AI Agents and the Future of Health | Stockholm MLOps Event #33 | May 18 2026

Insights from the Stockholm MLOps community

Healthcare AI is no longer just about pilots, chatbots, or isolated tools.

At Stockholm MLOps #33, hosted around the Arkus AI launch, the conversation moved into something much more operational:

How do AI agents actually enter healthcare systems?

Across founders, clinicians, researchers, infrastructure builders, and healthcare operators, one signal stood out clearly:

Healthcare AI does not fail because the models are not smart enough.

It fails when systems cannot integrate, validate, govern, and operate the technology safely inside real healthcare workflows.

The strongest discussions focused on:

  • interoperability

  • clinical workflows

  • patient empowerment

  • runtime evaluation

  • compliance

  • integrations

  • organizational readiness

  • and agents as operational healthcare infrastructure

Summary — Key Insights from the Event

  • Healthcare AI fails in workflows, not models

  • Interoperability is becoming strategic healthcare infrastructure

  • Trust depends on continuous evaluation, not one-time approval

  • Patient empowerment will force healthcare transformation

  • Compliance is becoming part of the deployment layer

  • AI agents are becoming workflow interfaces

  • Specialized healthcare software is becoming economically feasible

Key Insights from the Event

1. Healthcare AI Fails In Workflows, Not Models

“The biggest challenge is not the model. It’s evaluating if the system still behaves correctly after every change.” — Eskil Forsell, Karolinska University Hospital

“The difficult part is not the LLM. It’s the routing, integrations, security, and compliance.” — Johan Malmqvist

The event repeatedly pushed the same point:

The model is not the bottleneck.

The hard part is everything around it:

  • integrations

  • routing

  • compliance

  • validation

  • workflow fit

  • and operational safety

Healthcare AI has to work inside complex, high-stakes systems. That means deployment is not just a technical milestone. It is an operational responsibility.

2. Interoperability Is Becoming Healthcare Infrastructure

“Interoperability is foundational healthcare infrastructure.” — Helena Holma

“Healthcare platforms need to become more user-centric and integration-friendly.” — Helena Holma

The strongest infrastructure signal was clear:

Healthcare AI depends on system connectivity.

Without interoperability, AI agents become isolated tools. With interoperability, they can become part of clinical workflows, patient journeys, documentation systems, and operational decision-making.

This is why APIs, semantic interoperability, EHR integrations, and platform connectivity matter so much.

Healthcare AI is not a standalone application problem.

It is an infrastructure problem.

3. Trust Is Operational, Not Philosophical

“We need AI systems that communicate the degree of evidence behind recommendations.” — Svetlana Lagercrantz

“We need systems for evaluating evolving AI systems continuously.” — Eskil Forsell

Trust was not discussed as a branding issue or abstract ethical principle.

It was discussed as an operational system.

Healthcare AI needs:

  • evidence transparency

  • continuous validation

  • clinician verification

  • scoped permissions

  • auditability

  • and safety controls

The core challenge is not only whether an AI system works today.

It is whether it still behaves correctly after every update, every workflow change, and every new use case.

4. Patient Empowerment Will Force Healthcare Transformation

“Patient empowerment will force healthcare transformation.” — Johanna Hultcrantz

“What was controversial ten years ago is now normal.” — Johanna Hultcrantz

The event pointed to a major shift:

Patients will increasingly expect access, agency, and participation in their healthcare data and decisions.

That changes the operating model of healthcare.

Patient empowerment is not just a UX trend. It affects:

  • data access

  • care coordination

  • decision-making

  • consent

  • monitoring

  • and communication

AI agents may become one of the interfaces through which patients and families navigate complex healthcare decisions.

5. Compliance Is Becoming A Deployment Layer

“Operationalizing compliant AI at scale is the real challenge.” — Eskil Forsell

“You need systems that feel like a safe Volvo.” — Johanna Hultcrantz

The event did not frame regulation only as a blocker.

A stronger signal emerged:

Compliance is becoming part of the system architecture.

In healthcare, AI systems need to be designed with:

  • GDPR

  • MDR

  • auditability

  • safety

  • access control

  • and risk analysis

from the beginning.

The challenge is no longer simply getting permission to use AI.

The challenge is building systems that can be trusted, evaluated, and operated under healthcare-grade expectations.

6. AI Agents Are Becoming Workflow Interfaces

“We limit what the agent is allowed to do so it doesn’t accidentally update or delete the wrong information.” — Vladimir Li, Arkus AI

“AI agents are becoming operational healthcare infrastructure.” — Daniel Gillblad, Arkus AI

The agent discussion was practical, not abstract.

The important question was not:

Can an agent act?

