Decide clearly where AI can really help.
You know which AI step really helps your team and how it fits your workflows, before individual experiments become lasting isolated solutions.
I combine AI consulting with practical implementation for companies where software development, specialist work, and customer care work together.
Request an intro callIndividual AI solutions can create more work.
When each department builds its own tools, knowledge can remain in individual teams. Interfaces are missing, while checking and rework move to the next team.
The more work depends on them, the more involved joining them up later can become. That is why we agree shared measures before you buy more tools or expand pilots.
Improve the whole workflow.
We examine the whole workflow: how knowledge stays with the team, how systems connect, and how solutions can grow. Impact, data access, and internal requirements are part of the discussion from the outset.
Three independent horizons.
Each horizon can be selected directly. There is no required sequence.
Decide on AI before the direction is clear.
The team expects orientation. You need a next step you can stand behind professionally.
Understand and orient
Know which next step holds up
We use your real tasks to clarify where AI helps and which next step is worth taking.
Your team brings questions and work examples.
Approach
The compact workshop addresses one question we agree together, including preparation and documented results. Using examples from your work, we clarify what is worth testing and how progress can be recognised.
Clarify the reason, questions, and measure for good orientation
Look at or test real examples together
Record findings, open points, and the next decision
A workshop can provide requirements for a later pilot. Productive integration is not included automatically.
Assessment
When possibilities are suspected but priorities and prerequisites remain unclear, we examine conversations, representative work examples, and permitted data. Unknowns and data gaps remain visible.
A usable basis for the next decision without a follow-on engagement: possible applications, obstacles, and reasoned priorities.
No complete process inventory or guaranteed business case. We agree the number and depth of cases beforehand.
Examples
Here is how we could work together. We agree your project together.
Example: AI orientation
AI is on the agenda, but the department still lacks a shared view and a decision about where to begin.
In a workshop, we connect the technical possibilities with the department’s actual work, risks, and decisions ahead.
A short decision brief with open questions, priorities, and a reasoned next step.
Example: AI for specialist tasks
A specialist team sees recurring tasks, but does not know where AI genuinely helps and where it creates more checking.
We test selected tasks, record quality limits, and separate useful preparation from specialist decisions.
An assessment against real tasks, documented quality limits, and a reasoned decision to run a next test or not use AI.
The proposal arrives quickly. The work stays with the team.
Review, questions, and rework still take time. What matters is what genuinely reduces the workload in the end.
Implement together
Test a working approach the team can carry forward
I develop a bounded AI use case with your team and test its effect.
Your team retains expertise, approvals, and operation.
Approach
We bound one use case and establish the goal, baseline, quality limits, and total human effort. I develop, configure, and test alongside your team. The result is a working approach tested within the agreed scope that your team can carry.
Establish the case class, goal, baseline, quality limits, and total human effort
Develop, configure, and test with real tasks
Check the effect against comparable cases and document how the team carries it forward
We check the effect against comparable cases and decide whether and how to continue. Larger work is agreed separately.
Examples
Here is how we could work together. We agree your project together.
Example: Verified bug fixes
Bug reports, reproduction, proposed changes, tests, and approvals sit in separate steps. Handoffs and missing context consume time.
AI prepares reproduction and code proposals. People check tests and approve changes. We measure total human effort including support, review, and rework.
A tested approach for an agreed bug class, from a structured report through verification and approval to an accepted fix, measured against total human effort.
Example: Coding agents
A team wants to use AI agents on real changes without bypassing repository context, tests, or reviews.
We bound tasks, prepare project context and tests, and define where people decide and verify. The team tries the approach itself on real changes.
A documented approach for real changes: context, agent assignment, tests, review, and independent use by the team.
Example: Answers with sources
Recurring service questions mean people spend too long searching or find conflicting answers.
AI searches selected sources and prepares answers; specialists check evidence and compare search and checking time on similar questions.
A tested search-and-answer approach with source checks and escalation of unresolved questions.
Example: Check order data
Missing details or unsuitable files are often found too late; technical order data must be completed, checked, and handed over.
We check representative cases early for missing files and input errors, agree checking steps and quality limits, and make exceptions visible.
A tested approach in which AI prepares the data and the specialist team owns checking and handover.
Many AI initiatives. But no clear priority.
All need time and budget. Without a shared measure, it remains unclear what continues and what can wait.
Develop strategically
Set priorities you can explain
We align your AI initiatives around a shared goal and make progress measurable.
With your leaders, we clarify priorities and ownership.
