By Guest contributor |
Two years ago, we were just starting to talk and learn about AI agents. We wondered, what can they do? What should we enable them to do? Are agents appropriate to use when patients’ health is on the line? We now know the answers to these questions as the technology continues to evolve. But instead of sharing generalities, we can answer these questions through innovative work done by startups with their enterprise customers.
Atropos Health, Autonomize AI, IgniteData and Raapid are reshaping the sector and setting new standards for innovation by using AI agents.
Agentic AI refers to AI systems that possess a degree of autonomy: the ability to make decisions, take actions and adapt dynamically to changing scenarios with minimal human intervention. Unlike traditional rule-based or supervised AI, agentic systems operate with greater autonomy, learning from data, context and outcomes to continuously improve their performance. In healthcare, this translates into AI tools that analyse information and then can recommend treatments, optimise workflows and even initiate interventions, all while responding to real-time changes in patient status or operational needs. The relevance of agentic AI in healthcare is profound, enabling more proactive, personalised and efficient care delivery, while addressing challenges in resource allocation, decision support and patient engagement.
Atropos Health exemplifies the power of agentic AI in generating actionable clinical insights from vast datasets. The company’s platform deploys intelligent agents that autonomously query medical literature, patient records and real-world evidence to answer complex clinical questions. These agents synthesise evidence, evaluate what information is most relevant to answer each particular question, and, if existing literature does not answer the questions, allow for new evidence generation – all within the clinician’s workflow. By transforming data into actionable evidence, Atropos Health enables faster, more informed choices that improve patient outcomes and reduce administrative burden. The continuous learning capabilities of the system ensure that care decisions are based on evidence, supporting a culture of personalised precision medicine.
Meanwhile, Autonomize AI helps healthcare enterprises become AI-native organisations by putting healthcare-ready agents to work across the business of care. The company’s more than 100 configurable agents and ready-to-use workflows pull together clinical and administrative information from unstructured data, organise it in real time, and support teams with the context they need to make faster, more reliable decisions. Delivering value across healthcare enterprises, including three of the five largest health plans in the USA, clients begin seeing value on day one and typically reach three to five times return on investment within the first year, along with meaningful reductions in administrative cost, burden, and improved experiences for patients and providers.
IgniteData deploys agentic AI to automatically capture data from clinical trials. These agents ensure seamless, error-free, human-in-the-loop data exchange between healthcare providers, research institutions and trial sponsors, dramatically accelerating research timelines and enhancing patient safety monitoring. The reduced manual effort and real-time synchronisation improve the quality of patient care during studies and delivers financial efficiencies for hospitals by minimising redundant documentation and compliance risks.
Agentic AI can also be used to manage healthcare risk adjustment with minimal human intervention. Unlike AI-assisted tools requiring constant oversight, Raapid’s specialised agents work independently, identifying HCC (Hierarchical Condition Category) codes, validating MEAT (Monitor, Evaluate, Assess/Address, Treat) evidence, checking compliance risks and citing supporting documentation.
Raapid’s technology uses neuro-symbolic AI, which integrates neural networks for pattern detection with symbolic reasoning for powerful insights with clear explanations. Multiple agents collaborate through what Raapid terms “agent orchestration”, reaching consensus on coding decisions. In production environments, organisations reduce chart review time from 30-40 minutes to under eight minutes – a five times productivity gain – while maintaining higher than 98 per cent coding accuracy. Each recommendation includes traceable audit trails linking to source clinical evidence, addressing RADV (Risk Adjustment Data Validation) audit defence requirements.
As an M12 (Microsoft venture fund) backed company with HITRUST certification, Raapid provides deployment within organisations’ existing Microsoft Azure environments, ensuring data sovereignty. The system integrates with current EHR (Electronic Health Record) infrastructures enabling continuous learning from human coder feedback.
These Microsoft for Startups companies share several common themes that reflect broader trends in agentic AI for healthcare and life sciences. First, there is a clear emphasis on interoperability and integration, ensuring that AI systems work harmoniously. Second, explainability and transparency are prioritised to foster trust among clinicians and patients. Third, these startups demonstrate the value of continuous learning, with AI agents that adapt to new data and clinical guidelines over time. Looking ahead, agentic AI is poised to drive further personalisation of care, streamline research and operational processes, and support population health initiatives. As regulatory frameworks evolve and adoption accelerates, these innovations will likely become standard practice across the sector.
Sally Ann Frank is global lead for health and life sciences at Microsoft for Startups
Discover more insights from these partners and others, in the Winter 2025 issue of Technology Record. Don’t miss out – subscribe for free today and get future issues delivered straight to your inbox.