dialtone

Inspiration

We noticed that a lot of important work still depends on phone calls — reaching out to businesses, following up with leads, confirming appointments, making bookings, and collecting information. But making these calls manually takes a lot of time and often means repeating the same conversations again and again.

We wanted to build an AI agent that could actually pick up the phone and handle these conversations instead of just generating text.

What it does

Dialtone is an AI calling agent for real-world outreach and tasks.

You give Dialtone a goal, the people or businesses to contact, and the information it needs. Dialtone can make calls, have natural conversations, ask questions, respond to what the other person says, and complete the task.

It can be used for things like:

• Business outreach and lead generation
• Follow-up calls
• Appointment scheduling
• Bookings and reservations
• Confirming information
• Customer outreach
• And other tasks that can be handled over a phone call

The goal is simple: tell Dialtone what you need done, and let it handle the calls.

How we built it

We built Dialtone as a web-based AI agent platform using Next.js, TypeScript, Tailwind CSS, SQLite, and AI voice technology.

The application handles the workflow from the user's request to the actual phone call. The AI processes the conversation in real time, understands responses, and decides what to say next based on the task it was given.

We also designed the interface to make creating and managing calling tasks simple, so users don't need to manually manage every call.

Challenges we ran into

One of the biggest challenges was making an AI agent that could handle real conversations instead of following a fixed script.

People don't always answer questions in the expected way. They may ask questions back, change the subject, give incomplete information, or respond differently from what we anticipated. Making the agent flexible enough to handle these situations while still staying focused on the user's goal was a major challenge.

We also had to think about reliability, conversation flow, and making the experience feel natural rather than robotic.

Accomplishments that we're proud of

We're proud that we built a working AI calling agent that can go beyond simply generating responses and actually interact with people through phone calls.

Dialtone turns a simple instruction into an actionable calling workflow, allowing users to delegate repetitive outreach and booking tasks to an AI agent.

What we learned

We learned that building an AI agent for real-world conversations is very different from building a traditional chatbot. The agent needs to understand context, react to unexpected responses, and make decisions while the conversation is happening.

We also learned a lot about combining AI, voice, web applications, and automation into one complete product.

What's next for Dialtone

We want to make Dialtone more capable and reliable across different industries and use cases.

Our next steps include improving conversation intelligence, giving users more control over agent behavior, adding better analytics and call summaries, and expanding the types of tasks Dialtone can complete autonomously.

Our vision is for Dialtone to become a general-purpose AI calling agent that can handle any task that can be completed through a phone conversation.

Built With

• agents
• ai
• api
• call-e
• language
• large
• llm
• models
• next.js
• rest
• speech-to-text
• sqlite
• stt
• tailwind-css
• telephony
• text-to-speech
• tss
• typescript
• voice
• websockets