The first step is unclear
Leadership or teams see a need to act but do not yet have the shared AI Literacy required to identify and prioritise meaningful use cases.
AI consulting · AI literacy · responsible use
AI consulting for universities, research and science-adjacent organisations
I support universities, research services, Grants Offices, Graduate Schools, foundations, non-profit organisations and associations in integrating AI into their work in a meaningful, responsible and trustworthy way.
Together, we clarify which AI uses are genuinely useful, what competencies teams need and whether a workshop, clear guidelines or a limited pilot is the right next step.
AI literacy instead of tool hype
Understanding the research context
From use case to pilot
Responsibility remains human
Where does responsible AI use begin?
Universities and science-adjacent organisations rarely need another tool first. They often need to clarify which tasks are suitable for AI, which data may be used, which rules apply and which next step fits the organisation.
A short initial assessment creates direction without requiring a complete AI strategy or a large IT project.
Orientation or training is needed
Researchers, doctoral researchers, professional staff or governing bodies need a common foundation and practice-oriented learning.
AI use is already inconsistent
Different teams use different tools while shared safeguards, roles and review steps remain unclear.
Data and confidentiality matter
Personal data, research data, grant applications and internal documents require careful clarification before use.
Recurring questions consume time
Research support services, Grants Offices or central units repeatedly answer similar questions about procedures, regulations and support.
Organisational knowledge is scattered
Knowledge is distributed across PDFs, policies, websites and internal repositories, making it difficult to find and maintain.
Who this is for
Many AI questions arise in science-adjacent interface roles where research support, funding requirements, institutional rules, governance and implementation meet. Higher-education research sometimes describes this as the “Third Space”; in practice, concrete terms such as research support services, Grants Offices, research management, Graduate Schools, Transferable Skills programmes and science-adjacent central units are more useful.
Swiss higher education institutions
Universities, universities of applied sciences, universities of teacher education, vice-rectorates, faculties and science-adjacent central units.
Research support services, research management and Grants Offices
Teams working with calls, grant application processes, budgets, policies, confidential information and researcher guidance.
Graduate Schools and Transferable Skills programmes
Programmes for doctoral researchers and postdocs, academic careers teams and professional-development services.
Foundations
Grant-making organisations with distinct needs around grant intake, reporting, portfolio knowledge, internal guidelines and preparation for governing bodies.
Non-profits, associations and learned societies
Organisations working with limited resources, member communication, grant-seeking, project delivery and organisational knowledge.
Science-adjacent central units
International affairs, governance, science communication, quality development and other functions connecting research, institutional rules and implementation.
Why this distinction matters
Foundations
Foundations often approach AI from the perspective of grant-making: grant intake, reporting, portfolio knowledge, internal guidelines and preparation for governing bodies.
Non-profits, associations and researchers
These groups more often use AI from the perspective of grant-seeking, communication, project work, knowledge organisation and limited resources.
In both cases, AI can prepare, structure and reduce workload. Professional judgement and responsibility remain human.
Approach
AI consulting as empowermentFrom need to next step
Good AI adoption is not about delivering a finished solution that nobody can develop further. AI can make professional work better, but not automatically easier: tasks must be framed precisely, outputs reviewed critically and responsibility kept clear.
AI can temporarily support thinking, structuring and drafting, but it should not become permanent delegation of judgement. The process therefore connects orientation, use-case discovery, a workshop or pilot, and evaluation.
1
Clarify
Goals, workflows, audiences, data, existing rules and internal capabilities.
2
Prioritise
A small number of use cases based on value, risk, feasibility, data and organisational fit.
3
Pilot or train
Run a bounded pilot or build the relevant capabilities in a role-specific workshop.
4
Evaluate and embed
Review experience, document responsibilities and decide whether to adapt, expand or stop.
Consulting and implementation formats
Depending on the question, we begin with orientation, build capabilities in a workshop or test a concrete application in a clearly bounded pilot project.
AI Literacy and orientation
We clarify which tool fits the material, task and level of responsibility. The aim is understanding and decision-making capability, not a ranking of as many tools as possible.
Use-case discovery and prioritisation
Starting from real work, we identify a small number of applications that are useful, feasible and bounded enough for responsible testing.
Small, verifiable pilot projects
A pilot tests a clear purpose with defined users, sources, roles and evaluation criteria. Its findings support a decision to adapt, expand or stop.
A pilot can also test a recurring workflow with reusable AI skills or an agent. We define the steps, permitted data and tools, approval points and stopping criteria, then assess its usefulness against concrete examples.
Workshops: Prompting, AI skills and agents
I help researchers and teams formulate clear AI instructions, package recurring methods as reusable AI skills and delegate multi-step tasks to AI agents within defined limits. Formats reflect participants’ prior knowledge, tasks and institutional requirements.
The three areas in detail
- Prompting: Define tasks, sources, output formats and review criteria.
- Reusable AI skills: Develop reusable instructions with templates and reference material, then test them on your own examples.
- AI agents: Try multi-step tasks with defined access to data and tools, approval points and stopping criteria.
These areas can be covered individually or as a coordinated learning pathway. Participants work on their own tasks and review their results.
