Like many other SaaS vendors, Freshworks is working to build AI into its solutions. While many other vendors choose to make the AI as open as possible for customers to use on their own, Freshworks is keeping the AI as locked down as possible. In doing so, Freshworks says it is opting for the best user experience and consistent results.
At the Freshworks Refresh event in the Netherlands, we spoke with Mukesh Mirchandani, Senior Vice President of Global Engineering, and Mika Yamamoto, Chief Customer Marketing Officer. Mirchandani said Freshworks has been developing AI solutions for about six years now. The most important thing for Freshworks is to ensure that AI doesn't become an out-of-control science project. AI should primarily provide features that can be turned on and used immediately. Yamamoto added that instead of developing generic AI capabilities, the company looks closely at the customer's business case and how AI can be applied to that case.
Freshworks has a large partnership with AWS, primarily for hosting its SaaS solutions. For AI, Freshworks currently primarily uses Azure, which offers an enterprise version of OpenAI. However, Freshworks says development is moving forward rapidly and that it has developed proofs of concept with Amazon's LLMs. Additionally, it has projects underway with LLMs from other vendors, including Anthropic, Meta, and Google.
Domain and industry AI models are the future
Mirchandani said Freshworks is keeping its options open for now. Freshworks' AI, Freddy, is currently performing satisfactorily. Freshworks needs LLM for text generation, which can be done well on Azure today. Consider generating responses to support tickets, creating knowledge base articles based on support tickets, etc.
Mirchandani said they work on domain and industry models for other purposes, such as determining next steps in a process or understanding specific customer questions and matching them to features.
A good example of a domain model is a refund policy. A customer who wants a refund from a retail organization and wants to return a product is often subject to rules. Some products allow it, others don't. Some organizations give it 30 days, others 90 days. Then there are steps that need to be taken. Mirchandani says this is best expressed in a domain-specific model, so that a chatbot on a website can handle it better. He says this works better than, say, basing an existing model on knowledge base articles.
Understand your customers and initiate the right processes
The goal is for such a chatbot to understand what the customer wants and, based on that understanding, automatically initiate a process to resolve the customer's question fully automatically. In the case of refunds, such a process is largely predetermined, but situations always arise from time to time. Imagine someone received a broken product and complained. The company has always refunded customers based on previous support tickets with similar complaints. The domain model can decide to automatically refund the customer according to that policy.
Did you receive the wrong color dress?
AI chatbots bring efficient customer service and high customer satisfaction to organizations. However, not all customers are able to express their problems in words. If the problem is unclear, the chatbot often asks for clarification. In fact, a picture can say more than words. So Freshworks is also working on photo recognition. They showed a demo of a fictional clothing store where a person who ordered a red dress got a blue dress delivered. When a photo of the blue dress was sent to the chatbot, the chatbot immediately realized that there was a delivery error. It suggested sending the correct color for now.
Clearly, this is the future of retail. Not only can issues be detected and resolved sooner, but there is also a much lower sensitivity to fraud. When a customer submits an issue with a photo, suppliers have instant evidence, enabling a faster, more solution-focused response.
Freshworks is a leader in AI-powered software
Freshworks is currently updating its mission statement. The goal is now to deliver software that is powered by AI. In other words, AI will always be part of the solution. Mirchandani says Freshworks has taken the right technological steps in recent years to prepare for this.
All Freshworks solutions have been partially redeveloped on the new Neo platform. This cloud-native platform is made up of various microservices that allow Freshworks to innovate quickly. It also seamlessly integrates all Freshworks solutions and centralizes data into one customer data platform. There is one customer record with all data that all Freshworks solutions can access.
This allows Freshworks to innovate faster than competitors who are still using a different IT stack.
Developing AI to solve customer problems
Yamamoto is primarily concerned with the customer experience at Freshworks. She claims that innovation at Freshworks comes primarily from looking at customer challenges and how existing processes can be improved or further automated.
Adding innovation to business problems, with or without AI, ensures high customer satisfaction. At Freshworks, she is also looking at ways to make things work better together. For example, she has improved alignment between sales and marketing teams. Previously, both teams had reports that served specific sales or marketing functions. Now, the reports are more customer-based and serve both teams. This also improves alignment between sales and marketing.
Yamamoto says Freshworks has already made great strides, but there is still a lot of room for improvement. It's clear that self-service has a future in many industries, and AI can play a big role in that. It can greatly improve the customer experience and gain more loyal customers. But it also raises the bar even higher: without AI, Freshworks will lose out to companies that use it.
Yamamoto: “Freshworks' solutions are suitable for small and medium-sized enterprises and small business organizations.”
Yamamoto says Freshworks has a strong position in the market. The company's solutions are targeted at organizations with 10 to 5,000 employees and are available to all SMEs and small business organizations. Organizations can get started right away without needing an entire team of data and AI analysts.
TipFreshworks remains flexible and accessible as it embraces AI
