Overview

YAAC (Yet Another AI Chatbot)

In a nutshell, Persona Studios makes AI Chatbots (we call them personas) to not only answer user questions trained on your company data but also to automate repeat conversations. But what exactly does that mean to automate repeat conversations? Don’t other AI Chatbot companies do the exact same thing?

When we started Persona Studios in 2023, the challenge we sought to solve at Persona Studios was enabling longer conversations. Up to that point, all of the AI chatbots we used had a very clear paradigm: One user question → one AI answer. If you want more, you need to keep asking, and you need to know the right followup questions to ask as well as . In this paradigm, the user “owns” the conversation and must lead the AI to get to proper resolution / success.

Ask a question like “Where are the best cites to live in the world?”, and other AI systems would give you an answer based on crawling data like happiness indexes and top-10 livable city articles written by travel websites.

But if you’d ask “How do I move to Singapore?” for example, we found the answers in every system extremely lacking. At most, we would get a high-level overview of the kinds of things we should consider (e.g. getting a visa, finding a new job, moving our existing stuff, finding a new location). What we really wanted was: One user question → multi-part answer where the AI begins a series of questions and answers. In this paradigm, after the user asks the first question, the AI “owns” the conversation and leads the user without having the end user keep asking probing or clarifying questions.

The next obvious attempt was to improve our prompts to something like “Play the role of a moving expert and walk the user step-by-step through the process of moving, asking initial clarifying questions and then walking them through each of the steps.”

Every time we used a single system prompt to shoehorn and enforce a longer multi-step conversation structure, the conversation would inevitably break down, no matter how many tokens we would feed it.

What we felt was missing across the board, no matter how we prompted, was the ability to create conversation structures that we knew 100% would: Maintain an overall agenda after thousands of repetitions Not only be able to be used for personal consumption, but to help businesses that wanted to automate more than just simple one-off support questions. When these conversation structures were initiated, the AI would ask the right questions and take the user on a journey to complete a large, multi-step process rather than forcing the user to always ask the right question. Used AI where possible, but still leveraged non-AI code to enforce the agenda regardless of if the user tries to derail the conversation or inject a new prompt. No need for manual intervention. Many other systems that used AI would still very quickly “call for help” by referring the user to a contact number, an external webpage, or to a human agent. We felt like if we built this right, there would be no need to redirect to an external site or human to complete the user’s request. No need for coding, and to really mean it. No need for the user to have to know which model to select, or to write prompts from scratch with rules for how to handle the conversation, or to use a node editor to build workflows that looked like spaghetti just to get a working chatbot. If we did it right, the user could effectively use a single google doc to describe how they want the call to run, and our agenda builder would do the rest.

The problem now was, this conversation structure builder had to work for any vertical, for example: A recruiter who wanted to automate the first screening call that would qualify a candidate to move forward in the interview process. A restaurant that wanted to automate online orders, table reservations, or inbound vendor requests. A hotel that wanted to book customers and provide support during their stay. A small startup with high inbound that wanted to qualify inbound B2B customers.

As we started down this road, the word “agenda” kept coming up in our mind. For most businesses, people already create agendas to describe the order and sequence of discussion topics when they send an invite out for a call. This agenda typically covers things like intros, discussion topics, key questions or data points that need to be covered, and leaving room for the other party to ask any questions they may have. So we started to call our conversation structures “agendas” to mirror an existing, fairly unstructured process but one users already were likely familiar with (plus “conversation structures” was too long to say ove rand over).

So how could we create a conversation structure builder that could abstract these conversations into building blocks (almost like Legos) where any business in any vertical could define the conversation with. We call these building blocks “Agenda Item Types” and have built each one almost like a separate product entirely in itself. Discussion / Debate items have enough features in itself that rival companies that solely focus on screening candidates. Cart items have enough features that rival AI e-commerce companies. And so on…

Most importantly, we set out to make an AI Chatbot that would 100% pass a Turing test. No selectable pills, no “An agent will be with you shortly” bullshit, and when a user wanted more than an answer to a simple question but wanted to have a longer conversation like apply for a job, order a meal, have an inside sales call.

So back to our initial question… what does it mean when we say we “automate repeat conversations” and how is this different than other AI Chatbots?

Our persona already comes with the same features that other AI Chatbots have when it comes to omnichannel, multilingual, and giving answers based on training data from a company.

But the ability to define a longer agenda that feels human, that can be interrupted and still pull the user through the agenda, that can ask dynamic followups and not just read off a list of questions, and ultimately send that conversation to any CRM, and that this would work for any company in any vertical for any use case, we have not seen this literally anywhere else.

What is a Persona?

A Persona is an advanced AI assistant that goes beyond traditional chatbots. It not only provides answers based on a knowledge base but also leads structured conversations. For instance, a Persona can conduct a screening call, ask relevant questions, engage in discussions, and schedule follow-ups, effectively mimicking human interaction.

Personas have the following characteristics:

Basic info: Name, role (e.g. “You are Kevin Davis, CEO of Persona Studios, and can help users either get answers about your company or make a -persona themselves), style of communication
Omnichannel: Integrate with platforms like WhatsApp, Facebook Messenger, SMS, Web, Web Widget, Slack, Discord, and more.
Multilingual: speaks 30+ languages and can understand even more. Persona owners can also read and monitor conversations in the language of their choice.
Multimodal: accept images, resumes, or other file uploads as well as display text, images.
Voice: When communicating back with users, personas can use text only, voice notes, or both.
Knowledge:
Agendas: defined conversation structures that can screen candidates, making bookings, gather feedback, lookup customer info, run first round interviews, and can not only ask followup questions but allow the end user to ask questions as well.
Privacy: GDPR compliance

Key Features

Customizable Agendas: Define structured conversations tailored to specific business needs.
Multichannel Support: Integrate with platforms like WhatsApp, Facebook Messenger, Slack, and more.
Voice and Text Capabilities: Communicate naturally through speech or text.
Data Privacy: Ensures user data is secure and not used for model training.
Performance Tracking: Monitor interactions and effectiveness through an Activity Dashboard.

What is an Agenda?

An agenda is a defined conversation structure that has different agenda items (e.g. Convey, Gather, Debate, Q&A, Cart, API, Schedule) that can be used to automate any repeat conversation, especially first conversations. We go into a lot more detail later in this guide when it comes to preparing to create your first agenda, creating your first agenda, what are the different kinds of building blocks or agenda item types, and more.

Use cases

Persona Studios caters to diverse industries, including:

Recruitment: Automate candidate screening and evaluation.
E-commerce: Assist customers with product inquiries and order placements.
Customer Support: Provide instant responses and escalate complex issues.
Market Research: Collect and analyze qualitative and quantitative data.

Questions?

For more information, visit Persona Studios or contact our support team at hello@personastudios.ai.