Mingle Collaboration × Lernforum Oberursel
Results report from the interviews with Susanna, the AI voice agent
Produced in December 2025 by Bianca N. Rohrbach and Robin T. Sverd.
All rights to the content are held by Linie 22 GmbH.
Client: All-in-One-Spirit — Dr. Matthias zur Bonsen, Jutta Herzog, Myriam Mathys.
The full report (German, ~45 pages) is available via the form below.
Executive Summary
The report brings together findings from voice-based interviews with “Susanna”, run to gain qualitative insight into how participants of Lernforum 2026 perceive and use AI in facilitation, consulting, coaching, leadership and learning. It combines key figures (participation, age, conversation length, usage rates) with thematic clusters and quotes.
Usage is predominantly internal, in preparation, and less in direct client-facing work; a relevant share does carry AI into delivery, while many remain cautious because of questions of transparency and trust.
169 people started the conversation; 142 responded in some form.
94 completed full interviews. Only these results were analysed.
Age (n=92): median 56; the largest groups are those in their 50s (40.2%) and 60s (34.8%).
Conversation length: 617.5 seconds (≈ 10:18 minutes).
Key figures
169 people started the conversation, 142 responded, 94 completed full interviews. Only the complete interviews were analysed.
169
Conversations started
People who started the conversation with Susanna.
142
Responses given
People who answered the agent in some form.
94
Complete interviews
Only the data from the 94 complete interviews was analysed for the report.
Maturity of AI use
Participants fall broadly into three segments.
64%
Regulars
High usage and practical integration. They use AI routinely and are looking for optimisation rather than introductory “how-to” content.
29%
Explorers
The transition from experimenting to routine. They have tried AI and want to move from “trying things out” to mastering more complex applications.
7%
Observers
The critical minority. They stay at the edge, bring in a critical perspective and keep the forum from becoming an echo chamber.
Who the participants are
The group is professionally experienced and often works in “role-fluid” profiles that combine facilitation, consulting, coaching, leadership and training. It consists above all of systemic change agents and hybrid role profiles — complemented by a significant share of leaders with decision-making responsibility.
92 people gave their age. The median is 56. The largest groups are those in their 50s (40.2%) and 60s (34.8%) — a professionally very experienced cohort.
| Metric | Value |
|---|---|
| Count | 169 conversations |
| Mean (average) | 612.1 seconds |
| Median | 617.5 seconds (≈ 10:18 min.) |
| Minimum | 102.0 seconds |
| Maximum | 1,449.0 seconds |
Consulting & organisational development
The strategic designers. External specialists focused on structural and strategic change, agile transformation and systems design.
Coaching, facilitation & training
The process specialists. Often freelancers or boutique providers focused on team dynamics, workshops, conflict resolution and personal development.
Leadership (executive level)
The decision-makers. CEOs, founders and department heads with operational responsibility across a range of industries.
Public sector & specialist roles
The niche specialists. Professionals from public administration, healthcare and architecture as well as IT and AI training.
HR & people development
The internal designers. Specialists for workforce upskilling, culture and internal learning ecosystems.
What AI is used for today
AI is used above all as a productivity and thinking tool — for writing, process design, sparring and research.
AI is used predominantly as a central tool for efficiency in writing and shows considerable potential to save time on administrative and communication tasks.
Participants make heavy use of AI in the preparation phase of workshops as well as for developing materials and new ideas for workshops, training and coaching.
Beyond basic content tasks, AI is used as a thinking and sparring partner for complex questions — including challenging interpersonal situations such as conflict facilitation.
AI tools in use
The tool family built on the GPT architecture (ChatGPT and custom GPTs) is the primary and most frequently named option. Perplexity is mentioned most often as a complementary research tool.
The GPT architecture, including ChatGPT and tailored custom GPTs, underlines its market and mindshare dominance in general LLM use.
Perplexity is often named alongside general LLMs — an indication of the need for verifiable sources in consulting and content creation.
Usage extends to enterprise tools such as Microsoft Copilot, large cloud platforms as well as creative and productivity tools (e.g. Gamma).
Audiences for AI use
Use is concentrated predominantly on internal preparation, self-support and team processes. The majority (around 76% of responses, or 50 cases) defines their AI interaction strictly as “internal”.