It was:

What should the agent be allowed to do?

That means healthcare agents need:

  • scoped permissions

  • workflow boundaries

  • human oversight

  • safety rails

  • and clear operational roles

In healthcare, autonomy without boundaries is not useful. The agent has to operate inside a controlled system.

7. Healthcare Staff Are Already Moving Faster Than Policy

“Healthcare staff are already using AI tools outside official policies.” — Eskil Forsell

“Healthcare organizations are under immense pressure from staff who expect AI tools.” — Eskil Forsell

This is one of the most important adoption signals.

The demand is already inside the system.

Healthcare professionals are facing administrative pressure, documentation burden, and workflow inefficiency. Many are already experimenting with AI, even when official systems have not caught up.

That creates a governance gap.

The question is no longer whether healthcare staff will use AI.

The question is whether organizations can provide safe, compliant, operational alternatives fast enough.

8. Specialized Healthcare Software Just Became Economically Feasible

“Building this from scratch would have taken months.” — Leon Saidavi, Uppsala University

“As a solo builder, I was able to ship comparatively fast.” — Leon Saidavi

“AI lowers the cost of creating highly tailored healthcare solutions.” — Johanna Hultcrantz

The event showed how quickly AI is changing the economics of healthcare software.

Highly tailored tools for smaller patient groups, clinical workflows, multilingual support, wellness, prevention, and life sciences are becoming more feasible.

That matters because healthcare has many niche problems that were historically too expensive to solve with traditional software teams.

AI agents and platforms like Arkus may lower the barrier for health builders to prototype, validate, and deploy specialized solutions.

Operational Signals Emerging From Meetup #33

Several larger signals emerged across the event:

  • healthcare AI is moving from pilots into operational systems

  • agents are becoming interfaces into healthcare workflows

  • interoperability is becoming a prerequisite for adoption

  • trust depends on continuous evaluation and evidence transparency

  • compliance is moving into deployment architecture

  • clinicians are becoming AI validators and operators

  • patient empowerment will reshape care delivery

  • specialized healthcare software is becoming cheaper to build

  • organizational readiness is now a bigger bottleneck than model capability

The strongest overall signal:

Healthcare AI is becoming infrastructure inside complex systems, not software on the side.

What This Event Signals

Stockholm MLOps #33 showed that healthcare AI in Sweden is moving into a more mature phase.

The conversation was not about whether AI can help healthcare.

It was about how AI can be:

  • integrated

  • validated

  • governed

  • trusted

  • operated

  • and scaled

inside real healthcare environments.

The future of healthcare AI will not be defined only by better models.

It will be defined by the systems that make AI usable:

  • interoperable platforms

  • agent permissions

  • clinical validation

  • patient data flows

  • regulatory readiness

  • and workflow integration

The core insight from the event:

Healthcare AI is not limited by intelligence.

It is limited by interoperability, trust, and operational execution.

Speakers & Companies Featured

  • Daniel Gillblad — Co-founder, Arkus AI

  • Vladimir Li — Co-founder, Arkus AI

  • Ying Cheng — CEO / Co-founder, Arkus AI

  • Johanna Hultcrantz — Chief Innovation Officer, Cambio

  • Eskil Forsell — AI Lead, Karolinska University Hospital

  • Fredrik Borgström — CEO, Foja

  • Britta Stenson — Director Life Science, Vectura

  • Svetlana Lagercrantz — MD / Professor, Karolinska University Hospital

  • Johan Malmqvist — Founder, QVoxl / Cardiologist, eHeart Sophiahemmet

  • Jack Yu — CEO / Life Science Expert, Nordberg Medical

  • Leon Sami Saidavi — Uppsala University

  • Helena Holma — CEO / Co-founder, Leyr

  • Maja Magnusson — CEO / Co-founder, Care to Translate

  • Sajjad Saffari — CEO / Co-founder, Noord Ark

Topics covered:

  • AI agents in healthcare

  • healthcare interoperability

  • patient empowerment

  • clinical workflows

  • healthcare AI governance

  • operational trust

  • compliance

  • life sciences

  • wellness

  • healthcare infrastructure

  • AI-assisted healthcare software

Related Insights

Upcoming Events

👉 Designing, Embedding & Orchestrating Physical AI — Stockholm MLOps #34
👉 Unlocking Enterprise AI Productivity — Stockholm MLOps #35

Event Details

📍 Stockholm
📅 May 18, 2026
🤝 Arkus AI Launch
👥 Stockholm MLOps Community

👉 Stockholm MLOps Community:
https://www.meetup.com/stockholm-mlops-community/

Next
Next

Guardrailing Vibe Coding | Stockholm MLOps Event #32 | May 5 2026