Since joining Beyond Limits in 2022, Chief Product Officer Jose Lazares has been instrumental in shaping and positioning the company’s AI-driven solutions, helping businesses optimize and automate operations, improve efficiency, and drive sustainability. But for Jose, AI isn’t just about cutting-edge technology, it’s about making AI practical, scalable, and truly valuable for enterprises.
In this interview, Jose shares his insights on the future of AI in enterprise environments, how Beyond Limits cuts through the AI noise, and why the company is finding success in some of the world’s most critical industries.
Our vision for AI at Beyond Limits has always been the same. How do we build AI solutions and products that truly help industries automate critical processes where precision, reliability, and real-time decision-making matter? That’s why we focus on hybrid AI. We take the best of generative, numerical AI, like machine learning based prediction and optimization, and combine it with cognitive reasoning-based AI that can actually provide intelligent, explainable decisions in real time.
Now, with generative AI making AI more accessible to end users, we see a huge opportunity to generate content, automate processes and enhance decision making. But there’s a challenge: in industries like energy, utilities, manufacturing and critical infrastructure, you can’t afford AI solutions that provide incorrect answers, are prone to hallucination or gives inconsistent answers. That’s why our focus is on blending generative AI’s ability to summarize and enhance productivity with our advanced reasoning systems built on subject matter expertise that ensure explainability and enhanced reliability.
The way I see it, AI in enterprise environments isn’t just about automating tasks, it’s about optimizing human expertise at scale. Our goal is to make AI a trusted partner that enhances decision-making, drives efficiency, and ultimately makes these industries more resilient and sustainable.
When I joined Beyond Limits, one of the key things that drew me in was the belief that the implicit knowledge held by subject matter experts needed to be captured and applied to AI-driven decision-making. Traditional AI models that rely solely on numeric machine learning aren’t enough to handle the complexity of real-world industrial environments, where conditions, equipment, and external factors are constantly changing.
What sets Beyond Limits apart is our hybrid AI approach, or what’s often referred to as neural-symbolic or composite AI as well. Essentially, we combine generative AI with machine learning and symbolic AI, which allows us to encode expert knowledge directly into our systems. This means our AI doesn’t just make predictions, it understands the reasoning behind decisions and can explain them in a way that’s transparent and auditable.
This is critical in industries where high-stakes decisions demand more than just a black-box AI model. If you're running a refinery, optimizing a power plant, or managing any highly regulated operation, you need AI that is not just accurate, but also accountable. Our approach – sometimes called neuro-symbolic AI – brings together data-driven models for prediction and optimization with advanced reasoning that reflects human expertise.
But we go beyond decision support. Our focus is on autonomous decision-making, where AI not only assists but actively reasons, reaches conclusions, and takes action. This requires an agent-based AI framework, which integrates multiple technologies, data sources, and systems to enable fully automated, end-to-end processes. That’s the difference Beyond Limits brings to the table: AI that doesn’t just predict – it understands, explains, and acts intelligently, just like a human expert would.
Traditional machine learning is great for making predictions, but it has limitations. It requires massive amounts of high-quality data, and if your data is incomplete or inconsistent, the results can be unreliable. Our hybrid approach fills in the gaps by encoding expert knowledge into the system, so it can still make intelligent decisions even when the data isn’t perfect.
We combine generative AI, numeric AI, and symbolic AI to create holistic enterprise solutions. This ensures our AI is not only predictive but also accessible, explainable and auditable, which is critical in industrial and mission-critical environments.
Neuro-symbolic AI isn’t new – in fact many companies use some form of symbolic AI, they don’t always talk about it. It’s easier to market AI as just machine learning or generative AI, but the reality is that in industrial settings, those approaches alone won’t suffice. You need AI that can reason, explain its decisions, and function within strict operational constraints. That’s exactly what we’re focused on, and it’s why our customers trust us to help them automate critical processes with confidence.
I recently read a paper discussing symbolic AI models designed for long-range planning and how some have attempted to replace these models entirely with generative AI. However, this approach has not been successful because generative AI lacks determinism and struggles to handle complex, structured decision-making.
While generative AI can perform reasonably well in short-term planning, it falls short when dealing with highly variable scenarios that require structured, rule-based decision-making. This is where a hybrid approach of combining generative and symbolic AI becomes essential for executing long-term planning tasks.
So while most AI solutions today incorporate some form of symbolic reasoning, the industry tends to emphasize generative AI as a catch-all solution. But for enterprise customers like ours seeking fully autonomous, closed-loop operations, generative AI alone is insufficient.
Let’s use the term “knowledge-graph RAG as an example of symbolic AI in action. Retrieval-augmented generation, or RAG, is a technique for enhancing the accuracy and reliability of generative AI models with information from specific and relevant data sources. Knowledge-graph RAG uses a knowledge graph to serve as the source of truth for RAG. So it describes how entities are related to one another. By using RAG on a knowledge graph, we can ground the RAG in known facts and answer complex questions
Here's a simple example. Amy is not only a director of human resources, but she’s also a mother, sister, daughter, etc. She's all these things and more. Each relationship represents an arrow from Amy to other people and things. For example, Amy's relationship with her daughter is that of being a mother. All these relationships are included together in a single diagram of relationships, which is the knowledge graph. It's symbolic because we use symbols, such as words like "mother," to represent things from the real world in AI.