AI assistants with curated organisational knowledge
Regulations, procedures, FAQs and communication templates can become the basis for specialised AI assistants, chatbots or voicebots. RAG is only one possible background architecture. What matters are well-maintained sources, clear roles, good system prompts, tests with real questions and defined boundaries.
Chatbots, voicebots and internal assistants
These systems can support access to knowledge and recurring requests. Before implementation, we clarify the target audience, task, sources, escalation paths and situations in which a human must take over.
Copilot, not Autopilot
We distinguish where AI can assist productively from situations where it would take over too much judgement or responsibility. This supports practical safeguards for tools, data, tasks, outputs and human review.
Questions we clarify together
We begin with your real tasks and decisions, not a predefined list of tools.
Which tasks are suitable for meaningful AI support at all?
Which tools can we use with which kinds of content?
Where is AI a Copilot, and where would it take over too much responsibility?
Which use cases are useful and bounded enough for a pilot?
What rules do we need for public, internal and confidential material?
How do we build AI Literacy rather than relying on prohibitions alone?
When is a simple assistant sufficient, and when is RAG architecture useful?
How do we support interface roles across research, administration and governance?
Not simply “AI or no AI”: which use is responsible?
Whether AI is appropriate does not depend on the tool alone. The concrete combination of material × tool × task × output matters.
A public text, an internal record and a confidential grant application require different decisions, as do brainstorming, translation and professional assessment.
Material: Is it public, internal, confidential, personal or otherwise sensitive?
Tool: Where does the data go, what terms apply and is the tool suitable or institutionally approved?
Task: Is AI supporting a step, or is it evaluating, selecting or deciding?
Output: Will it remain internal, be published or affect a decision? Who reviews it?
Data protection, confidentiality and scientific integrity cannot always be addressed through simple yes/no rules. The material, tool, task, intended output and responsible reviewer must be considered together. The consulting work helps translate good scientific practice, transparency and accountability into concrete working rules.
Where legal requirements, data protection or internal approval are involved, the appropriate specialists in your organisation should be included.
Open the AI traffic-light tool for initial guidanceSelected collaboration contexts
Sollberger AI Consulting works on a mandate basis with organisations in the Swiss higher education and research environment.
Swiss Academy of Humanities and Social Sciences
Ongoing mandate-based collaboration with the Swiss Academy of Humanities and Social Sciences.
University of Bern
Ongoing mandate-based collaboration, including in the context of the Transferable Skills Program.
University of Fribourg
Completed mandate with the Research Promotion Service of the University of Fribourg / Université de Fribourg.
Why Sollberger AI Consulting
Dr. Michael Sollberger is the founder of Sollberger AI Consulting and holds a doctorate in philosophy. He combines analytical thinking with practical experience in the Swiss higher education, research and innovation sector.
He has more than ten years of experience in research management. His work combines AI workshops, presentations, introductions and consulting with the development or selection of suitable AI solutions for concrete tasks. This includes domain-specific AI assistants, chatbots and voicebots, as well as clarifying use cases, limits, data protection, quality assurance and human responsibility.
View profile and experienceFAQ
What is a useful first step for our organisation?
A useful start is to clarify goals, workflows, data, existing rules and internal capabilities. From there, we can decide whether orientation, a workshop, minimum safeguards or a bounded pilot is the appropriate next step.
Do we already need a complete AI strategy or policy?
No. For initial low-risk steps, a clear framework for purpose, permitted tools and data, responsibilities and human review is often sufficient. A broader strategy can build on practical experience.
Can confidential data or grant applications be entered into AI tools?
That depends on the material, tool, task and intended output. Public information, internal documents, personal data and confidential applications must be treated differently. Before use, the tool, necessary information, institutional rules and responsible reviewer should be clear.
What is the difference between an AI assistant and RAG?
An AI assistant, chatbot or voicebot is the working interface for a concrete task: finding information, drafting, structuring or answering recurring questions. RAG describes a technical architecture in the background. It allows the assistant, when needed, to retrieve information from a curated and maintained knowledge base and use it in its answer. The point is therefore not RAG as a buzzword, but designing a useful assistant with clear sources, roles, limits, tests and human responsibility.
Can a pilot begin without a large IT project?
Yes, if its purpose, users, sources, roles and review criteria are tightly defined. A pilot should test assumptions and create learning before any wider roll-out decision.
Which workshops are suitable for researchers and doctoral researchers?
Useful workshops connect AI literacy to participants’ own tasks: writing clear prompts, packaging recurring methods as reusable AI skills or delegating multi-step tasks to agents within defined limits. Literature and text work, source criticism, data-protection awareness and human review are part of the approach. The areas can be covered individually or together, depending on prior knowledge, needs and institutional rules.
Do foundations and non-profits have the same AI needs?
Not always. Foundations often approach AI from the perspective of grant management, reporting and portfolio knowledge. Associations and non-profits more often focus on grant-seeking, communication, project delivery and knowledge organisation. Human judgement and responsibility remain central in both cases.
How does a collaboration begin?
It normally starts with a non-binding initial conversation. We can then clarify the need, audiences, data situation, possible pilots and a realistic scope.
Discuss an AI project with institutional context in mind
If you want to explore responsible AI use for your university, research support service or organisation, a short first conversation is a useful next step.
Request an initial conversation