Agentic AI, stochastic, deterministic approaches, the Matrix and Health and Life Sciences
A quick definition of Agentic AI
"Agentic AI" refers to artificial intelligence systems that exhibit agency, meaning they can act autonomously toward achieving goals rather than just following predefined instructions.
Hang on hang on…if after reading that you are thinking
Figure 1 The impactful Matrix movie...this is Neo the hero...an entire generation was named Neo as result of this futuristic decisioning engine now named generative AI incidentally
…that is about right, and we are aging ourselves. Moving on…
In other words, these systems are designed to:
- Set and pursue objectives: They can decide what actions to take to accomplish a goal.
- Adapt dynamically: They adjust their strategies based on changing environments or feedback.
- Demonstrate initiative: They don't just react; they proactively plan and execute tasks.
This concept is often discussed in the context of next-generation AI, where models go beyond passive tools (like chatbots) and become active agents capable of reasoning, planning, and interacting with other agents or humans in complex ecosystems.
Agentic AI is closely tied to areas like:
- Autonomous decision-making
- Multi-agent systems
- Goal-directed reasoning
- AI alignment and safety, since giving AI more autonomy raises ethical and control challenges
Whirlwind tour of Agentic AI relation to SAS Institute
For us here at SAS Agentic AI is just a fluid continuation of our core Viya strengths, Data and Math (oh and a bit more).
RELAX…..here is where I typically launch into a somewhat complex crowded diagram and its associated description so you can be amazed….but I won’t. Suffice it here to show you the architecture below to help you rest assured that we have great and well thought through components behind every single step that I summarize below. You can also go read all about it in this great post…
Figure 2 SAS Viya - unified, governed and decision-driven approach that enables the design and deployment of AI agents oat scale
SAS Agentic AI Offering (via SAS Viya)
- Hosts a dedicated solution—Agentic AI on SAS Viya—to build and deploy intelligent AI agents with trust, governance, and explainability. [sas.com], [reworked.co]
- Uses Intelligent Decisioning, combining deterministic analytics (see short discussion below re: deterministic analytics and stochastic generative AI) and LLM capabilities to produce reliable, auditable decisions. [sas.com], [prnewswire.com], [reworked.co]
- Ensures a balanced human–AI partnership, allowing organizations to customize autonomy levels and oversight based on task risk. [sas.com], [prnewswire.com], [reworked.co]
- Emphasizes built‑in governance and transparency, with lineage tracking and ethics guardrails essential in regulated industries. [sas.com], [prnewswire.com], [reworked.co]
Key Pillars of SAS Agentic AI
- Decisioning Engine – Ensures agents make accurate, explainable, and trusted decisions—avoiding “black box” behavior. [sas.com], [prnewswire.com], [reworked.co]
- Human–AI Balance – Users define when AI operates autonomously and when oversight is required. [sas.com], [prnewswire.com], [reworked.co]
- Governance & Compliance – Embedded traceability, auditing, and ethical oversight across the agent lifecycle. [sas.com], [prnewswire.com], [reworked.co]
SAS Agentic AI Accelerator
- A no-/low-code toolkit built on SAS Viya for rapid development and deployment of AI agents. [sassoftwar….github.io]
- Includes ready-to-use integrations, deployment recipes for LLMs, embedding models, and governance features. [sassoftwar….github.io]
Real‑World Use Cases
- Financial Services: Detect fraud rapidly while maintaining auditability and compliance. [sas.com], [prnewswire.com]
- Healthcare: Optimize patient outcomes and resource allocation with intelligent, governed decisions. [sas.com], [prnewswire.com]
- Retail: Enable personalized experiences at scale via automated analysis and action. [sas.com], [prnewswire.com]
We will be looking at agentic AI’s influence of healthcare a bit further…but first….
The balance between deterministic analytics and stochastic generative AI
This is a critical design principle in modern AI systems, especially in regulated domains like healthcare, finance, and life sciences.
Deterministic Analytics
- Nature: Rule-based, mathematically precise, and repeatable.
- Examples: Statistical models, optimization algorithms, business rules, and predictive models with fixed outputs for given inputs.
- Strengths:
- High accuracy and auditability.
- Easy to validate and govern.
- Essential for compliance and risk-sensitive decisions.
Stochastic Generative AI
- Nature: Probabilistic, driven by large language models (LLMs) or generative models that produce varied outputs based on learned patterns.