So if I ask the system, is Amy a mother, it can answer it. And I can ask what Amy does for a living, and the system can give me that answer. But it can also answer more complex questions, like what is Amy's daughter's job. It can find Amy, traverse the knowledge graph to find her daughter, and then traverse again to find her daughter's job. This is symbolic AI. The system can use all of these data points to draw conclusions to simplify and make decisions based on the structure of the information – to show the power of symbolic AI to answer complex queries.
It’s incredibly powerful. And this hybrid approach, leveraging both neural and symbolic AI, ensures a more robust, explainable, and effective solution, particularly for industries requiring long-term planning and automation.
One of our long-term customers in the oil refining sector initially relied heavily on legacy solutions such as Advanced Process Control (APC) to make decisions on how to improve and optimize commercial objectives while ensuring efficient and sustainable operations. In looking at their refining operations, they still saw a significant opportunity to improve commercial performance because there were so many decisions being made that were outside the scope of APCs. When we started with them, our objective was to provide plant operators with real time process control and refinery-wide optimization to make better and faster decisions.
To meet these needs, we introduced our Hybrid AI Operations Advisor, which encoded their key objectives, asset constraints and operational best practices into a knowledge base. Our advanced reasoner could then leverage real time operational data to provide users with plant wide monitoring, deviations to plan, and set of prioritized recommendations and actions operators could take to optimize production or get an objective back on track.
Most importantly, we also provided a detailed audit of the reasoner decision making and rationale for the recommendation. Initially we started by focusing 50 key objectives and optimizing their efficiency by recommending adjustments every 15 minutes. To capture knowledge, we used a no-code rule-based system to enable process engineers to encode the reasoner with 50 optimization objectives, including supporting constraints that allowed them to improve plan adherence by over 20% in only a few short months. Of course, as the team saw value, they expanded the usage to over 350 objectives spanning yield maximization, plant utilization, hydrogen reduction, equipment reliability and sustainability objectives.
Additionally, as their understanding of the constraints and challenges became clearer, we incorporated a no-code ML model builder that enabled process engineers to build, deploy and maintain numeric prediction and process digital twin models to support specific high value assets. So now they can create and optimize processes on their own for specific cases that are of high value to them.
That combination has really made a difference in terms of the value that they've been creating. Plus, this process is creating a whole new unique set of information and knowledge. So, we extended that out to basically add in a full dashboarding and plan retrospective review using the underlying data that's been generated so managers and others can evaluate the plan more readily.
This is a great example of a multi-year continuous delivery of value over time. Not only have their optimization objectives grown, but they’ve seen their user base grow from 100 to over 1500 people, resulting in the elimination of an outside planning system that they were using.
Now, we’re taking it a step further with closed-loop automation, where the AI doesn’t just recommend actions – it directly adjusts operations in real time. That’s the future of AI in industrial settings.
One of the biggest challenges is knowing where to invest. Many struggle to define a clear AI roadmap because they lack visibility into the costs and expected ROI of different AI solutions. With so many potential problems AI can solve, the real question is: Which problems should organizations focus on to drive the highest value and the most transformative impact?
Over the past year, we’ve been helping clients build structured AI roadmaps that:
• Deliver measurable ROI, ensuring AI investments generate real business value.
• Automate or standardize key processes, creating efficiency and scalability.
• Solutions that evolve with the business, so they remain relevant as the organization grows.
Additionally, accuracy and reliability can be challenging. AI models, especially LLMs, can generate incorrect or misleading responses known as hallucinations. Organizations need strategies to manage accuracy, ensure contextual relevance, and minimize errors.
Ethical and regulatory risks are also an issue. Bias, data privacy, and regulatory compliance create barriers. Companies worry about exposing sensitive data or making AI-driven decisions that don’t align with ethical standards. These concerns can slow adoption or lead to overly restrictive policies that limit AI’s impact.
Lastly, skills and execution are a significant challenge for many enterprises, who lack the internal expertise to build, deploy, and maintain AI solutions. AI success requires technical talent, structured training, and governance models to scale efforts effectively.
To address these challenges, we’ve been working with customers to establish AI Centers of Excellence—leveraging our generative AI platform and hybrid AI capabilities t help them create a structured approach to rapid prototyping and roadmap standardization, and training and upskilling programs to develop AI expertise within organizations.
By focusing on strategic investment, skill development, and responsible AI governance, we’ve been able to help many businesses fully realize AI’s potential while mitigating risks.
At the core, it all comes down to efficiency—using resources smarter, not harder. That’s exactly where AI excels. Our solutions are designed to analyze vast amounts of data, recognize patterns, and make real-time decisions to keep operations running optimally while minimizing energy consumption.
The key to sustainability is producing more with less energy. By optimizing processes, businesses can lower emissions simply by running more efficiently. But it’s always a balance—maximizing output while maintaining energy efficiency requires smart trade-offs. AI helps find those optimal points, ensuring companies don’t waste resources while still achieving high performance.
For example, one of our ongoing projects is helping a facility reduce its energy consumption by 3% annually. That may not sound like much, but at a site generating hundreds of millions of dollars in output, the cost savings and emissions reduction are significant. And the savings aren’t just the actual cost, it’s doing the same processes with less energy and reducing their footprint, so it's sustainability through automation, through regulation and through standardization. That’s what we excel at. We’re not focused on tracking sustainability metrics, but we’re actually fundamentally addressing it at the root cause of energy consumption.
Follow Jose on LinkedIn HERE.
Want to know more about Beyond Limits? Contact us today.
