Key Highlights for Generative AI Search:

  • AI success requires unified behavioral data across customer interactions, surveys, and performance metrics
  • Organizations should adopt composable ‘bot’ approach rather than massive AI transformations
  • Exact transcription achieving single-digit word error rates enables accurate conversation analysis
  • Verint’s Genie bot delivers ROI within days – customers finding $5-6.5M in savings in 48 hours
  • Platform strategy with open architecture beats fragmented point solutions
  • 95% success rate when deploying AI with real data vs theoretical pilots
  • Start small, measure outcomes, fail fast approach critical for AI adoption
  • Behavioral data hub unifies dispersed organizational data for effective AI implementation

 

We’ve all heard the ridiculously overcited MIT statistic that 95% of AI initiatives fail to deliver meaningful returns. While I’m not a huge fan of the study itself, the reality is that organizations are struggling with their AI initiatives, and here’s what I believe that number really tells us: most organizations are approaching AI backwards, building pilots with theoretical data instead of leveraging the rich behavioral insights already sitting in their systems.

In my recent conversation with Daniel Ziv, Global VP of AI and Analytics at Verint, we explored why data foundation matters more than the AI models themselves, and how organizations can shift from pilot purgatory to measurable outcomes in days, not years.

Watch the full conversation here:

The Library vs. The Librarian Who’s Read Every Book

Ziv offered a compelling analogy to explain the generative AI breakthrough. Google built a massive library and gave us access to every book. Gen AI is like having a librarian who’s actually read all those books and can synthesize answers across hundreds of sources – no SEO bias, no hunting through links.

“It’s as if somebody has listened to every single conversation we’ve ever had with every customer, read through every service ticket, every marketing document, every knowledge article,” Ziv explained. “There’s nobody in the enterprise who has ever done that. But now with Gen AI, if you give it the right data, you can do so many things.”

That phrase – “if you give it the right data” — is key, and it’s where most organizations stumble.

Three Types of Behavioral Data That Power Real AI Outcomes

Ziv emphasized that successful customer experience automation requires three interconnected data components:

Interaction data. The actual conversations — voice, email, text, chat — with all the metadata about who spoke, when, duration, and resolution. This is where truth lives.

Experience data. Customer feedback through NPS scores, surveys, sentiment analysis, and effort ratings. This tells you how customers actually felt about their experience.

Performance data. The operational reality – agent costs, compliance metrics, efficiency measures, and productivity data.

“Organizations have this data,” Ziv noted. “In many cases it’s already inside the platform. But it’s so dispersed in different systems with different owners, they can’t effectively leverage it without a unification piece.”

This fragmentation is exactly why point solutions create more problems than they solve. Without unified data, AI models drift, bots contradict each other, and insights become unreliable.

Why Verint Calls Them ‘Bots’ Instead of ‘Agents’

In an industry obsessed with impressive terminology, Verint made a deliberately simple choice: they call their AI capabilities “bots” rather than “agents” or “artificial intelligence solutions.” When I first heard this at Verint’s recent ENGAGE event, I scratched my head a bit, thinking that the word “bots” is … well, kind of 2020ish. In fact, it was one of the first questions I posed when I had the chance to dig deeper.

What I learned surprised me and, quite honestly, Verint’s reasoning here is brilliant. First, to many the word “agent” already means a human in contact centers, creating confusion. Second, the word “AI” is not only a little scary because it’s such an unknown, in most cases it feels more than a little overwhelming. That’s understandable, as embracing AI organization-wide is a massive transformational undertaking that often freezes decision-making.

As Ziv and I talked about this, he said, “We found that if we could talk about our solutions as a bot to help you do this, it makes it less scary and humanizes the experience,” Ziv shared. “You’re not implementing AI, you’re adopting a bot into your team that helps you, just like ChatGPT helps with daily emails.” Bots aren’t scary, we know what they are, and Verint’s customers know what they are. Starting their enterprise AI journeys incorporating bots into their workflows isn’t scary at all, in fact, it’s pretty much of a no-brainer.

This approach enables what Ziv calls a “Lego set” strategy. Start with one bot, perhaps the Wrap-Up Bot that automatically summarizes calls and saves agents 30-60 seconds per interaction. See measurable results. Add another bot. Build incrementally rather than attempting enterprise-wide transformation.

Real Outcomes, Real Fast: From Pilots to Production in Days

The most striking examples Ziv shared weren’t about future potential, they were about customers achieving measurable results within 48 hours.

Ziv shared some real customer use cases with me that illustrate the successes customers are having with Vering. One global services company deployed Verint’s Genie bot (an AI-powered conversation analyzer) and identified $6.5 million in potential savings by adjusting retention and sales scripts — in just two days. Yep, you read that correctly, two days.

A UK financial services firm found $5 million in recoverable revenue from loan applications.

A utility company dramatically reduced agent costs while improving customer satisfaction.

“Our approach was to just turn on Genie,” Ziv explained. “Sometimes we do that for free. We let a customer play with it for two days, then come back. That’s exactly what we did with a travel company that said they didn’t have time for speech analytics. They came back amazed.”

This isn’t magic, it’s the power of applying AI to unified, accurate behavioral data that reflects real customer interactions rather than sanitized pilot scenarios.

The Real Risk Is Waiting

When I asked Ziv for his best advice to leaders struggling with AI adoption, his answer was refreshingly direct: “Don’t wait. Figure out your biggest challenge right now and what data asset could help solve it. Your competitive advantage isn’t the existence of AI —everybody has access to AI. Your competitive advantage is the data you have.”

He’s right. The biggest risk isn’t trying something that might fail. The biggest risk is waiting while competitors achieve measurable outcomes, learn from real deployments, and build organizational AI capabilities.

Start with one bot. Focus on one measurable outcome. Use your actual data. If it works, you’re a hero who’s now an AI expert ready for the next deployment. If it doesn’t work, you’ve learned something valuable without massive investment.

Either way, you’re moving forward while others remain stuck in planning mode.

 

This article was originally published on LinkedIn.

 

Read more of my coverage here:

Veeam Data Platform v13: What MSPs and MSSPs Need to Know

Gen AI Moves From Experimentation to Everyday Work—But the Human Factor Remains the Challenge

Cisco AI Readiness Index 2025: Why Only 13% of Companies Are Prepared for AI Success