How Kata.ai built an AI chatbot platform for 200+ enterprises

Executive overview

Most customer service operations drown in repetitive queries, high agent turnover, and rising costs. Kata.ai built a conversational AI platform that handles those queries at scale, with a human-agent fallback loop to catch what the AI misses.

The company pivoted from a B2C virtual assistant to a B2B platform by productising its NLP technology into developer tools that IT consulting firms could deploy for enterprise clients. The edge: deep training on Indonesian language, which has 13 variants of the first-person pronoun alone.

Falling in love with the problem, not the technology, is what separates sustainable AI companies from feature shops.

The Indonesian language challenge

• English has one first-person pronoun; Indonesian has 13 variants
• Abbreviations and slang create significant NLP complexity
• Kata.ai collected 3 million messages as the initial training dataset
• This hyperlocal data is the core competitive moat — not accessible to OpenAI or generic models

Pivoting from B2C to B2B

• Started as Yes Boss: a human-staffed virtual assistant (50 agents) for Indonesian consumers
• Recognised from day one that human agents were not scalable
• After 18 months building the AI model, pivoted to B2B and renamed to Kata.ai
• The pivot required letting go of 90 staff — described as the toughest three months
• Helped departing staff find jobs and build CVs to manage the transition humanely

First B2B breakthrough

• A major Indonesian consumer goods brand wanted to build a "virtual female best friend" on Line messaging
• Kata.ai ran a POC with its early NLP technology and won the contract
• The chatbot reached 2 million users within a year
• 40% (800,000 users) provided data enabling retargeting and remarketing
• Success with a prominent brand created a snowball effect — other enterprises followed

Platform and go-to-market model

• Early contracts required full customisation (software house model) — not scalable
• Built a developer platform so IT consulting firms could deploy Kata.ai NLP themselves
• Shifted to a partnership model: consulting firms and software houses as the distribution channel
• Rationale: in B2B, long-established relationships cannot be disrupted by technology alone
• Annual subscription pricing based on conversation volume

Human-in-the-loop design

• Every deployed chatbot works alongside a human agent in the back office
• If the chatbot cannot resolve a query, it hands off to a human agent
• Human agent interactions feed back into chatbot training
• Target: reduce the 80% of repetitive queries currently handled by stressed human agents
• Human agent cost: ~$3 per call; chatbot handles volume at a fraction of this

Building the right team

• Three traits to hire for: shared mission and why, growth mindset, humility and coachability
• Shared why makes the rest limitless once people buy in
• Growth mindset matters more than current skill level
• Humility enables learning from mistakes — both the founder's and the team's

Opportunity for local AI companies

• Generic AI algorithms are now widely accessible — the differentiator is hyperlocal data
• Data not on the public internet and not owned by large models is the edge
• Opportunities exist in agriculture, supply chain, healthcare, financial services
• Founders who combine AI with hyperlocal data solve problems global players cannot