Professor Bamshad Mobasher explores the evolution of artificial intelligence and emerging trends that are transforming industries

From Foundation Models to Autonomous Agents: How AI is Reshaping Business Innovation

When most people think about artificial intelligence, they think about ChatGPT and the explosion of AI tools that emerged seemingly overnight in late 2022. But as Professor Bamshad Mobasher reminded attendees at an AI webinar, artificial intelligence has a much longer and richer history than many realize.

As director of both the Center for Web Intelligence and the DePaul AI Institute, Professor Mobasher has spent decades working at the intersection of AI research and practical business applications. In his presentation “AI As A Catalyst For Innovation: Key Trends And Practical Strategies,” he traced the journey from AI’s 70-year academic roots to today’s transformative technologies, offering Chicago business leaders a roadmap for understanding where AI is heading next.

Understanding the Paradigm Shift

The history of AI stretches back to pioneers like Alan Turing, who began contemplating machine intelligence even before general-purpose computers existed. For decades, AI researchers worked to develop systems that could reason, learn from observations, and adapt to new conditions. While the grand vision of artificial general intelligence captured public imagination through science fiction, a more practical vision took hold in industry: using AI to augment human abilities, automate tasks, and increase productivity.

This practical vision led to AI quietly powering many tools we’ve used for years, from web search engines to recommendation systems on e-commerce sites. Professor Mobasher explained that what changed recently wasn’t AI itself, but rather a fundamental paradigm shift in how AI models are built and deployed.

Traditional AI models were task-specific. If a bank wanted to predict credit risk, it would train a model on banking data specifically for that purpose. A different task required a different model and a different dataset. This approach, while effective, was limited in scalability and required significant human oversight for data labeling and preparation.

The breakthrough came with foundation models, large-scale models trained on vast amounts of diverse data from across the internet. Rather than being built for one specific task, these pre-trained models can be adapted or “fine-tuned” for countless downstream applications. This is why we’ve seen such a rapid proliferation of AI applications recently. Companies can now build on existing foundation models like GPT, Claude, or Gemini rather than starting from scratch.

The Rise of Agentic AI

While foundation models represent one major trend, Professor Mobasher focused considerable attention on what he called agentic AI, which he identified as perhaps the most significant emerging trend for business innovation.

The difference between current AI assistants and agentic AI is profound. When you interact with ChatGPT today, you provide a prompt and receive a response. It’s a single-step interaction. Agentic AI, by contrast, involves autonomous systems that can break down complex goals into multiple steps, use various tools, and execute extended workflows with minimal human intervention.

Think of it this way: instead of asking an AI to draft an email, you might tell an agentic system to research a topic, compile findings from multiple sources, draft a report, schedule a meeting with relevant stakeholders, and send personalized summaries to each participant. The agent would autonomously plan and execute each step, using search engines, databases, calendars, and email systems as needed.

This capability emerges from combining large language models with agent architectures that include perception modules to gather data, reasoning and planning systems to determine optimal actions, execution layers to interface with external tools, and learning mechanisms that improve performance over time through feedback.

According to research cited by Professor Mobasher, investors have committed approximately nine to ten billion dollars to AI agent startups just between 2023 and 2025. Major technology companies including Google, Microsoft, and Amazon have launched enterprise platforms specifically designed for building and deploying these autonomous agents. Some projections suggest the agentic AI market could grow to $200 billion by 2034.

Verticalization: Domain-Specific Intelligence

The second major trend Professor Mobasher highlighted is verticalization, the development of AI models trained specifically on domain-specific data rather than general internet content. While foundation models provide broad capabilities, vertical models combine smaller, specialized language models with proprietary data from specific industries like healthcare, finance, manufacturing, or education.

This combination of domain-specific models with retrieval augmented generation (a technique that allows models to access and incorporate relevant information from databases and documents) and agentic capabilities creates particularly powerful tools for industry applications.

In healthcare, for example, vertical models trained on medical publications, patient data, and sensor information from wearable devices can assist with diagnostics, develop personalized treatment plans, enable remote patient monitoring, and accelerate drug discovery. In finance, specialized models support algorithmic trading that accounts for complex market factors, real-time fraud detection, dynamic risk assessment, and automated portfolio rebalancing.

Real-World Impact: The Walmart Case Study

To illustrate how these trends translate into business value, Professor Mobasher highlighted Walmart’s innovative use of AI technologies. The retail giant has deployed proprietary generative AI systems to deliver what Mobasher called “hyper-personalized” shopping experiences. Unlike traditional recommendation systems that might suggest products based on broad categories, these newer systems consider numerous characteristics across diverse product categories, creating connections that might not be immediately obvious but are relevant to individual shoppers.

