# Sven Jungmann

AI founder, physician and author. Founder and CEO of aiomics. His talks explore how AI changes the value of work, personal responsibility and competitive advantage, grounded in research and current operating experience. Available in German, English and French across Europe.

## Background and speaking

- [About](https://www.svenjungmann.com/about)
- [Talks](https://www.svenjungmann.com/keynotes)
- [Enquire about a talk](https://www.svenjungmann.com/booking)

## Original articles and practical guides

- [Can simulated customers support your next product decision?](https://www.svenjungmann.com/guides/synthetic-customer-responses-product-decision): An AI model can generate a thousand plausible answers. A short evidence brief helps assess which product decision those answers can actually support.
- [Does the next AI customer fit your development plan?](https://www.svenjungmann.com/guides/ai-service-new-customer-development-time): New commitments occupy the people who should improve the product. A calendar calculation makes onboarding, recurring delivery and deferred development visible before acceptance.
- [When may AI-assisted literature screening stop?](https://www.svenjungmann.com/guides/ai-literature-screening-stopping-unread-records): Finding few new relevant records is insufficient justification. How specialist teams and leaders can preserve what was assessed, what remains open and which decision a briefing may support.
- [Comparing AI offers: what counts as a billable outcome?](https://www.svenjungmann.com/guides/compare-ai-offers-billing-units-accepted-results): A price per outcome requires a precise definition. How purchasing and operational teams can compare billable units, accepted cases and the work that remains.
- [When only one person can maintain your AI tool](https://www.svenjungmann.com/guides/internal-ai-tool-handover-maintenance-ownership): How to assess whether a colleague can maintain an internal AI tool: rehearse tasks, record the support required and transfer a defined scope of responsibility.
- [After an AI keynote: which follow-up is worth commissioning?](https://www.svenjungmann.com/guides/ai-keynote-follow-up-deliverables-evidence): A guide for event organisers: commission a suitable follow-up, assign responsibility and assess reach, learning and workplace application separately.
- [What does a second AI model actually check?](https://www.svenjungmann.com/writing/second-ai-model-review-agreement): Two models agree. Which errors still survive? A study and a fictional calculation help leaders assess an additional AI review step before buying it.
- [Which usage targets can AI surveys justify?](https://www.svenjungmann.com/writing/ai-surveys-usage-targets-denominators): Before a survey becomes an organisational instruction, examine its population, outcome and inference. With BCG, Gallup and a fictional adoption-rate calculation.
- [Who reviews the extra work that AI prepares?](https://www.svenjungmann.com/writing/ai-approvals-attention-review-capacity): More prepared work requires available attention. How leadership teams can assess review effort, exceptions and backlogs before expanding AI use.
- [Who is missing from your successful AI pilot?](https://www.svenjungmann.com/writing/ai-pilot-volunteers-selection-transfer): Early volunteers help a team learn. Wider adoption also depends on who received an invitation, could actually participate and remained visible in the results.
- [Who can actually use your AI voice service?](https://www.svenjungmann.com/guides/ai-voice-service-language-speech-access): A fluent answer does not establish who can use a voice service. Examine languages, ways of speaking, abandoned attempts and whether the alternative communication route actually works.
- [What another hospital needs from an AI incident report](https://www.svenjungmann.com/writing/learning-from-ai-incident-reports): Share AI incidents so another hospital can investigate: describe relevant conditions, retain uncertainty and keep report counts distinct from event rates.
- [Which assumptions hold up your AI strategy?](https://www.svenjungmann.com/writing/test-ai-strategy-assumptions): Turn an AI opportunity into assumptions you can examine: separate benefit, purchasing and revenue, then justify the next resource commitment.
- [What should a negative result teach the research team?](https://www.svenjungmann.com/writing/ai-research-negative-results-experiment-selection): When AI supplies many hypotheses, informative laboratory work becomes scarce. A proposal for research leaders to distinguish explanations and keep negative findings useful for later selection.
- [What remains when an AI feature is copied?](https://www.svenjungmann.com/writing/ai-product-durability-beyond-features): An AI company can face feature imitation while accumulating useful capabilities. Examine which demonstrable customer benefits the next deployment can inherit from the work already done.
- [When AI changes: what must be checked before renewed release?](https://www.svenjungmann.com/guides/retest-ai-changes-before-renewed-release): Models, sources and tools change. A practical approach to identifying which prior commitments still hold and who authorises continued AI use.
- [Patient questions between visits need a responsible recipient](https://www.svenjungmann.com/writing/patient-questions-between-visits-responsibility): An AI companion creates a service commitment between appointments. A proposal for hospital leaders to follow open questions through acceptance, action and resolution.
- [AI sovereignty: who can actually change the system?](https://www.svenjungmann.com/guides/ai-sovereignty-data-model-operational-control): How hospital, pharmaceutical and medical technology leaders can examine powers over data, model changes and administration, using current European sources and a fictional example.
- [Who owns the outcome when AI work crosses departments?](https://www.svenjungmann.com/writing/ai-outcome-ownership-across-departments): A fictional procurement case shows how AI can shift work between teams. Outcome ownership needs a shared definition of success and clear decision rights.
- [Prepare an AI talk around your audience’s decisions](https://www.svenjungmann.com/guides/prepare-ai-talk-organiser-brief): A guide for organisers: define the audience, decision and format, distinguish scientific claims, and discuss accessibility, recording and medical education recognition early.
- [Five questions about an AI claim: TrialGPT for pharmaceutical leaders](https://www.svenjungmann.com/guides/pharma-ai-claim-trialgpt-five-questions): What 42.6% less screening time means, and the five questions pharmaceutical leaders should answer before turning that claim into an investment.
