Last year saw interest in agent-based AI approaches, such as AutoGPT. This year, we’re seeing startups implementing some of these “let AI do the work for you” use cases. One example is Mindy, who introduces herself as your executive assistant via email.Thank you for reading this post, don't forget to subscribe!
Using Mindy is pretty simple. You can email the assistant at firstname.lastname@example.org with a question such as “I want to know about all the events happening in my area (you must specify the area) this weekend with links to book tickets. or “Can you provide me with Apple’s revenue numbers in a table for the last eight quarters?”
Mindy searches the web for this information and responds with an email after a while.
Currently, Mindy is free to use as the company is in the process of determining the right price for the service.
Why did the founders create Mindy?
Mindy was created by Yu Pan, Cuong Do and Benoit Berthoux last year. Pan was one of the founding engineers of PayPal and YouTube. While Do joined PayPal in 2002, he was a founding engineer at YouTube. Berthoux first worked in consulting, notably at Capgemini and Accenture.
The company raised $6 million in a funding round led by Sequoia Capital (with Roelof Botha as lead partner) and Founders Fund.
“The germination of the idea came in 2022, while I was finishing my work in the Crypto field and Benoit was finishing his initiative. AI was behind the growing popularity of models like GPT-3.5. It took a lot of rapid engineering to achieve anything with these models. So we thought that since these models are supposed to analyze language, the tools shouldn’t require a lot of thought before using them,” Pan said in a call with TechCrunch about the company’s creation.
Berthoux said that when Mindy conducted market research, it found that people use apps like ChatGPT but forget about it. Notably, ChatGPT still has over 100 million active users despite this.
The company began testing Mindy in “semi-stealth” mode in November. But he didn’t say how many users interact with Mindy.
Choosing Email as a Differentiator
Pan said the company plans to rely on email because it has a higher “signal-to-noise” ratio in terms of information and also focuses on specific topics.
“Email is text-driven and our LLMs are text-driven. So we thought it was a great wedding. The way we look at these media is like we’re trying to marry basically the oldest technology in the book with the newest technology,” he said.
The startup says many of its users – ranging from financial analysts to students and defense contractors – use the bot to conduct research. And because the AI assistant works through email rather than a web interface, they don’t expect an instant response.
The company believes that because email and the conversational tone of the medium are familiar to people, they don’t need to change their workflow to fit the product into their lives. Plus, you can use it with a group of other people when planning or ideating.
The company said that due to the asynchronous nature of email, the agent may have more autonomy or automation to get you relevant results.
While email is the company’s main focus, it has also tested prototypes in other delivery challenges for different use cases.
The road to follow
The company currently has a team of 10 people and wishes to recruit staff to accelerate its growth with engineering training. It also explores different monetization models, including subscriptions, request limits or charging some of the purchases made. However, it has not been finalized on a specific approach.
The agent-based approach
As my colleague Devin Coldewey wrote in his year-end article, we’re looking at more AI-focused startups that take an agent-based approach. Most recently, The Browser Company, which makes Arc Browser, announced that it was developing an AI agent that browses the web for you.
One industry executive, who did not wish to be named, said startups like Mindy trying to build an agent will need to create enough differentiation to not be overshadowed by ChatGPT. They said these tools will need to deliver very high value to customers and find a way to acquire users at a lower cost.
Hemant Mohapatra, partner at Lightspeed, believes AI assistant-based startups need to focus on user delight and reliability.
“Outstanding consumer apps often experience rapid growth in their early stages, fueled by strong consumer satisfaction and word-of-mouth referrals,” he said. Mohapatra noted that for assistants to process requests end-to-end, they will need more user data, which means they will need to focus more on compliance and security.