SaaS companies are racing to add AI to their support offering, sometimes for the wrong reasons. Adding AI to support shouldn鈥檛 be to deflect tickets or make humans harder to reach. It should be to enhance the experience of the customer and get them the right help as quickly as possible.
We鈥檝e added AI to 海角天涯社区 support too. We built Gearbot, with Fin (formerly Intercom, recently acquired by Salesforce), and in this post I want to explain how and why.
海角天涯社区鈥檚 support is consistently rated among the best in Salesforce DevOps: top marks on G2 and the AppExchange, fast responses even at peak volume, and the same level of help for every customer, with no higher tier to buy your way into. The people answering you understand metadata, deployment errors, Git integrations and more, because that鈥檚 the work they do every day.
That鈥檚 why, when we decided to bring AI into support, we made it completely optional.
What 鈥渙ptional AI support鈥� actually means
When you reach out through the in-app chat, you choose the kind of help you want. If a support engineer is free, you talk to a human straight away. If there鈥檚 a wait, we tell you up front roughly how long it鈥檒l be 鈥� so you can wait a few minutes for a person or get going now with Gearbot.
Either way, you鈥檙e making an informed choice. We don鈥檛 default to AI first or make you fight your way to a person. We make your options clear and give you the power to choose.
Choice matters because 海角天涯社区鈥檚 users are dealing with deployments, data, and production Salesforce orgs, where being unblocked correctly matters more than being unblocked fast. Our users are engineers and admins who can tell instantly when a vendor cares more about closing a ticket than solving a problem. Forcing AI on that audience doesn鈥檛 build trust 鈥� it erodes it.
Different ways to get help
We think about support as three complementary channels, not a hierarchy:
- Documentation is for when you鈥檇 rather read and solve it yourself. Fast, self-serve, no waiting.
- Gearbot is instant and always available. It鈥檚 strong at the things AI is genuinely good at: pointing you to the right doc, walking through common how-tos, surfacing known issues, and handling the long tail of questions that have known answers. As it鈥檚 based on past support engineer responses and documentation, it鈥檚 not just what you find on Gemini or ChatGPT.
- Customer Support Engineers are the experts you want for edge cases, more complicated problems, and anything account-specific 鈥� the problems that need judgment, context, and someone who can dig in alongside you.
Our amazing Customer Support Engineers are great at delighting users with empathy and expertise. Gearbot is great at getting the right information to users even faster for known challenges and solutions. The key is getting you the right kind of help for each question.
How we trained Gearbot
For the six months leading up to release, our support engineers reviewed Gearbot鈥檚 answers, sharpened its accuracy, and worked out how to correct it when it got things wrong.
The results held up. When offered the choice, 35% of users chose Gearbot rather than wait for a person. Out of the box, before further refinement, our team rated 70% of Gearbot interactions as Good or Acceptable.
The principle underneath it all
AI hasn鈥檛 changed what good support is. It鈥檚 still about getting you to the right answer, the way that suits you, as quickly as the problem allows. AI helps us do that better. It鈥檚 not a replacement for our engineers, and it鈥檚 not something we鈥檒l force on you.
I鈥檓 proud to continue offering world-class support 鈥� on your terms.
