AI automation is the use of AI agents and workflow tools to replace or accelerate manual business processes.
And AI is an AI implementation agency based in Groningen, the Netherlands, that has helped more than 30 SME organisations with concrete automations. Our AI ROI Scan comes with a €20,000+ ROI guarantee: if the implementation does not deliver €20,000 within 12 months, you get the Scan fee back.
AI automation combines AI models (such as Claude or GPT) with workflow tools (such as n8n or Make) to automate processes that classical rule-based systems cannot handle. The difference with traditional automation: AI can deal with unstructured input. Free-form emails, recorded conversations, PDF documents, photos of receipts. All data that rule-based software gets stuck on.
In practice this means: an AI agent reads incoming customer emails, determines the topic, summarises the content, finds the right answer in your knowledge base, and drafts a reply. A human reviews and sends. What used to take 8 minutes now takes 30 seconds. The same principle applies to quotes, contract analysis, lead qualification, meeting minutes, social posts and hundreds of other repetitive processes.
According to Statistics Netherlands (CBS), 27 percent of Dutch companies with 10 to 50 employees used AI in 2025, up from 11 percent in 2023. Companies that use AI most often apply it to marketing or sales (35 percent) and to business administration or management tasks such as planning and translation (32 percent). That is also where most of our programmes start: customer communication, content and meeting notes.
For SMEs this is attractive because the investment is low (from €5,000), lead time is short (2 to 4 weeks for a first use case), and the ROI is directly measurable in hours or euros. Unlike large enterprise AI projects, SME automation is pragmatic, modular and short-cycle.
Our order from Tuesday has not arrived yet. Can you see when it will come?
Your order is scheduled for Thursday. Tomorrow you will receive the tracking link.
| Rule-based automation | AI automation | |
|---|---|---|
| Input | Fixed fields and formats, such as a web form or an export from your accounting software | Also free text, PDFs, recordings and photos, such as customer emails and receipts |
| How it decides | By rules you set in advance | A language model reads the content; rules and checks around it guard the outcome |
| Where it stops | At any deviation from the expected format | At unique, one-off situations without a recognisable pattern |
| Review | The outcome is predictable | A human checks where a mistake is costly |
| Example tools | Zapier or Make without an AI step | Claude or GPT, connected via n8n or Make |
In four steps from process analysis to a working automation in production. We start small and proven, only scale up once the first use case delivers value. No big-bang projects.
We analyse your processes, identify the 3 highest-scoring automation candidates by time saved x frequency x feasibility, and deliver a concrete roadmap with an hours-saved forecast per use case.
We build the chosen automation with our standard stack: Claude for reasoning, n8n for orchestration, Python/TypeScript for custom logic, MCP servers for tool access. Iterative, with interim reviews with your team.
Test with real production data, fine-tuning on edge cases, training your team, documentation of what the agent does and does not do (governance), and handover session.
The automation runs in your production environment with logging and quality monitoring. Monthly check-in to monitor performance and expand to next use cases.
The five SME categories with the highest ROI on AI automation. For each category we have a specialised solution. The order below is based on what we most often see as a first use case with SME clients.
Three cases from our practice with hard numbers. Different sectors, similar pattern: first use case live within 2 to 4 weeks, measurable hours saved or revenue impact within 8 weeks.
Sector: gym chain, ~12 locations.
What we built: AI-driven lead capture with WhatsApp automation and email follow-ups that personalise themselves based on conversation flow.
Result: +20 hours per week saved on manual lead follow-up, +40% leads in pipeline, +15% conversion from lead to trial class.
Sector: real estate valuation and advisory.
What we built: AI agent that produces a first valuation in 5 minutes based on cadastre data, comparable objects and market context. Output goes to the appraiser for final check.
Result: 5 minutes per valuation instead of 45 to 60 minutes. Appraisers do 3x more valuations per day, clients get faster indications.
Sector: care and social support, a consortium of five care organizations.
What we are building: an AI assistant in chat that clarifies the request for help, prepares the registration and refers people to the organization that fits. Which situation leads to which organization is set out in a decision tree that the care organizations manage themselves. At signs of acute distress, Noor shows the safe route straight away.
Result: one digital front door for five organizations, in 9 languages. Staff also use Noor to see how things work at the other organizations in the chain.
Our projects range from €5,000 for a targeted AI ROI Scan to €40,000 for a multi-agent system. The exact rate depends on three variables: number of integrations with existing systems (CRM, ERP, mailbox), complexity of the decision logic (simple text generation versus multi-step reasoning), and the agreed maintenance frequency.
Our guarantee: if the implementation does not deliver €20,000 in demonstrable returns within 12 months (in hours saved or extra revenue), you get the Scan fee back. We can do this because we only take on projects where we are confident upfront that the ROI is achievable.
Honest scoping: four situations in which we advise against or decline projects. AI is no silver bullet and not every process lends itself to automation.
If the process occurs less than about 3 times per week, the build and maintenance cost more than the savings. Manual is cheaper.
AI works best on processes with recognisable patterns. Fully unique, one-off situations require human judgement, not automation.
For processes where errors directly cause high costs or safety risks AND a human review step is not possible, we decline the project.
For highly sensitive data without room for cloud LLMs, we first recommend an on-premise or EU-only stack, or we advise against the project until that option exists.
AI automation is the use of AI models (such as Claude or GPT) plus workflow tools (such as n8n or Make) to replace or accelerate manual business processes. It differs from traditional automation because AI can handle unstructured input such as emails, documents and conversations, and can make decisions independently within predefined boundaries.
The most valuable starting points are: customer communication (FAQ, email, WhatsApp), content production (articles, social, newsletters), knowledge management (internal search, onboarding), meeting administration (minutes, action items) and marketing/sales flows (lead capture, follow-up). And AI has specific solutions for each of these.
A first working automation is usually live within 2 to 4 weeks. More complex multi-agent systems typically take 6 to 8 weeks. We always start with a 1-week AI ROI Scan to determine which process has the highest ROI before we build.
Our projects range from €5,000 for a targeted AI ROI Scan to €40,000 for a multi-agent system. The Scan comes with a €20,000+ ROI guarantee: if the implementation does not deliver €20,000 within 12 months, you get your money back. Main cost drivers are number of integrations, complexity of decision logic, and maintenance.
AI automation works poorly when the process occurs less than 3 times per week (manual is then cheaper), when input is fully unstructured or unpredictable, when there is no room for a human review step at high error costs, or when data is too sensitive for third-party LLM providers without an on-premise option. At And AI we decline projects where AI adds no value.
Our standard stack: Claude (Anthropic) for reasoning, n8n for workflow orchestration, Python and TypeScript for custom integrations, MCP servers for tool access, and Supabase as the data layer. We also work with OpenAI, Make, Zapier and local LLM stacks when the client context requires it.
Updated on
Schedule a free introduction call or request an AI ROI Scan. Together we determine which processes have the highest ROI for you.
Want to know which processes in your organisation are best suited for AI automation? Fill in the form below for a first piece of advice.
No sales pitch, just an introduction. Anthonie is happy to think along about where your organisation stands and what the logical next step is, no strings attached.
Schedule a meeting with Anthonie