How to Choose an AI Automation Agency: 8 Questions to Ask

The market for AI automation agencies is growing fast. According to Gartner, by 2026 more than 30% of SMBs will invest in some form of workflow automation. That means: more providers, more promises, and more business owners who discover after the fact that they chose the wrong agency.

We are an AI automation agency ourselves. So yes, this article is not objective. But it is honest. We will tell you exactly which questions to ask — including to us — and which answers should make you suspicious.

Red flags: walk away if you hear this

Before we get to the questions, the warning signs first. If an agency does any of these, it is a reason to keep looking:

• No case studies. "We have a lot of experience" without proof is worthless. Ask for names, numbers, references. Any serious agency has at least two clients they can name.

• Vague pricing. "It depends" without any direction is a red flag. Any agency that knows what it does can give you a range. If they cannot, they either do not know or do not dare.

• No GDPR awareness. If you ask where your data is hosted and the answer is "in the cloud" without specification, end the conversation. Your data belongs in the EU, on servers you can point to.

• "10x growth" promises. Automation saves time and reduces errors. It does not multiply your revenue by ten. If an agency promises that, it is selling air.

• "We do everything." An agency that does everything does nothing well. Look for specialisation. In your sector, in specific tools, in a type of problem.

Green flags: this is what you want to see

• Transparent pricing with packages or ranges. You know upfront what you are paying.
• Sector experience. They know the processes in your industry. They do not ask "how does that work for you?" but say "with other accountants we see this problem."
• Measurable results. Not "it works better" but "client X saves 12 hours per week on invoicing."
• European hosting. Data in the EU, explicitly stated.
• Clear scope. What is included, what is not, and what costs extra.

The 8 questions you should ask

1. What is your pricing model?

Do you pay per hour? Per project? Setup + monthly? Are there costs per workflow, per user, or per action?

Why this matters: Hourly billing gives an agency no incentive to work efficiently. Setup + monthly is the fairest model: you pay for the build, and a fixed amount for maintenance, monitoring and optimisation.

What you want to hear: A clear overview with packages and a scope clause that covers what happens with additional work. At Mindsora, we work with a setup fee plus a fixed monthly price. Any scope overruns are always discussed upfront.

2. Where is my data hosted?

This is not a technical detail. This is a GDPR obligation.

What you want to hear: Specific answers. "On a Hetzner VPS in Frankfurt" or "On AWS eu-west-1 in Ireland." Not "in the cloud" and not "with our partner."

Bonus question: "Do you have a data processing agreement ready?" If the answer is no, walk away. It is legally required as soon as a third party processes data on behalf of your business.

3. What is your response time when things break?

Automation runs 24/7. If a workflow fails on Friday evening and you have sent 200 duplicate emails by Monday morning, that is a problem.

What you want to hear: Concrete SLAs. "Critical issues within 4 hours, other issues within 24 hours" is clear. "We do our best" is not.

Bonus question: "Do you have monitoring that automatically alerts on failures?" An agency that only finds out about problems when you call has no monitoring.

4. What happens if I cancel?

This is the question nobody asks and everyone should have.

What you want to hear: "Your workflows are yours. We provide documentation, export your data, and help with the transition." If the answer is that you are locked into their platform or cannot take your data with you — red flag.

Also check: Is there a notice period? A minimum contract duration? Switching costs? Everything that binds you to an agency should be transparent.

5. Can you provide references?

Not a testimonial on the website. A real reference. Someone you can call.

What you want to hear: "Yes, here are two clients you can speak with." An agency that refuses to provide references either has something to hide — or simply no satisfied clients.

Note: Ask for references from your sector. An agency that does brilliant work for e-commerce can be entirely unsuitable for an accounting firm.

6. Who actually does the work?

At many agencies, you talk to a salesperson, but a junior does the work. Or worse: it is outsourced to a freelancer you never meet.

What you want to hear: "This is the team that will work on your project, and this is their experience." Bonus points if you meet the person building your workflows, not just the account manager.

7. What technology do you use?

This sounds like a technical question, but it is a business question. The tools an agency chooses determine your costs, your flexibility and your dependency.

What you want to hear: A clear answer with reasoning. "We use N8N because it is self-hosted and we run unlimited workflows" is better than "we use the best tools for the job."

Also check: Does the agency use tools you can also use without them? Or are you locked into proprietary software that ties you to their contract?

8. How do you handle GDPR and compliance?

Not "we take privacy seriously." Be specific.

What you want to hear:

• "We have a standard data processing agreement."
• "Data is hosted in the EU at [specific location]."
• "We do not log personal data in our automation logs."
• "We work with data minimisation: only what is necessary."

If the answer is vague, so is their compliance awareness. And that is your risk, not theirs. The fines are for the data controller — and that is you.

The scorecard

Use this table as a checklist during your conversations:

Question	Good answer	Bad answer
Pricing model	Clear package + scope clause	"It depends"
Data hosting	EU, specific location	"In the cloud"
Response time	Concrete SLA (e.g. 4 hours)	"We do our best"
On cancellation	Data is yours + transition help	Vendor lock-in
References	Yes, in your sector	"We have testimonials"
Who does the work	Transparent team	Vague or outsourced
Technology	Specific + motivated	"The best tools"
GDPR/compliance	DPA + EU hosting	"We take privacy seriously"

Score: Each good answer = 1 point. Below 6? Keep looking.

Why we are writing this

You might think: this is a sales pitch disguised as advice. Partly true — we hope you think of us when you are looking for an agency. But we also hope you ask these questions to every agency you speak with. Including us.

The SMB market for AI automation is full of agencies that give beautiful demos but disappear after the signature. That have no monitoring. That put your data on American servers without a data processing agreement. That build your workflows in tools you are locked into.

Asking good questions protects you against that. Not our article — your critical attitude.

Mindsora's answers to these 8 questions

Full transparency:

1. Pricing: Setup + monthly. Three packages, from EUR 4,500 + EUR 650/month to custom. Scope clause on every proposal.
2. Data hosting: Hetzner VPS, Frankfurt. Data does not leave the EU.
3. Response time: Critical issues within 4 hours. Other issues within 24 hours on business days.
4. On cancellation: Your data and documentation are yours. No lock-in, no switching costs.
5. References: Yes. Ask us and we will connect you with existing clients.
6. Who does the work: Wouter. Personally. No intermediary layers.
7. Technology: N8N (self-hosted), GoHighLevel, Supabase. All tools you can also use without us.
8. GDPR: Data processing agreement with every client. Data in EU. No PII in logs.

Want to experience how we work? Schedule a free AI Readiness Scan. We will look at your processes together and honestly tell you whether automation is profitable for you — and if so, where to start. No obligation, no sales pitch.