Automation & AI for SMEs · Hamburg Automating workflows, using AI where it helps.
I analyse workflows and introduce automation and AI assistants with governance and data protection, in Hamburg and remotely. The starting point is a target picture, not a tool.
Where time and knowledge get lost
A lot of work consists of moving data between systems by hand, and company knowledge sits in mailboxes, folders and people’s heads. At the same time the pressure to use AI grows, while it remains open which use case actually delivers value and which data may be used for it.
What should be running at the end
Automations that run reliably and can be followed, and AI applications with a clear purpose, defined permissions and data protection settled. The basis is a target picture worked out together: what should get better, which data may go where, what stays in-house.
What I take on for automation and AI.
The order is deliberate: need and target picture first, then automation, then AI.
-
Process analysis and target picture
Which processes cost time, where is knowledge stuck, which use cases are worth it? Prioritised by benefit and effort.
-
Workflow automation
Approvals, notifications, data transfers and recurring tasks with Power Automate or n8n, documented and maintainable.
-
System integration
Connecting line-of-business applications, Microsoft 365, databases and cloud services through APIs and scripts, so nobody has to copy data across by hand any more.
-
Copilot rollout
Preparing and rolling out Microsoft 365 Copilot: permissions, data quality, governance, a pilot group and training.
-
AI assistants and knowledge bases
Preparing documentation, contracts and manuals so AI assistants can answer from them in a way you can follow, with sources cited.
-
Data protection and on-premises models
Clarifying which data may go into which models and, where data sovereignty requirements are higher, running models on your own infrastructure.
From practice
One example: when a new employee is created in Personio, her accounts in the on-premises Active Directory and in Microsoft 365 are set up automatically, and hardware and licences are assigned to her. Workflows like this run for clients on Power Automate or on n8n, which I host myself.
I have introduced Copilot for clients, each time after permissions and SharePoint had been cleaned up. AI assistants built on the company’s own documents and local models are also in production use at client sites.
My stance
Most companies already use AI, often without shared rules. So I rarely advise against it. My advice is rather to engage more thoroughly with the tools and use them more deliberately: with a clear purpose, clarified data and staff who know what they are doing.
Tools, when they fit the target picture.
The choice comes only once the goal and the flow of data are clear.
- Power Automate
- Workflow automation within Microsoft 365.
- n8n
- Open-source workflow automation, self-hosted, connecting line-of-business applications, Microsoft 365 and AI models.
- Microsoft Graph & APIs
- Programmatic access to Microsoft 365 and other systems.
- Microsoft 365 Copilot
- An AI assistant in Word, Outlook, Teams and Excel, working within existing permissions.
- Cloud and on-premises models
- Models from OpenAI, Anthropic or Microsoft, or open-source models on your own hardware.
From use case to live operation.
-
01
Analysis
Observing how work actually runs, quantifying the effort, locating the data and permissions involved.
-
02
Target picture
Agreeing one or two use cases with a clear benefit and a data protection assessment.
-
03
Delivery
Prototype, test with real data, pilot group, then production.
-
04
Further development
Build on what works. Name what does not.
Questions about automation and AI.
Your question isn't here? Write to me directly.
None, until the target picture is clear. Only once it is settled which problem is to be solved and which data is involved can you judge which tools help. Sometimes it is Copilot, sometimes an on-premises model, sometimes no AI at all.
The first question: which data may go to which provider, and which has to stay in-house? That determines whether cloud models, Copilot or on-premises models are the right fit, and what rules apply to your staff. For Copilot, tidy permissions in Microsoft 365 are the most important prerequisite.
That is the goal. Automations are documented and built so your team can understand and adjust them.
Want to know where AI would genuinely help in your company?
Send me a couple of lines about it, or just call. You reach me directly, no call centre.