Smart Processes
AI that works – not just chats.
A ChatGPT account is not AI adoption. Adoption means agents and automations working inside your systems – CRM, ERP, email – measurably giving hours back. First the audit, then the build, then the measurement.
The five stages of AI adoption.
Tools
A few licenses, no rules. Everyone prompts on their own, nobody measures anything.
Islands
A few people use AI really well – the knowledge stays with them and leaves with them.
Processes
AI sits inside defined workflows: quotes, reports, support triage. First measurements.
Systems
Agents work inside CRM and ERP, grounded in your knowledge (RAG), escalating to humans.
Organisation
Measurement, governance, training. Adoption is routine, no longer a project.
Most companies are at stage 1 – and call it a strategy. The audit tells you where you stand and which step pays off first.
AI audit
Two weeks: process map, time sinks priced by effort and payoff, roadmap with a clear order. Afterwards you know what pays off first – even if you build it without us.
Workflow automation
Copy-pasting data from system A into system B? That was yesterday. We connect your tools end-to-end and build reporting that updates itself – quotes, onboarding, follow-ups.
AI agents & RAG
Agents that work inside your systems: pre-qualifying inquiries, drafting quotes, finding knowledge – grounded in your data, escalating to humans where it matters.
Enablement & governance
So adoption stays when we leave: training, prompt guidelines, data-protection rules and a measurement frame that shows whether it works.
AI adoption workshop
The low-barrier entry: in a workshop we find where AI concretely saves you hours – and what pays off to build first. You leave with a prioritised list, not buzzwords.
Measure before. Build. Measure after.
Baseline
How long does the process take today, how often does it run, what does it cost? Without a baseline there's no success story – just gut feeling.
Pilot
One process, one team, real data. Small enough to learn, big enough to prove.
Rollout
What the pilot proves gets rolled out – with training and guidelines so everyone uses it.
Re-measurement
The same metrics, three months later. 6 h → 20 min per quote is one of those numbers – measured, not claimed.
sevDesk agent for bookkeeping
An AI agent that takes over recurring tasks in sevDesk bookkeeping – wired into the existing toolchain instead of manual clicking.
Reutter Miniaturen – shop & process migration
Migrating a COBOL ERP and a NopCommerce shop to Shopware 6 – with a custom theme and plugins that map the existing business processes one to one.
We already use ChatGPT – isn't that enough?+
That's stage 0 to 1: individual people saving individual minutes. Adoption starts when AI sits inside your processes and you can measure the effect. That's exactly the jump we build.
What about data protection?+
EU hosting, data-processing agreements, clear rules on what data a model may see – that belongs in every build, not in a footnote. Where needed, models run in your environment.
Which tools do you use?+
Model-agnostic: Claude, OpenAI or open models – whatever fits the case. Orchestration mostly with n8n, connected to your systems via API. No tool religion, no license lock-in.
How do you measure success?+
Hours per task, cycle time, error rate – before and after, on the real process. If nothing measurably improves, it wasn't a success. That's the whole rule.
Tell us what's up.
30 minutes, free, no slides. Reply within 48 hours.