The Enterprise AI Inflection Point with Kevin Surace

The 95% problem with AI pilots

Watch Kevin Suraces's full Face Value session on-demand

Boardrooms are starting to process an uncomfortable number: over the last two years, 95% of enterprise AI pilots have delivered no measurable P&L impact.

We’ve moved past the "gee-whiz" phase of generative AI, according to Face Value guest and CEO at Appvance.ai, Kevin Surace. The question, he says, is no longer whether AI works (it does), but why, after all the investment and noise, so few deployments have changed the bottom line. If you're stuck between "science project" and real-world value, you're in the majority. The problem is that the majority is losing money.

Kevin — a pioneer in virtual assistant technology, a multi-time founder, and author of the forthcoming book The Joy Success Cycle — joined a recent Face Value session to discuss why most AI deployments fall short. He went beyond the technology to get to the operational, cultural, and architectural shifts that separate the top 5% of successful pilots from the 95% that fail.

Here's what came out of that conversation.

The "expensive autocomplete" trap that causes your pilots to fail

Kevin says the technology isn't the problem. Today's frontier models are extraordinarily capable. When AI doesn't work, it's rarely because the model is dumb. It's because the model is disconnected.

Most companies deploy AI agents into a vacuum. They wire a powerful model to a system of record and expect magic. What they get is what Kevin calls "expensive autocomplete." An agent deployed without context, with no understanding of how the business actually runs, produces brittle, disconnected results.

Automate decisions, not just clicks

For a decade, the enterprise obsessed over Robotic Process Automation (RPA), Kevin says.. It was a necessary first step but it had a hard ceiling: it could only automate the click. RPA was rule-bound, brittle, and expensive to maintain. Move a UI button one pixel and the bot broke.

Agentic AI changes the unit of automation. RPA automates the task but an agent automates the decision. Agents handle exceptions, adapt to new data, and work across multiple systems without breaking when a field name changes. As Kevin put it: "Agents are interns with root access, infinite patience, and no fear of HR."

You can't run an agent and forget it. Kevin says you need an architecture that is composable, modular, and interoperable, so that when a process changes the agent adapts instead of forcing a rip-and-replace.

The imperative to elevate talent

A persistent boardroom myth is that AI is here to replace the workforce. The math says otherwise.

Kevin says the U.S. is heading into a demographic crunch. Over the next three years, roughly 18 million people are projected to leave the workforce and only 13 million to enter it, a shortfall of about 5 million. Letting good people go is the last thing most companies can afford.

So the point of AI isn't to replace your teams, it's to take the joyless work off their plates. That frees your best people to do what only people can do, which is exercise judgment, build customer relationships, and decide what comes next. Free them from the busywork and their capacity to create value climbs. The move is from "data processor" to "strategic orchestrator."

How to scale without chaos

How do you operationalize this without trying to change everything at once? Don't roll AI out to 5,000 employees overnight. You can't control the cultural fallout at that scale, and your HR department won't thank you for it.

Kevin says to use tiger teams instead:

1. Small and incentivized: Pick a high-performing group of 7-10 people and give them the resources to make AI work on one specific, high-impact process.
2. Prove, then scale: Don't chase a proof of concept; build a repeatable playbook. Let the team nail it in the real world first, then scale what worked.
3. Governance and biometrics: Agents with root access mean security has to come first. Every agent should be registered and governed by IT. For high-stakes decisions, add human-in-the-loop gates, specifically biometric assured identity, so you know the agent is operating within its bounds.

Watch the full session

The inflection point isn't coming; it's here. In the full session Kevin covered everything from the "SaaS Apocalypse" to why natural-language interfaces are replacing the traditional dashboard.

If you're a leader on the hook for results, staying in the 95% that burns cash on pilots isn't an option. You need the architecture, the talent shift, and the security gates that hold up in practice.

You can watch Kevin’s full session – The Enterprise AI Inflection Point – on-demand here

AI isn’t delivering ROI because it has operational blind spots. Time to meet The Context Model: your guide to Enterprise AI’s missing layer. Get your copy.

Discover the difference between agents, assistants and copilots in the eBook, Destination AI: Unpacking assistants, co-pilots, and agents

Explore the I-Chart: The key to measuring your enterprise acuity and transformation vision in the AI era. Click around and find out.