As 2026 catapults toward the final quarter of the year, the pressure is growing to turn AI ambitions into tangible results. Throughout August, The AI Forecast explored what that looks like in practice.
Through conversations between host Paul Muller and guests Mark Ritcey, Varun Puri, and Shanea Leven, The AI Forecast examined the realities of putting AI to work across the enterprise. Read on for three perspectives on what it takes to turn AI’s potential into meaningful business value.
1. Start with behavior and outcomes over the technology
For Varun Puri, co-founder and CEO of Yoodli, successful AI adoption starts with the problem people are trying to solve and the behavior they want to change.
“I often say we are not an AI company. In fact, I don't even care if AI is a solution. We are a behavior change company,” Puri said. “The reason people struggle with the conversations that matter is that they don't practice. The number one way to get better is to record yourself and watch yourself and cringe and do that three more times.”
In this context, AI becomes a means of making that practice easier. It can surface specific areas for improvement and allow people to practice repeatedly. “AI gives you interesting insights. It gives org-level statistics, but the core is practice and being more self-aware,” Puri said.
That same outcome-oriented thinking can help organizations evaluate ROI. Puri offered a hypothetical example: “Say I'm Google. I need to train 15,000 reps on the new Gemini pitch. In the old world, it would take me six months to get everyone talking in the same language. Now it takes me six days. Great, we have a reduced ramp,” he said. “Is that a direct dollar value? Not yet, but my gosh, we can quantify that very, very quickly.”
For Puri, the lesson is to define the outcome first and evaluate AI according to whether it helps achieve it. That could mean greater productivity or stronger communication. Technology matters most when it serves the goal rather than becoming the goal itself.
2. Enterprise AI success starts with discipline
With organizations pouring resources into AI, the pressure to move quickly is intense. But for Mark Ritcey, vice president of AI and automation delivery at LatentBridge, successful enterprise AI starts with the simple first step of defining the business problem and building the right structure around it.
“I would say that I have certainly seen, experienced, and delivered successful AI use cases,” Ritcey said. “And they're in many different areas. A common one is an AI- or LLM-powered chatbot that answers HR-related questions based on an HR knowledge repository. There are credit assessment tools that analyze a customer's or company's financial information, aggregating and summarizing it for presentation to a credit analyst. So it's not the machine or AI that makes the decision; it's actually a human.”
Across these examples, Ritcey sees the same fundamentals determining whether an AI initiative delivers value. Organizations need to start with a clearly defined problem, then build the structure needed to turn an AI initiative into something employees can actually use successfully.
That becomes even more important as organizations move from generative AI experiments toward production systems and agentic AI. Ritcey's advice is to resist the pressure to chase the biggest opportunity immediately and instead build the capabilities required to scale.
As Ritcey put it, “You must crawl before you can run.” Enterprise AI requires the same level of rigor as any major business transformation, with an added layer of governance for systems whose outputs can be unpredictable. That degree of work can’t be rushed.
3. When everyone can build software, guardrails become essential
Generative AI is dramatically lowering the barrier to software development, allowing people with little traditional coding experience to turn ideas into working applications. But for Shanea Leven, co-founder of Empromptu AI, the ability to build quickly also raises the stakes around how those applications reach production.
“You can absolutely build production apps. You just need to know what you're doing,” Leven said.
The difficulty lies in the infrastructure surrounding those applications. “If you're building a generative application and you don't have AI in the generative application infrastructure, like drift or edge-case detection, are you monitoring how inputs are working, and are they getting the right inputs?” Leven asked. “If you don't have all of those things set up, that's how you start to get hallucinations.”
With that infrastructure in place, however, AI can broaden who participates in software development. Take the hypothetical “Debbie in accounting,” for example.
“Debbie in accounting is a subject matter expert in accounting,” Leven said. “And if you're putting an app or you're putting a workflow in Debbie's world, Debbie is the best person to know what those outputs are supposed to look like, what those guardrails can be.”
Leven calls this growing group of AI-enabled builders “newly technical.” As more employees gain the ability to create software, she believes organizations will need to help them develop some of the thinking traditionally associated with software engineers and product managers.
Expanding access to software development can unlock valuable expertise across an organization, but greater access also increases the importance of shared infrastructure and education. Giving more employees the power to build works best when organizations also provide guardrails for responsible development.
Turning AI potential into business value
AI is maturing, and success increasingly depends on how organizations use the technology. This month, Ritcey, Puri, and Leven explored what it takes to turn AI’s potential into meaningful enterprise value. Their perspectives point toward a more deliberate approach to realizing meaningful business value from AI.
Listen to these episodes of The AI Forecast and more on Spotify, Apple Podcasts, and YouTube.
