I’ve always been a tech builder—first with software, then analytics, and finally AI—spending close to 25 years across Finance, Hi-Tech, and Automotive industries. Along the way, I’ve built technology internally and bought technology to transform businesses. And somewhere in that journey, a pattern became impossible to ignore. The gear shift before and after buying software was stark. Pre-purchase, there was energy—engagement, curiosity, conviction. Sales teams spoke the language of outcomes and transformation. Post-purchase, that zest often gave way to slower engagement, diluted focus on business outcomes, and an uncomfortable amount of dilly-dallying on promised features. What followed was a familiar “sugar-drop” after the initial high of the deal. Digging deeper into why Customer Success was consistently punching below its weight, one thing stood out: customer success teams were simply under-equipped. They had very little technology to manage customer goals, track product usage, drive adoption, or think meaningfully about expansion and advocacy. This was particularly constraining for one of the broadest job families in SaaS—the Customer Success Manager. CSMs were expected to do everything: onboarding, ongoing support, adoption, renewals, expansion, and advocacy. Yet the software supporting them was largely software-defined—rigid systems where every new scenario, data condition, or workflow required explicit design and engineering. The real world, meanwhile, refused to cooperate. Then came LLM-powered AI agents. Suddenly, many of these constraints began to dissolve. AI agents could operate across diverse situations, ingest massive contextual signals about a customer, and—most importantly—work with fuzzy, incomplete, and imperfect data. For the first time, it felt plausible that CSMs could have a true companion, not another dashboard. Something that worked with them, not around them. Building a fully AI-powered SaaS platform forced us to reimagine everything. The entire business layer became encapsulated in AI, creating a flywheel of learning and execution. As LLMs grew more powerful, so did the agents—becoming increasingly “built to fit” the realities of customer success teams rather than forcing teams to fit the software. That reimagination touched many areas. Metrics had to change. Users needed visibility into how agents were operating, not blind faith. This led us to FAIR metrics—Functional, Action, Impact, and Review—designed to keep humans in the loop while still benefiting from autonomous execution. We also built a Universal Chat Agent, an “uber-CS champion” capable of spanning the entire customer lifecycle. In hindsight, that became one of the most powerful constructs in the platform, even though it was among the hardest to get right. But another challenge surfaced along the way—one that went beyond product design. AI introduced variability. Variable paths to outcomes. Variable costs. Variable effort. And that brought us face-to-face with what I now think of as ROVI—Return on Variable Investment. Businesses weren’t just asking whether AI worked; they were asking whether they could trust the economics of something whose costs and outcomes weren’t fixed. That insight pushed us toward BYOL—Bring Your Own LLM. By allowing customers to power the platform with their own LLM keys, we shifted control back to them—control over cost, visibility into usage, and direct ownership of how intelligence is applied within their workflows. In a world where intelligence itself is variable, that ownership matters. Looking back, this journey wasn’t just about building better software. It was about rethinking how work gets done, how impact is measured, and how trust is built when autonomy and intelligence are embedded deep inside the system. And for customer success teams—who have always operated in the messy middle between product, people, and outcomes—that shift might finally be the one that levels the field.

zeeproc

Agentic Customer Success Platform for modern AI native teams

Agentic Customer Success Platform for modern AI native teams

Forum Threads

Launching zeeproc on Product Hunt tomorrow!

Forum Threads

Launching zeeproc on Product Hunt tomorrow!