Beyond personalization, Walmart uses AI for inventory optimization, automated replenishment through real-time demand sensing, and dynamic pricing strategies. The company tracks performance through metrics including stock-out rates, basket sizes, promotional effectiveness, and markdown reductions. These AI-driven innovations directly enhance revenue growth while strengthening customer loyalty.

Sector Transformations

Professor Mobasher outlined how agentic AI is reshaping several key sectors beyond retail. In fintech, autonomous agents now perform algorithmic trading with greater sophistication, detecting fraud in real time, conducting dynamic risk assessments, and rebalancing investment portfolios based on complex signals from both user behavior and market trends. Some automated funds using these technologies have consistently outperformed larger non-automated competitors.

Customer service represents another sector where many of us encounter these technologies regularly. While chatbots have existed for years, they’ve often been frustratingly ineffective. Professor Mobasher acknowledged this reality but noted that agentic AI is changing the landscape. Modern virtual assistants like Google Duplex can autonomously handle complex tasks including appointment booking, provide proactive support by anticipating customer needs, coordinate seamlessly across multiple channels from email to voice to web interfaces, and deliver genuinely personalized responses and recommendations.

Managing Risk Responsibly

After painting an optimistic picture of AI’s potential, Professor Mobasher shifted to address what he called the serious risks that accompany these powerful technologies. He emphasized that as AI systems become more capable, organizations must implement key strategies to mitigate potential harms.

Transparency and explainability present significant challenges. Many AI models, particularly those based on deep learning architectures, function as black boxes. Even the engineers who develop them cannot always explain exactly how they arrive at specific outputs. This problem compounds with agentic AI, where autonomous systems execute sequences of tasks, each step potentially opaque. Organizations need to build in mechanisms that can explain the rationale behind autonomous decisions.

Bias mitigation and fairness require careful attention throughout the development process. Because these models learn from data, any biases present in training datasets will propagate to the model’s outputs and decisions. This requires both careful data curation and algorithmic approaches that explicitly account for fairness in system design.

Data privacy protection becomes especially critical in domains handling sensitive information like healthcare or finance. Accountability frameworks establishing clear legal and ethical guidelines must address who bears responsibility when autonomous agents make decisions or take actions that cause harm. Professor Mobasher emphasized that accountability is essential for building trust, both within organizations and with end users and customers.

Professor Mobasher noted that while many organizations recognize these risks, implementation of robust mitigation strategies remains inconsistent. In the United States particularly, regulatory mechanisms to ensure responsible AI development and deployment remain limited, though he pointed to the European Union’s AI Act as an example of more comprehensive governance, while acknowledging concerns that excessive regulation might stifle innovation. Finding the right balance between preventing egregious harms and supporting beneficial applications represents one of the central challenges facing the AI community.

The Path Forward

The presentation raised important questions about how AI will reshape work and employment. Professor Mobasher acknowledged that certain jobs are indeed vulnerable to automation, as we’re already witnessing across various industries. However, he suggested that this pattern reflects broader technological advancement rather than something unique to AI, drawing parallels to previous waves of automation.

Rather than eliminating entire job categories, AI often transforms how work gets done. Professor Mobasher used computer programming as an example. Where students once spent hours learning syntax and debugging basic code, AI tools now handle much of that low-level work more effectively than most beginners. This shift allows programmers to tackle bigger, more complex problems by having AI handle routine coding tasks, potentially making the profession more creative and strategic rather than simply eliminating it.

Looking Ahead

For Chicago business leaders attending the session, Professor Mobasher’s message was clear: AI represents not a distant future but a present reality that’s already transforming competitive landscapes across industries. The combination of foundation models, agentic capabilities, and domain-specific verticalization creates unprecedented opportunities for innovation, automation, and value creation.

At the same time, responsible deployment requires thoughtful attention to transparency, fairness, privacy, and accountability. Organizations that successfully balance innovation with ethical considerations will be best positioned to harness AI’s transformative potential while building trust with customers and stakeholders.

The DePaul AI Institute, under Professor Mobasher’s direction, continues to explore both the promises and challenges presented by artificial intelligence through interdisciplinary research, educational programs, and partnerships with industry and nonprofit organizations.

As Professor Mobasher’s presentation demonstrated, understanding AI requires looking beyond the headlines about artificial general intelligence to grasp the practical, powerful technologies already reshaping how businesses operate, compete, and serve their customers. For Chicago’s business community, staying informed about these trends isn’t just about keeping pace with technology – it’s about recognizing and seizing opportunities to innovate in an AI-enabled world.