Article
Published
3 August 2026
The AI landscape: traditional, generative, and agentic AI
Traditional artificial intelligence (AI) refers to machine learning and statistical models that operate on structured data to automate, predict, and optimise decision-making processes. These systems are designed to identify patterns and improve outcomes in areas such as quality, clinical efficiency, patient safety, and person-centred care.
Generative AI (GenAI) extends these capabilities by using large language models (LLMs) and other generative architectures that are built to handle general inputs and outputs. GenAI can analyse and create content from unstructured data such as large documents, clinical imagery, or audio from patient interactions, enabling more interactive and context-aware use of information. An assistant (sometimes used interchangeably with agent) is an LLM configured to perform a specific task through prompt engineering and data access.
Agentic AI systems equip LLMs with autonomy and access to a predefined toolbox, enabling them to complete multi-step clinical and operational tasks, such as drafting discharge summaries, querying EHRs, and coordinating referrals.
Data, artificial intelligence, and new technologies offer fantastic opportunities to bind the healthcare system more closely together […] How do we create a humane healthcare system founded on care, trust, and security, where every citizen, no matter where they live, has access to the highly specialised treatment they need?
Rasmus Møgelvang, Hospital Director, Rigshospitalet
State of AI in healthcare
The healthcare sector faces sustained cost pressure on already high public expenditure, alongside a demographic squeeze in which rising demand meets a shrinking workforce. That combination forces a structural change to reduce the workforce-intensity in care delivery.
Clinicians often work on several tasks in parallel, and the complexity of that everyday reality leaves little room to absorb new tools efficiently, which slows in-situ adoption. To continue delivering quality, patient-centred services, there is a need to adopt novel technologies responsibly and create real impact for patients and HCPs in their everyday tasks.
To understand the current level of AI adoption within the healthcare sector, it is worth distinguishing between two types of use cases:
- clinical use cases built on complex patient data
- office/administrative tasks which resemble use cases in other sectors
Most documented AI implementations reside in the former category, specifically within diagnostics. Medical image analysis is the most mature use case, radiology AI is running in selected regional hospitals, and audio-based triage analysis is partially deployed.
The latter category of use cases like predictive prevention, capacity optimisation, cross-entity knowledge sharing, etc., sits in small pilots and shadow initiatives that are not being scaled efficiently. Leveraging knowledge across healthcare and from other sectors, there seem to be substantial opportunities to capture within these impact areas.
The lack of broadscale AI implementation and adoption in healthcare does not seem to be held back by an absence of strategy in Denmark. On the contrary, strategic ambition exists at several levels; nationally, regionally, locally, and increasingly at European level. The challenge is that these efforts are not always connected into a clear implementation pathway, resulting in many pilots and initiatives that struggle to scale into everyday clinical and administrative practice.
The structural obstacles on the clinical side are concrete: records are fragmented, unstructured, and biased across systems and formats; staff have little room in the average workday to take on new tools; and the systems lack the legal and commercial competencies to scale AI built in-house. AI today acts mainly as an efficiency lever to ease workforce pressure rather than as mission critical infrastructure.
Trust transfer is non-trivial in a profession built on strong professionalism, and the question of who is responsible for the patient when AI is in the loop remains unresolved. Crucially, this liability cannot sit with individual doctors and nurses on top of an already busy day; it is a leadership responsibility to remove that risk from clinicians and to establish clear ownership of accountability before an AI ever makes an error. Workforce concern about redundancy meets a real risk of widening digital inequality between resourceful and less resourceful citizens, while AI-prepared patients arrive with self-generated diagnoses, treatment plans, and follow-up questions that time-pressured clinicians struggle to absorb.
AI is improving at a remarkable pace, and the potential to assist within a healthcare system under increasing pressure calls for efforts to bring the boardroom strategies into the everyday work of HCPs. Realising this at scale requires the benefits to be made tangible: leadership will only invest once the concrete gains of a given AI application are clearly demonstrated.
Healthcare's AI ambition is real, but fragmented data, thin capacity, and an overstretched workforce keep adoption from scaling
Current state of AI in the healthcare industry…
AI adoption in Danish healthcare rests on a strong foundation: high digital maturity, a deep health data asset base, sustained political attention, funding for pilots, and active collaboration across national, regional, and clinical actors. Strategy and ambition sit at multiple levels, both centrally and locally.
Adoption is therefore happening, but unevenly. Diagnostic imaging is the most advanced area, already in production at scale in Danish radiology, and AI-based triage support on the 1813 and 112 emergency lines is partially deployed. Speech to text scribes, predictive readmission models, and capacity optimisation tools are advancing through pilots. Most other activity still sits in small-scale experimentation rather than scaled rollout, with initiatives coming both from national and cross-regional flagship projects and from bottom-up clinical teams.
The data, governance structures, and talent within the workforce are largely in place to build and scale responsible AI in Denmark.
… but what is hindering adoption?