A significant minority (around 18%, or 12 cases) uses AI both for internal preparation and for external, client-facing interactions — bridging the gap between internal and external.
Direct, dedicated “client-facing” use without a preparation buffer remains rare (only 4 cases). Most professionals maintain clear boundaries and human oversight in order to preserve trust.
The client perspective
The reaction is not uniform but a clear polarisation between intense curiosity/enthusiasm and scepticism — depending on industry, role and personal experience.
A common pattern is the deliberate non-disclosure of AI use in order to maintain the perception of human control over the result.
Where AI use is made transparent, clients respond positively to the efficiency gains — particularly technically minded clients.
Experiences with AI
Enthusiasm about high accuracy and the ability to turn large, unsorted volumes of data into focused, actionable results.
The sheer speed and the immediate generation of answers were described as outstanding — usable as a starting point straight away.
AI helps to overcome creative blocks and to make sure nothing is missing.
The tendency of AI to produce generic, superficial or factually unreliable information offsets the speed advantage and creates mistrust.
Users find that high-quality results take unexpected effort — a collaboration that is hard to steer and gets compared to a “poor intern”.
The risk to professional authenticity and the question of what facilitators add once AI takes over the basic structuring.
Impact & adoption
4.67 / 10
Average rating of the current impact
At present AI has a moderate influence on professional work. The distribution is broad, with most responses in the middle range. This shows that for most people AI is helpful but not yet transformative.
Despite the moderate average, a clear polarisation emerges: around 33% fall into the low-impact group (values 1–3), while 15.7% (values 8–10) report a high to transformative impact. Ratings 1 to 5 together account for 61.4% — a clear potential for targeted training.
| Value | Count | Share |
|---|---|---|
| 1 | 8 | 11.4% |
| 2 | 5 | 7.1% |
| 3 | 10 | 14.3% |
| 4 | 10 | 14.3% |
| 5 | 10 | 14.3% |
| 6 | 3 | 4.3% |
| 7 | 5 | 7.1% |
| 8 | 4 | 5.7% |
| 9 | 3 | 4.3% |
| 10 | 4 | 5.7% |
| Total | 62 | 88.6% |
Distribution of the values (n=62) — 10 = very high impact, 1 = no impact.
AI in projects
A clear majority has already integrated AI into client or internal projects — AI is becoming a professional baseline standard.
70.6%
have integrated AI into projects
AI is increasingly seen as a widespread standard tool for professional delivery.
29.4%
no use in projects
A substantial minority, presumably shaped by barriers (uncertainty, skills, time).
The most common use is groundwork ahead of client interaction: brainstorming, workshop designs, draft agendas as well as working out complex proposals.
AI is used widely to speed up documentation and content production: emails, minutes and marketing content.
AI is also used for conceptually demanding, systemic tasks that require structural insight.
Forecasts for the changing role
The overwhelming majority sees AI primarily as relief in the back office.
Offloading cognitive and manual tasks creates more time for human-centred processes — collaborative decision-making and empathetic support.
AI is developing beyond a mere tool into an active co-worker; advising on AI integration is becoming a new service area.
A clear minority voiced cautious scepticism and ethical concerns — particularly about a possible loss of “truthfulness”.
Surprising experiences with AI
Surprise at the creative capability — with unusual concepts and convincing texts that also stimulate one’s own thinking.
The pace, precision and depth of the analytical capability made an impression — for instance when turning raw data into structured summaries.
The most profound surprises concerned human-like support — up to the feeling of a psychological “relationship”.
Not all surprises were positive: users were shocked by factual errors and unreliable analyses.
Advanced AI usage
The large majority has not yet taken the step towards custom GPTs and agents. The hurdle around configuration and light “programming” remains considerable — but a small, highly specialised “power user” segment is emerging.
12.9%
have built custom GPTs
11 of 85 participants — 87.1% have not yet taken the step.
14.1%
have worked with AI agents
12 of 85 participants — more than 85% do not work with agents.
For AI workflows the ratio between non-users and users is roughly 8:1 — a critical need for structured enablement to guide the majority to their first successful AI workflow.
Future learning goals
Moving beyond simple prompting to build complex, integrated AI workflows, custom GPTs and autonomous agents — automating entire processes.
Exploring advanced, hands-on applications in the core competence: how AI can support relational spaces, strengthen presence and “hold” the process in large groups.