Approach
We clarify intended direction and indicators that show whether the department is moving towards it. This makes next steps, owners, and a review rhythm reasoned and ready for a decision.
Clarify decisions, direction, and indicators for department progress
Prioritize development steps, ownership, and capabilities
Review the effect and sharpen decisions
Larger technical implementation is agreed separately.
Examples
Here is how we could work together. We agree your project together.
Example: AI in your department
A department wants to use AI, but direction, sequence, roles, and investment are not yet decided.
We connect direction and priorities with capabilities, responsibilities, and implications for actual work.
Options ready for a decision and prioritized development steps with owners and a regular review rhythm.
Example: Review AI initiatives
Several applications compete for budget; effect, total cost, risk, and strategic contribution are not comparable.
We develop criteria, compare the available evidence, and prepare the decision.
Reasoned expand, replace, or stop decisions with a next step and owners.
How do we check the effect?
For orientation, we agree criteria for learning, shared understanding, and decisions. For implementation, we assess quality and total human effort on comparable cases before and after the change. For strategy, meaningful KPIs show whether the department is moving towards its intended direction.
Against the agreed measure. Embedded in the team.
You can see what has changed and what your team can do next. We clarify the starting point and measures for this at the outset.
How we begin working together.
Discuss the reason
Send two or three sentences on the work and question; no slide deck is needed. We clarify fit and a bounded next step.
Agree the engagement
We agree the goal, scope, collaboration, responsibilities, and fee before work begins.
Clarify the starting point and measure
At the outset, we agree how progress will be recognised. This clarification itself makes unclear goals, assumptions, and missing information visible.
Work together
I contribute structure, technical knowledge, and hands-on work. Together, we test and review progress.
Check the effect and carry it forward
We compare with the starting point and decide to continue, adapt, or stop. A follow-on is not automatic.
Data and responsibilities first.
Before we begin, we clarify permitted access, processing locations, privacy requirements, internal rules, responsibilities, and agreements. A test with approved sample data can begin before production access.
Your team owns specialist knowledge, decisions, approvals, administration, and operation.
Milan Kappen
I am a software engineer, have been a CTO and Head of Engineering, and have worked with machine learning since 2015. I bring technical depth to decisions that need to hold under time pressure.
At Ogulo, I built AI-assisted floor-plan generation and automated production workflows. At DeepUp, I contributed to digitisation as Lead Engineer and Head of Engineering; during my time there, 3D digitisation per operator rose from about 200–300 metres to 15–25 kilometres a month.
I know the friction between Engineering, Customer Success, Support, Sales, and Marketing: when handoffs and rework arise because promises, processes, and data do not fit together. My work is to make that friction visible in a concrete workflow - with clear boundaries, responsibilities, and checkable steps.
For you, this means seeing where work gets stuck, making decisions you can explain, and checking from the outset whether the change actually helps your team.
Frequently asked questions
Do we have to begin with an Assessment?
No. It is one possible, separately commissioned engagement within “Understand and orient”. For a clear implementation initiative, engagement clarification and the baseline belong to that initiative.
Do we need a business case already?
No. AI orientation can be a useful reason to begin. In the intro call, we clarify which question and next step fit.
Why clarify this now rather than later?
Each additional isolated solution can add dependencies: knowledge stays in teams, while interfaces and privacy questions arrive later. Before buying another tool or expanding a pilot, we clarify whether workflow, responsibilities, and scaling fit together.
What does working together cost?
For orientation: compact workshop including preparation and documented results from €2,500 net; bounded implementation or pilot from €12,000 net; strategic initial engagement from €6,000 net, then advisory support from €1,200 net/month for about eight flexibly allocated hours. These starting prices are not fixed quotes: we agree scope and fee before engagement. Extra tool, operating, and, where applicable, travel costs are agreed separately.
Do we need AI for this?
We assess the concrete use case openly. It can be right not to use AI or to automate one step conventionally.
Who is this a fit for?
It is especially suited to companies where their own software development, specialist work, and customer care work closely together, and many tasks repeat within individual customer services.
What happens in the intro call?
Send two or three sentences on the work and question; no slide deck is needed. We discuss fit and a bounded next step. If I am not the right person, I will say so.
What do you do, and what does our team do?
I develop alongside the team, configure, test, investigate errors, and share knowledge. Your team provides specialist knowledge, named owners, and agreed participation; it owns decisions, approvals, administration, and operation.
Bring clarity to the next AI decision.
A clear next question makes the decision easier. Before a budget or tool decision, we clarify in the intro call whether and how I can support you.
Request an intro call