platform (Clutch, Google) is worth more than a page of logos. Then ask what the provider does with AI in its own company, because a provider that sells you agents without running any is selling you a theory.
Get started
Put the seven questions to us. We'll answer them on screen.
A 30-minute call: you bring your context, we show you our agents in production and tell you, with price ranges to hand, where to start. No obligation.
Last updated: