At the Portfolio Private Health Forum 2026, presenters argued that artificial intelligence on its own is largely an expensive piece of software, and that the real competitive edge will come from hybrid organisations that combine human staff with digital agents. Orbán Előd, co-founder and CEO of Enterprise Group Technologies (EGT) and European Life Technologies, emphasized that human–machine collaboration functions multiplicatively rather than additively — organisations with rigid, outdated cultures risk nullifying their digitalisation efforts.
Three levels of AI integration
Orbán outlined three abstraction levels for applying AI in organisations:
- Tool level (vector database): solutions used for individual tasks (for example, dictating reports). They can produce fast gains but typically plateau after delivering a few tens of percent improvement.
- Process level (knowledge graph): technology permeates the entire patient journey, generating multiple-fold benefits but requiring significant data integration.
- Organisational level (ontology and agents): networked digital agents appear; this level can increase efficiency by orders of magnitude and is primarily a management and strategic issue.
His conclusion: the aim should not be to "buy AI," but to build a hybrid organisation.
Digital agents in hospital hierarchies
In envisioned 2027-era clinical organisational charts, alongside traditional roles (institution head, finance, medical leadership) digital colleagues — known as agents — are present. Orbán explained that an agent is more than a simple algorithm: it has a name, defined responsibilities, can make autonomous decisions within set boundaries, and is supervised by a human custodian.
He underlined that severe staff shortages are one of healthcare’s biggest current problems. Digital agents are intended not to replace existing personnel but to substitute for missing staff: by handling administration and data organisation they free time for doctors and nurses to spend with patients.
Knowledge graphs, ontologies and rule governance
A knowledge graph stores not only data but the relationships between data, so the system becomes ‘‘smarter’’ with each new piece of information. An ontology is a machine-interpretable description of a company’s operational rules; together these enable agents to act autonomously while remaining within the organisation’s rule set.
Orbán also pointed to an important economic dynamic: tasks that can be automated (such as data entry, patient routing, or preparing reports) will become extremely cheap, whereas pure human attention, image interpretation and empathy — elements only a flesh-and-blood clinician can provide — will become premium services patients are willing to pay for.
Data readiness is decisive
According to the presentation, 78 percent of companies are currently unprepared to adopt generative AI, largely because they lack adequate data foundations. Orbán stressed a critical point: corporate knowledge that a digital agent cannot query in structured form effectively does not exist for the agent. Keeping the corporate knowledge base continuously updated is essential for success.
EGT’s operational practices were cited as examples: the company prepares roughly 100 knowledge-base updates per week via agents, which are then approved by humans. This ongoing self-check allows quality assurance audits (such as ISO certifications) to be part of everyday operation rather than lengthy, separate projects.
Implementation approach and field data
Orbán warned against treating transformation as a single large IT project: the recommended approach is ‘‘pilot & pivot’’ — start small with focused experiments and iterate based on results. He noted that roll-out begins with inspiration and understanding, not with a licence purchase.
In EGT’s system today there are more than 150 agents, and 80 percent of employees use them daily. Agents support a three-person business decision-making board in a way that effectively lets the team perform the work of nine people. Orbán concluded that agents can be autonomous where rules are deterministic; for decisions requiring nuanced human judgment, people will remain responsible.