The step from trial-and-error prompting to power-user level: understanding the “inner logic” of AI and developing a critical sensitivity to bias and ethical risks.
Barriers to AI use
The most important systemic barrier: the conflict between AI tools and strict internal security and data protection requirements.
A loss of trust caused by the tendency to “hallucinate” and to put probability in the place of verifiable truth — with considerable effort for checking and validation.
The barrier often lies less in a lack of ability than in the time and energy that good prompting demands, and in keeping up with a growing tool landscape.
Building trust
Trust requires clear ethical and technical guardrails: data-protected processing environments, transparency about data use and reliable control mechanisms.
A wish for intensive, structured learning formats (e.g. AI coaching days) in order to reach power-user level and gain a solid overview of what the tools can do.
Trust requires verifiable results and clear source references — plus the wish for AI to be able to process proprietary documents securely and locally.
Expectations for the Lernforum
About half expect concrete “show me” examples of how AI can be used in real facilitation settings — as a bridge from abstract potential to applicable methods and prompts.
Around a third emphasise exchange: seeing how colleagues work with AI, comparing levels of maturity and sharing lessons learned. The Lernforum as a practical community of practice.
About one in six puts trust topics front and centre — often as a precondition for use in client-facing contexts.
An accessible overview of tools, providers and market developments — plus clarity on what facilitators should know without having to become AI specialists themselves.
Protected spaces to try things out, to observe how AI changes dialogue and to reflect on limits as well as on “where humans are better than AI”.
Feedback on the interview experience
The overall experience is predominantly positive — average 7.4/10, median 8.
7.4 / 10
Average rating
Median 8 — the overall experience is predominantly positive (excerpt n=70).
≈77%
rate it 7–10
More than half (≈54%) give 8–10; 10/10 occurs regularly (≈11%).
≈11%
experience friction (≤5)
A small but relevant minority reports a weaker experience.
The dominant positive signal: the agent comes across as attentive and “human” — an empathetic, trustworthy voice and the feeling of really being heard.
A frequent criticism is over-affirmation: repeated reactions such as “fantastic/great” feel excessive and make the agent seem more “scripted”. Several people would prefer a more neutral tone.
The agent jumps in too quickly during thinking pauses, interrupts longer answers or asks several questions in one go. Wishes: more “wait time”, clearly separated single questions.
Mirroring and summarising is valued as evidence of understanding, but some experience it as formulaic. The nuance: keep the reflective technique, reduce the repetition.
Technical problems (dropouts, echo, latency) and interface aspects (white screen, no avatar) break immersion and comfort. Requested: a pause button, better audio performance, optional visual cues.
The voice agent
The voice agent Susanna collects qualitative data from Lernforum participants through a voice-based survey. This anonymised preliminary report came out of it, preparing the audience thematically for the event.
Susanna was designed as a calm, clear and curious conversation partner who invites reflection — in the style of an experienced facilitator. During onboarding it is made transparent that she is an AI voice collecting anonymised thoughts on AI in facilitation, coaching, consulting and leadership.
The basic design emerged from two co-design workshops with the clients and was developed over seven iterations into a final instructed prompt — with a clear role description, tone of voice, conversation logic and question structure.
Technologies used
Background
This report presents the results of an AI-supported survey conducted in the context of the Lernforum Oberursel (January 2026). The Sunday of the Lernforum is dedicated to the theme “Facilitation and AI” — a topic that triggers fascination and concern in equal measure and is deliberately designed as a space for experience and reflection.
In order to involve participants actively before the event, a privacy-compliant, intelligent interaction system was offered. Its core building block was an interactive survey conducted by a voice agent (“Susanna”): a voice persona developed specifically for this audience and accessible via a link.
The conversations were transcribed automatically and analysed with AI; from this came a compact pre-event report with trends, verbatim quotes and visual analyses. The development and implementation were supported by Mingle Collaboration (Bianca Rohrbach and Robin Sverd).
Practical insight into usage patterns, opportunities and limits of AI in work with people and groups. The full report runs to around 45 pages.
Get the full report here:
Your email address is processed in line with the GDPR and stored securely. You will receive the report and occasional information about offerings from Mingle Collaboration. No spam — promised.
Find out how Mingle Collaboration integrates AI into facilitation, consulting and collaboration — from voice agents to tailor-made workflows.