The world of artificial intelligence is evolving at breakneck speed, and nowhere is this more conspicuous than in the emergence of AI agents. As organizations grapple with separating genuine innovation from marketing hype, we sat down with Ed Keisling, Chief AI Officer at Progress Software, to cut through the noise and understand what AI agents really mean for businesses today. Ed brings a unique perspective, having taken on his new role in February 2025 at a time when the industry is proclaiming this as “the year of agents.” His insights reveal both the tremendous potential and the current limitations of this transformative technology. As always, time is of the essence.

AI Agents, Beyond the Hype

The Rise of the Chief AI Officer: A Strategic Imperative

The creation of Chief AI Officer roles across the technology industry signals more than just a trend—it represents a fundamental shift in how businesses view artificial intelligence. As Ed explains, “AI needs to be a strategic pillar of a business to drive innovation and growth. It really reflects the pace at which technology is evolving, and having somebody that is accountable to follow all these latest updates and really look at it through the lens of new risks and opportunities.”

This observation resonates with the broader digital transformation patterns we’ve witnessed over the past decade. Just as Chief Digital Officers emerged to guide organizations through the digital transformation revolution, Chief AI Officers are now stepping up to navigate the AI transformation. The role isn’t merely about implementing technology—it’s about strategic thinking, risk assessment, and identifying genuine business opportunities in a rapidly changing landscape.

Defining AI Agents: Beyond the Buzzwords

One of the most persistent challenges in the AI space is the confusion surrounding terminology. AI agents, in particular, have become an overloaded term that means different things to different people. Ed provides valuable clarity by positioning agents on a spectrum of AI capabilities.

“When generative AI came out, it was generally reactive,” Ed notes. “We would go to ChatGPT, provide a prompt, and it would generate a response based on its training patterns. Agents are moving along that spectrum in terms of capabilities—they have the ability to perceive their environment, access to audio, video, documents, and crucially, the ability to reason, plan, and learn from their actions.”

<figure class="wp-block-image size-large"><figcaption class="wp-element-caption">Unfortunately, Rover isn’t always willing to search in the right direction… — photo by antimuseum.com</figcaption></figure>

Traditional automation relies on strict rule-based systems—the digital equivalent of if-then-else logic. Chatbots, while more sophisticated, remain predominantly reactive. AI agents represent a step toward proactive, reasoning systems that can adapt to changing circumstances.

AI agents represent a step toward proactive, reasoning systems that can adapt to changing circumstances

The evolution doesn’t stop there. Ed introduces the concept of “agentic AI“—a broader paradigm where multiple agents collaborate, passing context between each other to accomplish complex tasks. This represents the holy grail of AI automation: systems that can dynamically adapt to real-time situations without constant human intervention.

Reality Check: Why Perfect Automation Remains Elusive

Despite the exciting potential, Ed provides a sobering reality check about current capabilities. His observation about the Pareto principle in AI is particularly insightful: “AI is the ultimate manifestation of the 80/20 rule. You can very rapidly get to value with 20% of the work achieving 80% of the results, but actually getting it to work 100% of the time is still very, very difficult.”

This phenomenon explains why AI demonstrations look so compelling while real-world implementations often fall short of expectations. The gap between proof-of-concept and production-ready systems remains significant, requiring careful planning, clean data, and well-defined business processes. As always, I should add, “the more it changes, the more it stays the same,”as the French poet would have it.

RAG Technology: Making AI Practical for Business

While pure AI agents may still be evolving, Progress Software’s acquisition of Nuclia, an agentic RAG (Retrieval Augmented Generation) provider, demonstrates a more immediate and practical application of AI technology. Ed explains the fundamental problem RAG solves: “Large language models have been trained on the entirety of the Internet, giving them broad general knowledge, but they don’t have access to data stored behind firewalls or on personal computers.”

This limitation is critical for businesses. While public AI models are impressive, their real value emerges when they can access and reason about proprietary business data. RAG technology bridges this gap, allowing organizations to leverage AI’s reasoning capabilities while grounding responses in their specific knowledge base.

The practical implications are significant. As Ed points out, “Small to medium-sized businesses have lots of unstructured data—audio, video, log files, recordings, PDFs, charts—that represent proprietary business value, but they have no way of indexing or finding or correlating the data within it.” RAG technology makes this data accessible and actionable.

<figure class="wp-block-image size-large"><figcaption class="wp-element-caption">It’s high time to stop AI-Washing Ed Keisling advises — image created with Midjourney</figcaption></figure>

Separating Innovation from AI Washing

Ed’s experience at the AI4 conference provides valuable insights into the current state of the AI industry. His observation about AI washing is particularly relevant: “There was a lot of AI washing—companies that weren’t sure they understood the problem to be solved, with very thin wrappers around language models to solve point problems. It felt like a hammer looking for a nail.”

The key differentiator, according to Ed, lies in problem-solving focus rather than technology-first thinking. “AI allows you to solve old problems in a new way and to make seemingly impossible problems possible. You’re thinking about how to drive outcomes—making developers more productive, automating tedious workflows, providing better insights that weren’t possible before.”

This perspective aligns with successful technology adoption patterns throughout history. The most successful implementations focus on specific business outcomes rather than showcasing technological capabilities.

Real-World Value: ShareFile’s Document Intelligence

Progress Software’s ShareFile platform provides concrete examples of AI delivering measurable business value. The platform serves client-facing teams in regulated industries—doctor’s offices, law firms, and tax accountants—where document management is critical but time-consuming.

The AI implementations are practical and measurable: “We can create curated lists of documents appropriate based on your situation, and as you upload documents, we can figure out which document relates to which checklist item. We’ve measured this at being three and a half times faster.”

More importantly, th