ChatGPT and the future of Support
Zendesk and Freshdesk create desks. Modern SaaS needs a round table.
“Nobody wants to stay in support long-term,” said the Head of Support of a modern SaaS company recently. I wondered why there was such little pride or purpose in the craft. And if GPT — and AI at large — truly take away the repetitive chores of level-1 and level-2 agents, you’re left with level-3 support engineers who need a path to be celebrated, to be proud of what they do, to feel like their chain-of-thought reasoning is way better than that of an emergent language model.
What support folks get today is a “desk” for filing, emailing, notifying, searching, deflecting, classifying, deduplicating, clustering, (copy-paste) responding, and routing — all chores that a machine and its AI models can be really good at in the coming years, especially as the underlying data gets broken down into models that capture specificities in product parts and customer identities. AI models will be most effective on modern object models that organize information in “rows and columns” of parts and users.
That is what amazon.com did so well in its war with eBay — organize every product listing in its catalog on (unique) parts and then capture every search query and shopping cart behavior of every user in real time. They knew demand and supply at the back of their palm. They knew their inventory — and their customer — better than any retailer on the face of this earth. Most importantly, their customer service knew everything about their customer’s buying history and behavior to support them in real time. Having an elegant product catalog that described inventory and organized sellers around those listings nicely closed the loop on customer experience. Amazon was unsurpassable. Their support agents were not sitting on a desk passing messages between departments. They were information machines who could make service decisions autonomously and with pride.
Pride for their jobs is elusive for support engineers in modern day SaaS. Many technical folks are made to do repetitive tasks because chat search is weak, deflections are poor, responding with shortcuts is rarely specific, routing is a game of hot potatoes, and daily KPIs are bogus. And no one dares to classify, cluster, and label tickets — on behalf of PMs or developers — because a support engineer is supposed to keep busy fulfilling daily KPIs of ticket creation, SLA violations, and closure. Customer success… that’s another department 🤯
DevRev believes that support engineers need a space to collaborate with developers and PMs. They need to know about roadmaps, incidents, enhancements, deployments, sales opportunities, customer purchase history, product usage, browser crash reports, and user analytics in real time. Only then will they have confidence (and joy) in what they do — so over time, they’ve a career path that leads them to customer success, solution architecture, product management, and general management.
To be better than ChatGPT, service folks of the future don’t need a desk. They need a round table!
