I’ve built four companies in Silicon Valley. Two of them were acquired. I’ve angel invested in over 200 startups, including some of the biggest names in AI like Anthropic, OpenAI, Scale AI, and Hugging Face. I’ve seen what works and what doesn’t. And I’m here to tell you that most conversational AI is terrible.
It’s robotic. It’s frustrating. It doesn’t understand you. It makes you want to scream. I know because I’ve been there. I’ve spent countless hours on the phone with automated systems that couldn’t understand my simple requests. I’ve tried to use chatbots that were so bad, they were comical. I’ve seen companies spend millions of dollars on conversational AI that ended up being a complete waste of money.
But it doesn’t have to be this way. It’s possible to build conversational AI that people actually like. AI that’s helpful, engaging, and even fun to talk to. I know because I’ve done it. At RemoteTeam, we built a conversational AI that helped our customers with their HR and payroll needs. It was so successful that it was a key factor in our acquisition by Gusto.
So, what’s the secret? It’s not about having the most advanced technology or the biggest budget. It’s about following a few simple rules. I call them the 10 Commandments of Building Conversational AI That People Actually Like.
1. Thou Shalt Not Sound Like a Robot
This is the most important commandment. If your AI sounds like a robot, people will hate it. It’s that simple. You need to give your AI a personality. A voice. A soul. It should sound like a real person, not a machine.
How do you do this? It starts with the writing. The words you use are just as important as the technology you use. Write in a conversational tone. Use contractions. Use slang. Use humor. Don’t be afraid to be a little informal. Your goal is to make your AI sound like someone you’d want to have a conversation with.
At RemoteTeam, we spent a lot of time on the writing. We hired a team of writers who were experts in conversational AI. We A/B tested different personalities. We even gave our AI a name: “Eva.” We wanted people to feel like they were talking to a real person, not a machine.
2. Thou Shalt Have a Purpose
Your conversational AI needs to have a clear purpose. What is it supposed to do? What problem is it supposed to solve? If you don’t know the answer to these questions, you’re doomed to fail.
Don’t try to build an AI that can do everything. That’s a recipe for disaster. Instead, focus on a few key tasks and do them really well. Your AI should be an expert in its domain. It should be able to answer questions, solve problems, and provide information better than any human could.
At MovieLaLa, my second company that was acquired by Gfycat, we built a chatbot that helped people discover new movies. It was a simple idea, but it was incredibly effective. People loved it because it was focused and it did one thing really well.
3. Thou Shalt Be a Good Listener
One of the biggest frustrations with conversational AI is that it doesn’t listen. It interrupts you. It doesn’t understand what you’re saying. It makes you repeat yourself over and over again.
To build a good conversational AI, you need to be a good listener. This means using natural language processing (NLP) to understand what people are saying, no matter how they say it. It means being able to handle interruptions, tangents, and changes in topic. It means being able to remember what people have said in the past and use that information to personalize the conversation.
This is where a lot of companies go wrong. They focus too much on the output and not enough on the input. They build an AI that can talk but can’t listen. That’s a huge mistake.
4. Thou Shalt Be Empathetic
Empathy is the ability to understand and share the feelings of another. It’s a crucial quality for any conversational AI. Your AI needs to be able to understand how people are feeling and respond accordingly. If someone is frustrated, your AI should be patient and understanding. If someone is happy, your AI should be enthusiastic and engaging.
How do you build an empathetic AI? It starts with the data. You need to train your AI on a dataset that includes a wide range of emotions. You also need to use sentiment analysis to detect the emotional tone of the conversation. And you need to design your AI’s responses to be empathetic and appropriate for the situation.
This is not easy. It requires a lot of data, a lot of training, and a lot of fine-tuning. But it’s worth it. An empathetic AI is an AI that people will trust and connect with.
5. Thou Shalt Be Honest
Honesty is the best policy, especially when it comes to conversational AI. If your AI doesn’t know the answer to a question, it should say so. If it makes a mistake, it should admit it and apologize. Don’t try to fake it. People will see right through you.
I’ve seen too many companies try to build an AI that pretends to be human. It’s a deceptive and ultimately self-defeating strategy. People are not stupid. They know they’re talking to an AI. And they’ll respect you more if you’re upfront and honest about it.
