Show notes
In this episode of Marketing Spark, host Mark Evans talks with Jiaqi Pan, co-founder and CEO of Landbot, an AI-powered platform revolutionizing customer service.
Jiaqi shares Landbot’s journey from a B2C service to a leading B2B SaaS solution.
Topics include recognizing unsustainable business models, effective brand positioning, and balancing AI with human interaction.
Jiaqi offers valuable advice for B2B SaaS entrepreneurs and discusses metrics for measuring AI-driven customer service success.
Tune in for practical insights into leveraging AI, driving leads, and enhancing customer experience.
Connect with Jiaqi on LinkedIn and learn more about Landbot at landbot.io
From the conversation
How Landbot pivoted from a personal assistant service to chatbots
Jiaqi Pan, co-founder and CEO of Landbot, explains that the company began as a B2C service that worked like a personal assistant. Customers chatted with human agents to book trips, run errands or buy things online. The team built technology to improve its own customer operations, and Jiaqi says it became clear that the technology could also help other businesses talk to their customers.
The business model was the problem. While preparing for a new financing round, Jiaqi looked at a graph of costs against revenue. Costs were growing exponentially and revenue was growing more linearly. His conclusion was that the company would either need to raise something like 100 million dollars to fund the operation, or it could not scale.
Chatbots were becoming popular around 2017, so Landbot repositioned itself as a chatbot solution. Jiaqi says it took an intuitive, easy-to-use approach that let non-technical business users build their own bots. This was before the term no-code was in common use.
Advice for B2B SaaS founders on problem, market and pricing
Jiaqi says his advice applies to any type of business. The first step is to identify the problem, and he sees many founders go wrong at this point.
He also says founders need a clear target customer, a clear pain and a reason their solution is better. With thousands of SaaS products already available, Jiaqi does not think a product has to be 10x better, but it does need some angle for telling prospects why they should choose it.
On pricing, Jiaqi describes what Landbot did and what it should have done. The team looked at comparable products and priced about 30% lower. That worked, with people paying quickly and little friction. In hindsight, he says, the team should have done proper customer discovery to understand the value of the solution, either in cost savings or in added revenue.
Once that value is clear, Jiaqi suggests charging a part of it. His recommendation is about a tenth, which gives the customer roughly 10x the ROI, though 5x can also work. Without that anchor, he says, a company ends up in a pricing war.
"I think the problem many people or many entrepreneurs have is that they are starting with the solution in mind."
How AI technology for chatbots has changed
When Landbot first launched, the important technology was natural language processing (NLP). Jiaqi says it tried to spot patterns and keywords and return structured responses. It was hard to map every intent because people can say the same thing in a million ways, so bots failed often. He gives the example of a bot that hears "I only speak Spanish, please" and replies by treating "Spanish, please" as the request.
Large language models, the technology behind ChatGPT, removed the need to map intents because they understand what users are saying. The drawback is hallucination, where the bot misinterprets something or says things that are not true.
Jiaqi says many larger businesses now use a hybrid approach. For sensitive topics such as pricing and product information, the bot uses NLP or standard linear flows. For non-critical information it uses a large language model, which gives more flexibility and a better user experience.
Balancing AI and human agents, and measuring results
Asked about AI replacing human agents, Jiaqi compares it to earlier technology shifts such as the car. He believes new opportunities will follow. He adds that much of what agents do today, such as answering repetitive questions and filling out the CRM, is not motivating. At Landbot, the 10-person customer service team is encouraged to automate work so they can become trusted advisors.
For metrics, Jiaqi separates output from input. In support, output metrics include how many tickets the AI handled without a human and how much time that saved. In sales and marketing, they include leads captured and the conversion of those leads to opportunities and closed deals. Input metrics include satisfaction, abandonment and engagement. Early on, with low volume, he suggests reading five randomly chosen conversations.
The right mix of AI and humans depends on the customer journey. Landbot uses bots across the journey. At the start, they help filter low-quality leads from some low-income regions that rarely convert. Managed accounts, the best and highest-paying customers, are mostly handled by humans to protect trust.
Brand positioning at Landbot
Jiaqi recommends the work of positioning expert April Dunford, and Landbot ran a workshop with her to define its positioning. The approach starts with existing customers. Talk to the best ones to learn their industry and sub-segment, what they struggle with and what differences they see in the product. The narrower the niche, the better.
Jiaqi stresses using customers' own words rather than the company's view of itself. Finally, he advises smaller players to pick an existing category instead of inventing a new one. Landbot stays in chatbots and looks for its niche there.
Jiaqi says Landbot's positioning is not perfect and keeps changing as new players arrive and technology shifts. Two years ago AI was barely relevant to the message, and now everyone is talking about it.
Lead quality, content marketing and customer discovery
Jiaqi says he does not think Landbot has this figured out, so he shares mistakes. About 80% of customers are self-serve, and the product generates around 10,000 sign-ups a month. Conversion is only 1% to 2%. After adding an onboarding question about why people use Landbot, the team found only 30% were using it for business. The other 60% were doing personal projects or academic work and would never convert.
Landbot had been optimizing for low-cost campaigns that bring volume, and that attracted people looking for free chatbots. Jiaqi now prefers high-intent business leads, which he says convert far better and have stronger retention and unit economics.
Content marketing remains one of Landbot's channels, and it needs strategy around keywords that bring quality rather than only volume.
Jiaqi's biggest regret is delegating customer interviews to the customer success and sales teams instead of doing them himself. He says nuances get lost, and founders hear the customer's words directly. His one piece of advice is not to skip this work, even though it is hard and boring and does not give immediate results.
"A lot of people say with all these AI-generated content, SEO will die. It's been two years, and content has still been growing nicely for us and keep growing, keep driving volume."