You do not have to solve it all at once.
Adopting AI at a university can begin with a single course, one administrative process, a workshop or one concrete integration.
We start by finding where there is a real problem and measurable value.
With one concrete problem.
AI is not a goal in itself.
A course
An AI tutor, practice, or a teaching workflow.
An administrative process
Support for internal knowledge, document handling or case work.
An integration
Connecting an existing university system or data source to AI.
An organisational unit
A concrete problem in one department, faculty or administrative area.
A workshop
When the question is still where using AI would make sense at all.
A good first use case is somewhere that
- has a real user need
- takes a lot of time or manual work today
- makes information hard to reach
- is a well-bounded task
- can be tried at small scale
- makes it measurable whether it added value
First we understand what happens today.
Before choosing technology, we map the problem, the people who would use it, and the systems already in place.
What we look at
- what the current process is
- where the largest friction sits
- who uses it
- which data and systems are involved
- what institutional constraints apply
- what AI can, and what it is not worth using AI for
- how success can be measured
Outcome
A prioritised first use case and a realistic pilot proposal.
Try it first, scale it after.
A pilot is not there to show off an impressive demo.
What we want to find out is whether the solution works with real users, in a real institutional setting.
A pilot can include
- a working AI solution
- the integrations it needs
- user access
- permissions
- basic governance
- measurement points
- user feedback
What we measure
Depending on the use case, for example:
- whether it is used
- how much time it saves
- what tasks it is used for
- what errors appear
- whether it improves the user experience
- whether it reduced repetitive work
- whether it is worth building further
When it is not yet clear what to build.
Many institutions already know they have to engage with AI, but not yet where to begin.
That is what a focused workshop is for.
Leadership workshop
- institutional AI opportunities
- priorities
- risks
- governance
- roadmap
Teaching workshop
- AI in teaching work
- student use of AI
- course development
- skills and good practice
Administrative workshop
- repetitive processes
- access to knowledge
- agents
- tasks that can be automated or supported
IT workshop
- integration
- MCP
- identity
- models
- data handling
- infrastructure
Outcome
A workshop should not be inspiration alone.
It should end with:
- a prioritised use case list
- a next step
- pilot candidates
- the decisions that need making
Adopting AI takes more than a system.
The people using it need to understand what AI is good for, what it is not, and how to work well with it.
Audiences
- teaching staff
- administrative staff
- leadership
- IT
- internal AI champions
Possible formats
- short practical training
- role-specific training
- a workshop
- preparing internal trainers
- training tied to using a specific system
Topics
- AI fundamentals
- thinking in tasks and workflows rather than prompts
- the strengths and limits of models
- data handling
- verification
- agents and skills
- institutional rules
The technology decision is only part of the work.
At institutional level, technological, organisational and regulatory questions usually have to be handled at the same time.
What we help with
- AI strategy
- use case prioritisation
- an institutional roadmap
- technology decisions
- choosing models and providers
- governance
- AI literacy
- shaping a pilot portfolio
- build versus buy decisions
Principle
We do not start from the assumption that every problem needs something built from scratch.
The goal is a solution that works, not a technology decided in advance.
Once the problem is clear, we build what it needs.
A pilot or an institutional solution can be an assembly of ready components, an integration of existing systems, or custom development.
This can be
- an AI tutor
- an internal assistant
- an agent
- a skill
- an MCP server
- an API integration
- a workflow
- a custom web or mobile interface
- a white-label solution
- AI connected to an institutional knowledge system
Worth saying
We are not trying to turn every project into a new product.
If a small, targeted integration is the right answer to the problem, that is what we build.
Control is not added afterwards.
Permissions, data handling and human oversight are part of the design of the pilot itself.
Questions
- what data the AI can reach
- who may use it
- what it may do on its own
- where human approval is required
- what has to be logged
- which models may be used
- what regulatory requirements apply
What works can be built further.
Not every pilot has to become an institutional system.
But what creates real value can take wider use as its next step.
Scaling can mean
- bringing in more courses
- use across more departments or faculties
- building a shared skill library
- introducing SSO
- further institutional integrations
- supporting external clients alongside our own
- formalising governance and operational processes
- training at organisational level
Worth saying
Scaling does not mean everything has to be merged into one central system.
Separate solutions should connect where there is a real advantage in connecting them.
What an adoption can look like.
Problem
A department sees its teaching staff spending a lot of time answering the same student questions.
Pilot
An AI tutor is built for one course.
Use
We watch how students use it, what they ask, and where the system gets stuck.
Learning
The course AI and the teaching processes are adjusted on what that shows.
Spread
More courses join.
Integration
SSO, an LMS connection or another institutional function is added.
New use cases
The technical and organisational experience gained makes pilots possible in other areas too.
There is no need to design one large transformation project up front.
A first step that works is something you can learn from.
What would be the first problem worth trying?
It could be a course, an administrative process, an integration — or simply the question of where to begin.