About us
We built the support tool we wanted to use ourselves
Quack started as a frustration: every AI support product we evaluated ignored the best signal a team has, its own resolved tickets. We decided to build the one that doesn't.
The origin
Why we started Quack
We spent six months talking to B2B SaaS support teams before writing a single line of product code.
The pattern was consistent: teams had good documentation and they had archives of resolved tickets going back years. They also had support agents spending a third of every day answering the same eight questions, worded seventeen different ways.
The AI tools available to them were either FAQ bots that couldn't handle novel phrasing, or generic LLM wrappers with no product knowledge that hallucinated product-specific steps. Neither category used the resolved ticket archive at all. That was the gap.
Quack is built on the thesis that a support AI trained on your actual product history is categorically different from one trained on the internet. It knows your specific settings paths, your exact feature names, your pricing edge cases, your migration steps. That specificity is what makes it actually useful to customers instead of just technically functional.
The team
The people building Quack
Nadav Kemper
CEO and co-founder
Nadav spent five years running support operations at two B2B SaaS companies before co-founding Quack. He wrote about the ticket patterns that led to Quack in the blog series that became our founding thesis.
Arjun Mehta
CTO and co-founder
Arjun built the ingestion and retrieval architecture that makes Quack accurate on product-specific questions. Before Quack, he spent years building ML infrastructure for high-throughput data pipelines at growing B2B software companies.
Tomoko Hayashi
Head of Customer Success
Tomoko onboards every early-access team personally. She came from a support leadership background at a growth-stage SaaS company where she managed a team of 22 agents. She knows what good support looks like from the inside.
How we work
What we actually believe
Specificity beats generality
A support answer that names the exact settings path is worth ten generic suggestions. We optimize for responses that are right about your product, not responses that could apply to any product.
Escalation is not failure
A well-designed escalation that gives the human agent full context is better than an overconfident AI response. We built Quack to know its limits and hand off gracefully, not to pretend it knows everything.
The loop matters more than the model
The quality of a support AI is determined by its training loop, not its base model. Every resolved ticket that comes back as training data is what makes Quack improve. That feedback loop is our product.
Early customers teach us what to build
We are in early access for a reason. Every onboarding call teaches us something we didn't expect. We talk to every customer personally during this phase, not because it's efficient, but because it's how we learn.
Join our early-access cohort
We are onboarding B2B SaaS teams personally. If you have a support queue, documentation, and closed tickets, we would like to talk.