- [Before your AI pilot, decide what would make you stop](https://www.svenjungmann.com/guides/precommit-ai-pilot-decision): Agree interruption rules, sufficient evidence and the next decision before an AI pilot. Includes a 20-page booklet with fillable working pages.
- [Verify AI outputs: sources, omissions and contradictions](https://www.svenjungmann.com/guides/verify-ai-output-against-sources): A reproducible source exercise demonstrates five planted errors, two omissions and an unresolved discrepancy in the primary paper. Two working instructions help structure your own review.
- [Who maintains the knowledge your AI uses?](https://www.svenjungmann.com/guides/ai-organisational-knowledge-ownership): A practical guide to reliable definitions, ownership and current knowledge, using a capacity-planning example and a small maintenance agreement.
- [What an AI benchmark can tell a hospital](https://www.svenjungmann.com/writing/medical-ai-benchmarks-hospitals): Technical capability, collaboration and patient outcomes answer different questions. Eight supplier questions help hospitals connect benchmark evidence to the intended use.
- [Evaluate AI pilots: quality, effort and value](https://www.svenjungmann.com/guides/evaluate-ai-pilots): A practical process for leaders: define one task, compare representative cases and count the complete effort required to produce a usable result.
- [AI saves ten minutes. What happens next?](https://www.svenjungmann.com/writing/ai-time-savings-organizations): How less drafting time becomes a real benefit: a complete worked example including review, rework, abandoned attempts and the decision about where the time saved should go.
- [AI agents: define permissions and approvals](https://www.svenjungmann.com/guides/define-ai-agent-permissions): Which actions may an AI agent take? A guide to bounded assignments, technical permissions, approval points and verifiable execution.
- [AI and early careers: how leaders can read conflicting employment evidence](https://www.svenjungmann.com/writing/ai-early-career-employment-evidence): US payroll data, a Danish study and an international task index answer different questions. A guide to interpreting them when making workforce decisions under uncertainty.
- [What can we still do after AI has helped?](https://www.svenjungmann.com/writing/ai-independent-learning): Two education studies show how strongly the effects of AI depend on the learning environment. They raise practical questions for clinical education and leadership development.
- [When AI shares expertise: what should experienced professionals do?](https://www.svenjungmann.com/writing/ai-expertise-transfer): A large customer-support study raises a leadership question: how can organisations share expertise while continuing to develop the capacity for new decisions?
- [An AI learning session that rehearses the next decision](https://www.svenjungmann.com/guides/ai-training-decision-practice): An adaptable learning sequence for healthcare and business: form a judgement, use assistance, work through a new case and assess learning again later.
- [When AI becomes unavailable: prepare for continuity and vendor exit](https://www.svenjungmann.com/guides/ai-service-continuity-vendor-exit): A guide for leaders to define minimum operations, identify dependencies and rehearse recovery using a fictional case. Includes preparation for a provider change.
- [Maintaining shared AI instructions: test changes and preserve versions](https://www.svenjungmann.com/guides/maintaining-shared-ai-instructions): A practical maintenance agreement for shared AI instructions: purpose, test cases, changes, approval and returning to an earlier version.
- [When people and AI disagree: resolve the disputed decision](https://www.svenjungmann.com/guides/resolving-human-ai-disagreement): An adaptable procedure for organisational decisions: define the disagreement, check authoritative information and record the cause for future cases.
- [Ending an AI initiative while preserving what was learned](https://www.svenjungmann.com/guides/ending-ai-initiative-preserving-learning): How leaders can compare continuing, pausing, redirecting and ending an initiative, with clear next commitments and an orderly handover.
- [The responsibility behind an AI draft](https://www.svenjungmann.com/writing/ai-drafts-review-responsibility): An AI draft needs inspectable evidence and a clear handover of responsibility. A fully fictional administrative workflow shows who approves each claim and the action it may support.
- [An AI-designed antibody binds: which decision comes next?](https://www.svenjungmann.com/writing/ai-antibody-binding-development-decisions): What a binding result means for drug development, and how pharmaceutical companies can select the next experiments by the decisions they inform.
- [A robot’s capacity includes the people who support it](https://www.svenjungmann.com/writing/robot-capacity-human-support): How organisations can include human assistance, interruptions and changing environments when planning the use of physical AI.
- [AI electricity use: assess requests, workflows and infrastructure separately](https://www.svenjungmann.com/writing/ai-electricity-request-workflow-infrastructure): How leaders can interpret AI electricity figures and turn individual requests into an accountable view of a complete workflow.
- [Buying or building AI: the complete cost calculation](https://www.svenjungmann.com/guides/buy-or-build-ai-whole-cost): A fully hypothetical worked example includes integration, review, corrections, operation and exit. Inspect the assumptions and the costs that can change the decision.
- [When experienced professionals hesitate over AI: what remains unresolved?](https://www.svenjungmann.com/writing/experienced-professionals-ai-hesitation): How leaders can turn reservations about AI into concrete questions without inferring motives from experience, age or a refusal.
- [AI makes the task easier. How should collaboration change?](https://www.svenjungmann.com/writing/ai-task-capability-team-redesign): A reflection on Benedict Evans’s conversation: what individual results tell us about team composition, and how a new way of working can be examined.
- [What an AI disclosure changes about a medical answer](https://www.svenjungmann.com/writing/ai-disclosure-medical-answer-trust): Identical answers can receive different evaluations depending on their attributed source. What research shows about AI disclosure, clinical review and clear explanations.
- [Ambient documentation: what remains after working hours?](https://www.svenjungmann.com/writing/ambient-documentation-after-hours-work): What clinical studies show about note-writing time, exhaustion and after-hours work, and which measurements hospital medical leaders need for a local pilot.