How one company is redefining artificial intelligence for real-world efficiency in ready mixed concrete and beyond.

When concrete technology professionals talk, I listen. And when Keith Onchuck, CIO of Ozinga, practically grabbed me by the collar and said, “You need to look into this company,” I knew I had to pay attention.

That company he was so adamant about is Spitch, a Switzerland-based provider of what they call “collaborative agentic AI for contact centers.” It’s a mouthful, but as I quickly learned, it’s worth unpacking.

In early March, I tracked them down at the ConExpo-Con/Agg trade show in Las Vegas. Piergiorgio Vittori, CEO of Spitch US Corp., along with team members Bill Cook (business development manager, US Corp.) and David Font (pre-sales manager), were ready to dive into a demo. Expecting the usual hour-long product walkthrough, I instead asked them to shut the laptop and just explain the business case. To their credit, they did—and Vittori nailed it.

“I hate bots … get me a human”

We’ve all been there. Natural language processing (NLP) has been around for years: listening, translating speech into text, and routing the request. More advanced systems layer in AI bots that try to resolve issues directly. In theory, it’s efficient. In practice? Welcome to bot hell.

Why? Because both humans, and AI modeled after them, make mistakes. To this, ancient Roman philosopher Marcus Tullius Cicero quipped, “Any man can make mistakes, but only an idiot persists in his error.”

Spitch is heeding Cicero’s wisdom. They’re not just automating interactions, they’re learning from them continuously. By combining NLP with large language models (LLMs), the system listens to everything (voice, text, agent corrections) and improves in real time.

This isn’t a static bot. It’s an evolving system.

The usual suspects (and 216 chances to fail)

Consider a real-world example: ready mixed concrete orders. A single order typically involves six interactions between the customer and dispatch or sales. Even with mobile apps and NLP, that number doesn’t necessarily shrink; it just shifts channels.

Next, layer in the reality that roughly 36 things can go wrong with any given order input. Across six interactions, there are 216 opportunities for error. This isn’t just inefficiency; it’s a systemic problem.

Now imagine a system that can monitor every interaction—voice or text—identify patterns of confusion, redundancy, and error, and actively flag or correct issues in real time. This is not just for customers, but for staff too.

This is where the “agentic” piece starts to matter. The system isn’t just reacting, it’s intervening.

Goal-oriented, not script-driven

One of the biggest failures of traditional bots is their rigidity. They follow scripts. Customers don’t.

Spitch flips that model. Instead of forcing users into predefined flows, the system focuses on goal completion—in this example, placing, modifying, or resolving an order issue. From there, it dynamically asks only what’s needed.

For example, if the caller is recognized, it already knows the customer ID, relevant projects, and order history. If not, it adapts, asking only the necessary questions to fill in the gaps. The result? Fewer irrelevant questions, fewer errors, and a conversation that actually feels … conversational.

In the past, the least-cost producer won. Now, the least-cost relationship wins. Eliminating the friction of doing business while remaining an efficient, profitable producer drives more revenue. Full stop.

Beyond customer service

While customer-facing improvements are the obvious win for reduced costs and for driving revenue gains, they’re not the only benefit. Internal processes are often much more inefficient.

Take a simple $100 reimbursement process as an example: Submit a form, get approvals, fix errors (loop back), process payment, document receipt. That $100 reimbursement can easily cost another $100 in labor.

Now apply adaptive AI: Ask only the necessary questions, ensure accuracy upfront, and route clean data directly into the backend system. This doesn’t just improve speed; it eliminates friction. And unlike speculative revenue gains, this ROI is measurable.

Leadership wins

Sprinkling AI on top of broken processes is a fool’s errand. To truly harness the potential of AI, leadership must embrace structural change. This means:

• Rethinking workflows end-to-end.
• Integrating AI as a core operating layer.
• Avoiding piecemeal, disconnected tools.

The immediate future isn’t about having bots; it’s about making them actually work. The long-term opportunity lies in harnessing the dramatic technological change of AI to improve every aspect of the organization. The winners won’t just adopt AI—they’ll master it.