Although ambition and strategy exist at several levels, a gap to scaled implementation persists. Activity is spread from European to local levels, with unclear ownership and few pathways from pilot to production; pilots are rarely shut down but scaling them afterwards is where efforts stall.
The barrier profile differs sharply by use case. Administrative and logistics-oriented AI can move quickly with limited clinical risk and is consistently underused. Clinical support tools require workflow fit and trust before they scale. Diagnostic and treatment decision support carries the heaviest evidence, governance, and regulatory load, and moves slowest. A large part of the challenge is capacity: implementation power at department level is thin, IT functions are focused elsewhere, and the gains at scale are not yet visible enough to pull investment and attention.
Denmark's health data foundation is genuinely strong, built on shared national infrastructure and high citizen trust, but fragmentation across systems and ownership structures makes operationalising it for AI harder than the headline position suggests. Clinical scepticism is professional, not reflexive: imaging or screening AI that flags more cases is only useful if the system can absorb the follow-up, and high false-positive rates can swamp downstream capacity.
Regulation adds a further layer. GDPR, the EU AI Act, and the incoming EHDS raise the compliance bar, governance practice is still catching up, and informed consent is hard to maintain when patients have no working model of how AI reaches its outputs.
Future outlook of AI across the healthcare value chain
We see AI reshaping healthcare across the value chain, starting with operational and documentation efficiency and moving towards de-hospitalised care, predictive prevention, and AI-augmented clinical decision-making.
Highlighted future AI use cases in healthcare
The next wave of AI in healthcare carries the dual promise of releasing clinical capacity and reaching patients the system currently underserves.
Reference cases
Implement reference cases
The two cases illustrate how Implement Consulting Group has embedded AI into core healthcare and public sector processes with clients, delivering tangible results and meaningful business impact
Strengthening ward-round documentation with AI for a Danish public hospital
Clinical documentation // Healthcare AI
A cardiology department in a Danish public hospital struggled with a disproportionate documentation load on doctors, nurses, and medical secretaries, and wanted a path towards digital, AI-supported documentation of ward rounds and patient consultations.
Working through a multidisciplinary working group, Implement ran ward-round observations and workshops to clarify viable technical solutions, assess their fit to the documentation task, and surface the legal frameworks each would require. The output was a prioritised, ready-to-execute implementation plan.
Impact
- Four viable solutions surfaced through observation and co-creation; two selected for deeper work
- Legal prerequisites mapped per solution, removing the most common clinical AI rollout blocker
- Concrete implementation plan delivered, enabling immediate pilot start
Shared AI governance and a lighthouse pilot for a regional healthcare authority
AI governance // Generative AI
The Centre for HR and Education at a Danish regional healthcare authority needed a shared direction, governance structure, and responsible framework for AI.
Over the engagement, Implement upskilled 300+ employees across eight units through seminars, workshops, and hands-on training; established an AI ambition, assessment framework, and a prioritised use case pipeline refined from 66 ideas down to five minimum viable product (MVP) candidates; and built and user-tested a lighthouse pilot, RAG-based holiday-guidance chatbot for HR Service, to validate the move from pilot to operations.
Impact
- Shared AI vocabulary and competency baseline built across eight units
- MVP cut email response time on standard HR queries by ~50%, with 80% of answers ready-to-send after iteration two
- Reusable governance toolkit handed over with decision gates and RACI roles
Mission-driven AI adoption starting from healthcare's fundamental challenges, not the technology – five strategic choices must be considered
Future outlook
The future of healthcare is impossible to predict. But what is certain is that AI can no longer be parked as a curiosity – the pressures on the system are now larger than the barriers that have held adoption back.
Across healthcare, organisations are moving in a common direction. Not because they share a destination, but because cost pressure, workforce strain, and the need for de-hospitalisation are too sharp to ignore. The early leaders are not those with the longest roadmaps, but those who confront the organisational, regulatory, and trust barriers head-on rather than treating AI as a purely technical question.
In the near term, progress starts with pragmatic use cases like capacity and logistics optimisation, AI-assisted pre-patient interaction, voice-based interfaces for less digitally capable patients that serve as entry points to build governance, scaling muscle and clinician trust. The harder work lies inside the organisation: closing the gap between top-down lighthouses and bottom-up clinician energy, resolving accountability when AI is in the loop and designing rollouts so they do not widen digital inequality.
At Implement, we see early movers investing not only in technology, but in data foundations across many sources, legal competencies for MDR, GDPR, and the EU AI Act, and most importantly, in workforce transformation that reframes AI as enabling new roles rather than threatening existing ones. Lasting impact will depend on organisations' ability to move AI from pocket-deep maturity to system-wide capability as a mission aimed at solving the fundamental challenges of healthcare – and not driven by tech savvy and curiosity.
Reach out to our experts
Related0 4
Article
Read more
CFO Advisory #2: Lead finance through the uncomfortable middle
Set the direction. Design the approach. Protect the capacity to learn.Article
Read more
