Market research firm Gartner recently released some eye-opening predictions about agentic AI:
- By 2028, 33% of enterprise software applications will use agentic AI
- By 2028, at least one-third of interactions with GenAI will invoke autonomous agents to complete tasks
- By 2028, at least 15% of day-to-day work decisions will be made autonomously through agentic AI, up from 0% in 2024.
According to its assessment, Garner expects explosive growth in agentic AI within the next three years. What does that mean for health insurers? Let’s take a look:
What is agentic AI?
Agentic AI refers to a class of artificial intelligence (AI) systems designed to act autonomously to achieve specific goals, often with minimal human intervention. Unlike traditional AI that primarily reacts to commands or follows rules, agentic AI exhibits a degree of “agency” – the ability to perceive, reason, plan, and execute actions independently within a dynamic environment.
What are some agentic AI use cases?
Consider processes that take a significant amount of time or create bottlenecks. Those are potential agentic AI opportunities. Also, don’t be afraid to think big. Remember that agentic AI has agency to make decisions. Instead of just looking at “claims processing,” map the entire claims lifecycle from submission to payment. Identify every handoff, every manual review, every point of delay, and every instance of human cognitive load.
Agentic AI works best when it augments human effort. Think about instances where employees access multiple disparate systems, copy-paste information, or manually reconcile data. Or, where are employees spending significant time on repetitive data analysis, information retrieval, or routine decision-making that distracts from core responsibilities or leads to burnout? These are prime candidates for agentic AI that can orchestrate data flow across systems.
Examples may include using an agentic AI to prepare a comprehensive member profile for a care manager, summarize complex medical records for an underwriter, or triage customer inquiries to ensure a human agent reviews the most urgent ones immediately.
If you can quantify the pain, do so. For each bottleneck, estimate the volume, time spent, error rate, and associated costs. High-volume, high-error, or high-cost areas are strong indicators.
Additionally, identify areas requiring proactive, adaptive intervention. Because agentic AI can anticipate issues, it can consider scenarios where early intervention could prevent a larger problem. Examples include predicting members at risk of non-payment before a premium is due, identifying potential care gaps before they become chronic conditions, flagging suspicious activity before a fraudulent claim is fully processed.
Finally, focus on use cases with measurable outcomes. Be sure to define success metrics upfront. Whether that’s reduced processing time, lower denial rates, increased member retention, or improved compliance scores, make sure you can measure the effectiveness of your agentic AI use cases.
How should I get started with agentic AI?
The first step is to get everyone on the team on the same page, including leadership and front-line staff. Start by educating your team on the benefits and drawbacks as well as how agentic AI differs from automation or generative AI that you may already be using.
Then, build internal expertise. Encourage key personnel to learn about agentic AI concepts, capabilities, and ethics. Foundational understanding can help everyone better understand how to leverage agentic AI.
Then, using the use cases you identified, find one that has a specific, well-defined outcome and use it to pilot the use of agentic AI. Because agentic AI thrives in environments where tasks are predictable but require some adaptive decision-making, look for repetitive, rule-heavy processes for your pilot. It’s also best to focus on agentic AI’s ability to augment the work human workers do. You’ll get more buy-in by explaining how agentic AI enables workers to do more productive work.
You should also assess if your data is ready for agentic AI. You’ll likely need high-quality, accessible data. If you don’t have clean, structured, and readily available data, you may be less successful. Additionally, it’s always important to establish clear policies for data collection, data storage, security, and privacy.
The team you choose to implement agentic AI also matters. Agentic AI shouldn’t be an IT initiative, but a cross-functional team that includes operations, compliance, technology, data science, and other teams. Before the pilot, ensure you have an AI governance framework in place. Organizations should address accountability, ethical guidelines, and bias detection before piloting any AI. If anyone in your organization has an understanding of existing or emerging AI regulations, they need to be involved in your governance process.
Finally, tools and partners matter. Discover which agentic AI platforms and tools best suit your use case. Consider industry-specific platforms instead of general tools. Cloud-based solutions may also offer better scalability. If you lack in-house knowledge, don’t be afraid to leverage vendors or consultants with a track record of responsible AI deployment.
How will agentic AI integrate with our existing legacy systems and data silos?
Agentic AI is more like an employee that interacts with your existing technology stack and data than it is an entity that requires unique technology. It should be able to sit on top of your technology infrastructure. The goal is to leverage existing data and processes while adding a new dimension of goal-oriented automation.
So what’s required? Agentic AI systems typically use APIs to send data to and securely retrieve data from various health plan systems. Systems include core administration, enrollment, premium billing solutions, EHRs, CRMs, provider directories, and other data sources.
If APIs don’t exist, agents may need to interact with middleware or integration platforms designed to translate data. Alternatively, some agentic AI solutions can use robotic process automation (RPA) to extract or input data in the user interfaces of legacy systems.
What should I know about AI governance?
Earlier this year, we discussed Health and Human Services’ AI Strategy Plan, which can provide a framework for health plans aiming to implement an AI strategy. But at its core, AI governance defines accountability, establishes guardrails and boundaries, prioritizes transparency, and includes clear points for human review.
Articulate clear data privacy and data security measures. Because agentic AI has autonomous access to data, you’ll need strong access controls and must adhere to privacy regulations, like HIPAA. Also, AI can hallucinate, so actively monitor and audit for accuracy as well as adherence to ethical guidelines.
How do we measure the performance and ongoing effectiveness of agentic AI solutions?
Measuring the performance of an agentic AI can be similar to measuring an employee’s performance. You’ll have a mix of quantitative metrics, qualitative metrics, and operational metrics.
Quantitative metrics may include measures like a task success rate (percentage of prior authorization requests processed), efficiency metrics (reducing claims processing time by x%), and accuracy metrics (percentage of accurately coded claims).
Qualitative metrics may include measures of user or member satisfaction or feedback from humans overseeing the agentic AI. Operational metrics measure the impact the agentic AI has on operations. Process adherence (percentage of grace period notices sent on time), workload redistribution (reduction in administrative tasks), and auditability (can the agentic AI’s decisions be traced and explained?) would be examples of operational metrics.
Measuring effectiveness is not a one-time event. It’s an ongoing process that might involve real-time dashboards, automated alerts, A/B tests, regular audits, and an integration of feedback from users, audits, or performance data. By combining these quantitative, qualitative, and operational metrics with a continuous monitoring strategy, health plans can effectively measure the performance and ongoing effectiveness of their agentic AI solutions, ensuring they deliver value.
Certifi’s health insurance premium billing and payment solutions help healthcare payers improve member satisfaction while reducing administrative costs.
