.svg)
Enterprise AI adoption isn't happening as fast as leaders expected. For most, the greatest hurdle has been integrating new models with legacy systems, disparate platforms, and convoluted workflows. Now, the limitless possibilities of AI are starting to paralyze the very people it was designed to empower.
CIO News spoke to Dave Kreitter, CEO and Automation Consultant at Grail Automation, to learn more. With a career as a solutions and technical architect for tech giants like HubSpot, Broadcom, and Workato, he's had a front-row seat to significant technological shifts, including as an individual contributor to Marketo's IPO and its eventual $4.75B acquisition by Adobe. For Kreitter, the clear solution is a new mindset, one with unwavering commitment from the top.
Swinging and missing: While acknowledging that the early, 'hit-and-miss' phase of AI exploration can be frustrating for leaders, Kreitter also framed it as a prerequisite for success. A period of trial and error is the only way to build institutional knowledge, he said. "When you start using AI, you'll find a lot of it yields mediocre results. But as you start to land more successful solutions, your home run rate increases. You have to go through the gray area to get to the black."
Effective adoption also requires absolute conviction from the C-suite, Kreitter said. "If leadership expresses hesitancy, the whole team feels it. You have to be all in." Fortunately, the technical fix is just as direct: move beyond classic APIs to Modern Connectivity Platforms (MCPs), an orchestration layer that sits on top of existing APIs.
One tool to wield them all: The goal is custom, multi-system workflows, Kreitter explained. Rather than giving it access to hundreds of small tools, he recommends building one powerful tool for AI to wield instead. "The real value is going to be in exposing tools to these LLMs that actually execute complex workflows across multiple systems."
Kreitter recently put this theory into practice for a client looking to automate workflows across banking partners after the SVB collapse. The lack of off-the-shelf connectors was a challenge, he admits. But the most significant bottleneck was the level of specialized expertise needed to build new ones. The task was too complex for even an AI to handle alone.
Lost in translation: Even though AI is a powerful assistant, it often fails when confronted with the niche syntax of specialized platforms, Kreitter said. "These domain-specific languages (DSLs) have so many nuances. Getting the syntax perfect is next to impossible for an LLM without serious guidance."
Chain of command: The solution is to manage the AI's process, according to Kreitter. Start by breaking down the enormous task into a series of smaller, sequential steps, he recommends. "You have to organize its chain of thought. An action block is enormous, so you have to guide it action by action."
But if AI can assist in building these solutions, does that make structured platforms obsolete? Not at all, argued Kreitter. Instead, their value is simply changing.
From creation to curation: The shift changes their core value from creation to governance. "It's not about creating solutions anymore. It's about how you organize them and enable people to edit them in the future."
Kreitter's final mandate for leaders was simple: get your hands dirty. For busy executives, that doesn't mean boiling the ocean, he clarifies. "Rather than trying to explore 10 different opportunities, pick one or two and go deep."
Get your hands dirty: A hands-on approach is the key to understanding the next frontier, where AI's role evolves from following instructions to generating new user interfaces, workflows, and even entire digital environments on the fly. In the near future, direct, personal engagement will be a matter of strategic survival for any leader. "Unless you get your hands dirty, you're going to miss out on these insights."
The insights needed to win the next wave won't be found in reports, Kreitter concluded. They'll be found in the work. Enterprises willing to navigate the gray area will move beyond scattershot proofs of concept and into the scalable, strategic impact that has, until now, remained just out of reach.
