Part 2: Stop Limiting the Power of Your AI — Build with Rules Engines for Agentic AI

This blog is Part 2 of a three-part series building on the ideas explored in “Stop Limiting the Power of Your AI”.

At the HITEC Summit last year, I delivered a keynote on a topic: “Stop Limiting the Power of Your AI.”  Immediately after, I moderated a panel discussion expanding on the same theme—bringing together diverse perspectives on how organizations can unlock AI’s true potential. That’s where I first met Marinela Profi, Global Market Strategy Lead for AI and GenAI at SAS. What struck me during the panel was the clarity and conviction with which Marinela spoke about rules engines. Her perspective stayed with me. So, I reached out to continue the conversation—and this article captures the highlights of that exchange.

Why AI Needs Rules

Question: Marinela, tell us a little bit about your role at SAS and the types of AI work that you do? How did you become interested in the role of business rules engines in successful AI implementations? How do you align your business strategy with your AI solutions’ rules engines to ensure they make relevant recommendations?

Marinela:

At SAS I work to help organizations make better and trusted decisions at scale using data and AI. I’m passionate about integrating data, traditional analytics and large language models (LLMs) in a secure way to augment the power of data and AI workflows. That leads to increased productivity, cost efficiency and customer experience. Right now, I feel like AI is widely misunderstood as some magical technology that people are either wildly gung-ho for or terrified of. With the rapid advancement of LLMs becoming more integrated into existing business processes, organizations are essentially adding a probabilistic layer to their analysis, which is great at certain tasks, but is prone to hallucinations and can generate output that are factually incorrect. So how do we govern that? How do we ensure it aligns with business’ goals? That’s where business rules engines come into play. You have to keep in mind that AI is so much more than just technology. It’s about people, processes and technology. From a technology perspective, though, AI can only help us after we set up proper business rules.

Before AI can offer real help, executives must decide what they want it to accomplish, what their strategies are, and where things have been going right or wrong – all the same things they would tell a new employee if they want that person to get off on the right foot. Only after your business has done all that should you start building AI technology. By aligning your AI solutions' rules engines with business strategy, you can ensure that the recommendations made by AI are not only factually accurate but also truly aligned with your business objectives.

Where Leadership Comes In

Question: I strongly believe that ‘rules set based on company strategy’ can and should include high-level executive buy-in to ensure the AI tool is effective and aligns with the company's growth and development plans. What are your thoughts on this?

Marinela:

You're absolutely right. As I discussed previously executive buy-in is not just beneficial, it’s essential for the success of AI tools, especially those involving LLMs and decisioning systems. Executives must ensure that AI systems serve the company’s vision, not just their technical objectives. I always like to say that LLMs alone do not solve business problems. They make good talkers, incredible at conversation, summarization and text generation, but when it comes to high-risk decisions such as fraud detection, hiring, health care, etc., the real power comes from integrating strong analytical models and business rules. It’s important to let LLMs do what they do best: being the voice, but not the brain.

Executives also play a critical role in overseeing AI governance, ensuring the system’s rules reflect the company’s evolving priorities and risk management strategies. As decision-makers, executives should shape the rules that guide AI models, ensuring that they drive business outcomes and avoid mistakes like hallucinations in AI output. In short, executives are not just stakeholders; they are the architects of AI success. They can’t afford to be passive bystanders in the age of LLMs and decisioning tools. Executives must own the rules by establishing strategic frameworks, validating AI decisions and ensuring that the tools they deploy directly contribute to company growth and strategy.

What It Looks Like in Practice

Question: Can you describe scenarios where rules-based engines help companies solve difficult problems?

Marinela:

We’ve had a lot of success with rules engines, but off-hand, I’d say these use cases really stand out:

• Fraud detection: SAS has made huge strides with banks, insurance companies and other financial institutions in helping uncover fraudulent transactions. It’s a simple matter of setting rules that trigger alerts based on specific transaction patterns. Not every alert indicates fraud, of course, but just bringing suspicious activities to the top of the list means that companies aren’t having to spend time casting suspicion on every transaction every minute of every workday.

• Ethical guidelines and compliance: There are many scenarios where companies have sensitive or private information they must ensure is never released. We can set rules engines that ensure that none of those sensitive attributes are used in the decision-making process. That’s especially useful, for example, in the hiring process. Even more importantly, those rules engines can be continuously refined to make sure companies are always improving.

• Supply chain optimization for global retailers: A large global retailer faced challenges in optimizing its supply chain, leading to excess inventory, stockouts, and delayed deliveries. They implemented an AI-driven system that combined AI and LLMs with rules engines to enhance forecasting and automate restocking decisions. The rules engine was configured to adjust inventory levels based on these predictions, factoring in real-time shifts in demand or supply chain disruptions (e.g., supplier delays, transport bottlenecks). The system was designed to allow the AI agents to execute decisions, such as triggering restocks or reallocating inventory across regions without manual intervention.

What’s Coming Next

Question: What future development do you foresee for rules engines, and more broadly - the integration of strategy and AI in your organization?

Marinela:

Just to give an example of how essential rules engines are becoming, I think we have to talk about agentic AI. For those who may not be familiar with it, agentic AI is about having AI work autonomously to achieve goals, which means we’re giving it agency to make decisions and complete tasks, with a lower level of human involvement. If that seems too far in the future to think about today, consider Gartner’s top prediction in their 2025 Top Strategic Technology Trends: “By 2028, at least 15% of day-to-day work decisions will be made autonomously through agentic AI, up from 0% in 2024.” Anyone who understands the importance of rules engines today can see how it will grow exponentially when AI agents gain the authority and autonomy in enacting rules and decisions previously made only by humans – essentially raising the stakes of poor decision-making. It’s crucial that rules-based engines evolve to ensure humans remain an integral part of the process – providing oversight, setting boundaries and stepping in when necessary – especially in high-risk scenarios. The traditional, static nature of rules engines will no longer suffice in environments where AI is making decisions that can have profound consequences, such as financial trading, health care or safety-critical systems. A key component of this evolution is integrating accountability mechanisms into the rules engine, where AI decisions are traceable, auditable and subject to human review. AI should augment human decision-making, not replace it entirely. By evolving rules engines to allow humans to maintain strategic oversight and intervene when needed, we can ensure that AI acts responsibly, mitigating risks while benefiting from the speed and efficiency AI provides. This balance will safeguard against potentially dangerous blind spots and ensure accountability in the most critical decisions.

SAS and Microsoft are closely partnering to empower organizations to drive better decisions faster. We recently announced SAS Decision Builder, now available in private preview on Microsoft Fabric.  In addition, we’ve introduced SAS Viya Copilot, an AI-driven conversational assistant embedded into Viya to help developers, data scientists and business users with a personal assistant that accelerates analytical, business and industry tasks. I am very proud of the work that we are doing as partners and looking forward to continuing delivering value in the agentic AI era.

Final Thought

As I reflect on this conversation, my three takeaways continue to hold true:

• Effective and transformative AI strategies must have deep involvement from senior executives to meet their potential

• Agreeing on priorities for AI transformation as well as detailed strategic tradeoffs takes time – but is critical

• This strategic direction and alignment can then be codified into agentic AI solutions as business rules that support their reasoning, a valuable reference that keeps AI solutions delivering the intended business value

And as we move toward more agentic AI—where systems make decisions with greater autonomy—rules engines become even more important. We need to put as much time and focus into agent governance, thinking about the guardrails, and setting the rules engine as we do into building the models themselves. Rules engine help ensure AI operates within the boundaries of your strategy, values, and intent. That’s what makes AI not just powerful—but purposeful.