Topcon Scales Support with Level AI by Turning their Knowledge Base into a Real‑time, AI‑Powered Resolution Engine
The results
Precision When It Matters Most
Topcon Positioning Systems doesn't just manufacture equipment; they power the infrastructure of our world. They develop GPS, laser, and optical technologies that automate massive machinery for the construction and agricultural industries.
We aren't talking about consumer gadgets; we are talking about ensuring that the race track is perfectly flat so a car doesn't crash at 200 mph, or automating a bulldozer to grade a site to the millimeter. Being a strong player in a highly technical B2B market, Topcon supports a global network of distributors who rely on complex, 300-page technical manuals to keep the world’s infrastructure moving.
"We make precise positioning equipment... If we know precisely where you are, we can do amazing things with equipment. We will then automate bulldozers, motor graders, asphalt, concrete pavers, and farm equipment." — Ben Huber, Director of Business Performance Analytics
The "Unforgiving Audience" & The 9-Month Failure
Topcon faced a scalability problem at their contact center: the products were so niche that finding new talent to support it was nearly impossible. They needed to scale and optimize their operations by implementing an autonomous agent that could handle complex customer queries, but they weren’t sure if AI could generate contextual and accurate responses to resolve customer queries.
Ben Huber describes their users as a "very unforgiving audience." If a concrete paver breaks down, a fleet of trucks is left waiting, costing the customer tens of thousands of dollars per hour. Customers would feel uneasy having a bot as the initial point of contact in such a demanding situation.
To make matters more difficult, Topcon had just spent nine months trying to implement a solution with a previous vendor. The project failed miserably because the bot relied on manual decision trees that required customers to guess the right menu option. It was, as Huber put it, "neither artificial nor intelligent."
"We basically knew we were going to have one chance to get this right... We have a very unforgiving audience... [Downtime] could be turning into tens of thousands of dollars per hour waiting." — Ben Huber, Director of Business Performance Analytics
A Holistic AI Ecosystem
Topcon partnered with Level AI to build a solution that wasn't just a "chatbot," but a comprehensive support ecosystem that is driven by real human-grade AI. They moved from zero to a live in just four to six weeks by executing on four critical use cases.
1. Integrating the Virtual Agent with Deep Knowledge
Instead of manually programming answers, the team integrated Level AI’s Virtual Agent directly with Topcon's massive library of 12,000 to 15,000 knowledge assets. The goal was to eliminate "search cost" with intelligent knowledge retrieval. By instantly being able to digest 300-page manuals, Level AI’s Virtual Agent helped Topcon create quick guides, with specific answers for customer queries that humans struggled to locate.
2. Multiplying Workforce with Agent Assist
To address the difficulty of hiring niche talent, Topcon leveraged Agent Assist as a "multiplying agent" for their human team. By delivering contextual knowledge to agents in real-time, the system helps retain talent and reduce the onboarding time, ensuring that even newer trainees can support complex B2B queries without needing years of experience.
3. Tuning Conversational Flow & Query Handling
Topcon worked closely with Level AI to refine the logic of the Virtual Agent, enabling AI to distinguish between similar widgets used in different applications and avoid confusing terms like a ‘planting machine with a concrete paver’. By implementing "checks" early in the conversation to confirm user intent, they were able to re-steer the conversation correctly, ensuring high accuracy for their diverse product portfolio.
4. Integrating with Telephony and CRM
Finally, Topcon ensured the AI wasn't a silo. They integrated the agent with their telephony systems and CRM (Salesforce). This allows them to track case history and capture metrics, ensuring that if a call is deflected or handled by the AI, the data still "lives in the real world" and contributes to business insights.
"We now need to scale and have deflection capabilities so that we can free up our agents and subject matter experts, to go create more content, which makes the AI smarter... perfect snowball of heading towards perfection."— Ben Huber, Director of Business Performance Analytics
The "Perfect Snowball" Effect
By implementing Level AI, Topcon effectively removed the "search cost" for their dealers. The AI now instantly combs through thousands of complex SKUs to find the needle in the haystack exactly at the moment agents urgently require information.
This success has created a flywheel effect. Because the AI handles information retrieval from a vast knowledge base, Topcon’s Subject Matter Experts (SMEs) are no longer bogged down by repetitive questions. They now have the time to write new support content. This new content is fed back into Level AI, making the agent smarter, and providing resolutions for even more questions.
The Results:
- Time to Value: Went from a 9-month failed implementation with a previous vendor to a live MVP in 4 weeks with Level AI.
- Multilingual "Surprise": Level AI’s Virtual Agent can now successfully handle Spanish queries by reading English manuals and translating the answer for faster resolution on the fly, helping Topcon scale their contact center operations seamlessly.
- Future Expansion: Due to the success of the Virtual Agent, Topcon is now preparing to launch Level AI’s Voice Agent to intercept and resolve phone calls before they ever reach a human agent.
"I would recommend Level AI to anyone who's looking for a partner in an AI environment where you're looking for true collaboration... willing to collaborate with us on needs that we have and truly wanting to partner in a bilateral way." — Ben Huber, Director of Business Performance Analytics