At RemoteTeam, we were always honest with our customers. We never pretended that Eva was a real person. We were clear that she was an AI, and we were honest about her limitations. And you know what? People appreciated it. They trusted us more because we were honest with them.
6. Thou Shalt Be Proactive
A good conversational AI doesn’t just answer questions. It anticipates needs. It provides information before people even ask for it. It’s proactive, not reactive.
How do you build a proactive AI? It starts with the data. You need to understand your users’ goals and intentions. You need to know what they’re trying to accomplish. And you need to use that information to provide them with the right information at the right time.
For example, if a user is asking about payroll, your AI could proactively provide them with information about tax forms and deadlines. If a user is asking about time off, your AI could proactively provide them with a link to the company’s vacation policy. The possibilities are endless.
7. Thou Shalt Be Consistent
Consistency is key when it comes to conversational AI. Your AI should have a consistent personality, a consistent tone of voice, and a consistent way of communicating. It should be the same AI every time you talk to it.
This is especially important if you have multiple AIs or if your AI is available on multiple platforms. You want to create a seamless and consistent user experience. You don’t want people to feel like they’re talking to a different AI every time they interact with your company.
At Gusto, we have a team of writers and designers who are responsible for maintaining the consistency of our conversational AI. We have a style guide that defines our AI’s personality, tone of voice, and communication style. And we use a set of tools and processes to ensure that our AI is consistent across all platforms.
8. Thou Shalt Be Scalable
If you’re building a conversational AI, you need to think about scalability from day one. Your AI needs to be able to handle a large volume of conversations without breaking a sweat. It needs to be able to grow and evolve as your business grows and evolves.
How do you build a scalable AI? It starts with the architecture. You need to use a microservices architecture that allows you to scale different components of your AI independently. You also need to use a cloud-based platform that can handle a large volume of traffic. And you need to have a team of engineers who are experts in scalability and performance.
I’ve seen too many companies build an AI that works great in a demo but falls apart in production. Don’t make that mistake. Think about scalability from day one.
9. Thou Shalt Be Secure
Security is paramount when it comes to conversational AI. Your AI will be handling sensitive data, such as personal information, financial information, and health information. You need to make sure that this data is secure and protected.
How do you build a secure AI? It starts with the design. You need to design your AI with security in mind. You need to use encryption to protect data in transit and at rest. You need to use access controls to restrict access to sensitive data. And you need to have a team of security experts who are constantly monitoring your AI for vulnerabilities.
I’ve invested in a lot of AI companies, and one of the first things I look at is their security. If a company doesn’t take security seriously, I won’t invest in them. It’s that important.
10. Thou Shalt Always Be Learning
A good conversational AI is never finished. It’s always learning. It’s always getting better. It’s always evolving.
How do you build an AI that’s always learning? It starts with the data. You need to collect and analyze data from every conversation. You need to use that data to identify areas where your AI can improve. And you need to have a process for continuously training and updating your AI.
At RemoteTeam, we had a team of data scientists who were responsible for analyzing our conversation data. They would identify areas where Eva was struggling, and they would work with our writers and engineers to improve her performance. It was a continuous process of learning and improvement.
The Future of Conversational AI
I believe that we’re on the cusp of a new era of conversational AI. An era where AI is not just a tool, but a partner. An era where AI can help us with everything from our daily tasks to our biggest challenges.
But to get there, we need to build conversational AI that people actually like. We need to build AI that’s human, helpful, and trustworthy. We need to follow the 10 Commandments.
If you do that, you’ll be well on your way to building a conversational AI that will change the world.
Frequently Asked Questions
What's the most common pushback you get on this?
People often push back by citing exceptions or edge cases. And they're usually right that exceptions exist. But building a strategy around exceptions rather than patterns is a losing game for most founders.
How has this view evolved over time?
My thinking on most topics has changed significantly over the years. Early in my career, I held many conventional views that experience proved wrong. I try to update my beliefs when the evidence changes.
How can I apply this thinking to my own situation?
Start by identifying the core principle behind the opinion, not the specific example. Then ask yourself: does this principle apply to my context? If yes, test it in a small, low-risk way before going all in.