Lovelytics has partnered with Ecolab since July 2025, working embedded within both the internal AI platform team and the commercial software-as-a-service team.

On the internal side, Ecolab’s momentum on AI was already strong: half a dozen groups across sales, service, supply chain, R&D, and HR had each stood up their own agents to solve for their function’s needs, based on templates Lovelytics had created. As adoption scaled, the natural next step was giving employees one consistent way to access all of it.

One Front Door for Employee AI

To deliver that, the team designed an Agent Registry and Orchestration (ARO) platform on Azure and Azure Databricks. 

Here’s how it works.

A planner agent, backed by Claude Sonnet, reads each incoming request and routes it to the right specialist agent based on the employee’s role and access, with new agents able to register into the system as they’re built.

Development was accelerated by Claude Code, taking the platform from handoff to production in roughly a quarter of the time. Today the platform unifies 12+ agents behind one interface, is onboarding more than 5,000 employees (starting with sales and service, with finance and supply chain next), and has produced a reported 4x reduction in time spent on supported tasks.

Scaling AI for the Frontline

In parallel, Ecolab’s commercial team was pursuing an ambitious retail and restaurant intelligence agent, the kind of tool that helps major retailers and quick-service restaurant chains manage cleanliness and equipment maintenance by pairing IoT sensor data with inspection insights. 

Built for high complexity and with a ~$20M revenue opportunity on the line, the initiative needed a partner who could help scale the model and stand up production-grade infrastructure with speed.

Lovelytics rebuilt it from the ground up as a single, unified Claude-powered agent. Under the hood, a tiered model strategy does the heavy lifting: Haiku handles fast, lightweight routing, while Sonnet takes on the complex reasoning, grounded in real operational and performance data.

Layered on top, semantic SQL querying lets the agent understand and answer questions naturally, short- and long-term memory lets it build context over time, user profiling tailors each interaction, and built-in multi-language support makes it ready for a global frontline.

Because the architecture was modular, the team could replicate the full agent for new quick-service restaurant clients in under a week, delivering frontline teams context-aware guidance inside the product experience itself. 

Reusable templates cut time-to-product by 70%+, and the platform now has 12+ agents live in production. 

New-client onboarding, once a manual multi-hour process, was automated into a guided, Claude-driven workflow with pre-commit validation and branch protection, cutting setup time by 93%, from roughly three hours to about thirteen minutes.

Ecolab needed a partner willing to operate as a genuine extension of its own engineering team rather than a vendor working at arm’s length. They ultimately chose Lovelytics for their hands-on Databricks and Anthropic platform expertise and ability to turn fast-moving prototypes into governed, scalable systems across both internal and commercial use cases.

Claude Code gave the team real velocity, but that speed came with a cost: early development produced a sprawling five-repository platform. The code was modular, but changes were hard to trace through it; documentation was strong, but CI/CD validation was inconsistent from repo to repo. 

Lovelytics addressed this head-on by simplifying the code and pipeline patterns, guiding the internal foundations pod with proven examples from the internal build, applying Claude Code with scoped skills and vetted patterns, and adding validation, error handling, and endpoint checks that hadn’t existed before. 

The lesson carried forward into every later phase of the engagement: Claude Code accelerates development significantly, but production-grade reliability still depends on experienced engineering practices layered on top of it.

Across both the internal and commercial sides of the engagement, Ecolab replaced a fragmented collection of one-off agents with a single, reusable platform pattern, cutting the time it takes employees to get help, protecting a significant piece of commercial revenue, and making it dramatically faster to bring new agents and clients online.

• 4x reported reduction in time spent on supported tasks by employees using the unified agent interface.
• 5,000+ employees onboarded to the platform, with finance and supply chain rollouts planned next.
• 12+ agents unified behind a single registry and orchestration layer.
• ~$20M revenue opportunity protected by reviving a stalled commercial agent project.
• 93% faster new-client onboarding (roughly 3 hours down to 13 minutes) and 70%+ faster time-to-product via reusable templates.

Ecolab is preparing to extend the internal agent platform to finance and supply chain teams, while continuing to onboard new commercial retail and restaurant clients onto the same reusable architecture. Lovelytics remains embedded with the team to support that next phase.