- Examples: Text generation, summarization, reasoning across unstructured data.
- Strengths:
- Handles ambiguity and complex reasoning.
- Generates insights from unstructured data.
- Enables natural language interaction and creativity.
The Balance
Hybrid Approach:
- Use deterministic analytics for core decision logic, compliance, and numerical precision.
- Use stochastic GenAI for interpretation, summarization, and contextual reasoning.
Workflow Integration:
- GenAI proposes options → deterministic models validate and enforce rules.
- Example: In healthcare, an LLM drafts a prior authorization summary, but deterministic rules check eligibility and compliance before approval.
Governance Layer:
- SAS and similar platforms embed traceability, lineage, and human-in-the-loop oversight to ensure stochastic outputs don’t override deterministic safeguards.
Figure 3 Balancing deterministic analytics and Stochastic GenAI
There will be as many opinions on this area as there are brains thinking about it.
I’ll leave it for now with this statement, more detail on this balance as this generative AI area matures.
Agentic AI transforming healthcare and life sciences
Clinical & Operational Applications
Healthcare
- Diagnostics & Early Warning Agents integrate EHRs, labs, vital signs, and imaging to detect conditions (e.g., sepsis) and deliver preliminary findings for clinician review. [healthcare…eaders.com], [healthtech…gazine.net], [mckinsey.com]
- Ambient Documentation & Prior Authorization Use ambient AI to streamline clinical notes and deploy agentic workflows for forms, validation, and communication in prior authorizations. [beckershos…review.com], [mckinsey.com]
- Decision Support & Care Coordination Agents assist with identifying care sites, appointment prep, and case management—assisting humans while maintaining oversight. [mckinsey.com]
Life Sciences
- Drug Discovery & Target Identification In labs, multiple agentic modules (data analysis, literature review, hypothesis generation, PI-level oversight) collaborate to accelerate research with traceability. [drugtargetreview.com]
- Clinical Trials & Regulatory Affairs Agents orchestrate recruitment logistics, monitor compliance, adapt study design, collect data, and prepare regulatory submissions. [ontoforce.com], [iqvia.com]
SAS Institute’s Role in Health & Life Sciences
- SAS Viya + Intelligent Decisioning Enables creation of trustworthy, governed AI agents across healthcare—automating resource allocation, care-cost optimization, medication adherence risk scoring, and clinical pathway analysis. [sas.com], [blstimes.com], [reworked.co]
- Agentic AI Accelerator A no/low‑code toolkit enabling rapid, compliant deployment with LLM integrations, governance, and full decision traceability—applicable to health and life sciences use cases. [sassoftwar….github.io], [sas.com]
- Governance & Human-in-the-Loop Control SAS emphasizes explainability, traceability, and adjustable autonomy, ensuring agents act under robust human oversight—critical in clinical and regulatory environments. [reworked.co], [sas.com], [healthaigo…e.duke.edu]
Key Benefits & Considerations
| Benefit | Description |
| Speed & Efficiency | Accelerates processes—from target discovery to clinical trials—using intelligent workflows [drugtargetreview.com], [iqvia.com] |
| Consistency & Accuracy | Reduces variability in diagnostics, imaging, and regulatory documentation [healthcare…eaders.com], [naviant.com] |
| Scalability | Operates across multiple departments, scaling without loss of oversight [sas.com], [mckinsey.com] |
| Compliance & Safety | Embeds auditability, safeguards, and human checks within every agent lifecycle [mckinsey.com], [sas.com], [reworked.co] |
| Initial Pilot Focus | Best used for well-defined tasks (e.g., prior auth). Expansion requires strong governance and data integrity [beckershos…review.com], [drugtargetreview.com], [iqvia.com] |
The future…What’s Next?
Agentic AI is advancing from pilot to production, underpinning:
- Self-driving labs and adaptive discovery tools that replicate scientific reasoning [drugtargetreview.com]
- Integrated trial management that adapts designs in real time based on data [ontoforce.com], [iqvia.com]
- Cost- and resource-based decisioning via SAS health analytics, including payer-provider AI workflows [blstimes.com], [aithority.com]
…more on this topic in future blogs….suffice it to say in the words of another film hero….
Figure 4 image from Toy Story movie and probably owned by them
Find more articles from SAS Global Enablement and Learning here.
