Skip to content

TRURL is here! 🎉

Author: VoicelabVoiceLab

White and orange cartoon robot with a letter T on its head and headphones, next to a phone call icon on a blue background

We are excited to announce the release of the Trurl model, a conversational AI solution designed to provide users with accurate, detailed, and polite responses to their inquiries that works similarly to ChatGPT. Trurl takes natural language processing and machine learning to the next level, offering an unparalleled degree of accuracy and relevance in its responses.

With the Trurl model, we are delivering on our commitment to providing the highest quality conversational AI solutions to our clients and partners. Whether you are a large enterprise, a small business, or an individual user, Trurl is designed to meet your needs and exceed your expectations.

Advantages

We created Trurl with business purposes in mind.

  • Data privacy: It can be installed on-premise on the client’s server,
  • No limits: If you install it on-premise there are no monthly usage limits.
  • Well-adjusted to your needs: For bigger budgets we can securely finetune it for you on your data or for your task, For smaller companies you can “train” it yourself simply by uploading your documents and website address (no programming skills needed!),
  • Knows business: It was trained on speech transcripts, books, and articles, for many conversation-related tasks,
  • Affordable prices: Prices depend if you are using the model on-premise or on our servers. In both situations prices are affordable.

Use cases

In what follows, we will provide an in-depth overview of the Trurl model, highlighting its key features and capabilities. We hope that you will find this information as exciting as we do, and that you will join us in exploring the possibilities that this innovative solution has to offer.

In this post we presented six sample use cases:

  • Detect arguments and objections,
  • Rate conversations & help make better calls,
  • Search through documents quickly (Information extraction),
  • Text summarization,
  • Reason for calling detection,
  • Creating social media posts.

Check our TRURL here for free: https://trurl.ai/

To begin with let’s see what Trurl has to say:

Trurl chat: asked for an introduction, Trurl presents itself as an AI language model developed by Voicelab

So let’s see how Trurl responds to our tasks.

Detect arguments and objections

Prompt:

“As a customer service quality controller, rate the following conversation. Was it an effective sales conversation? Did the client have objections during the conversation? Did the agent use arguments against the client’s objection?
conversation:
[CONVERSATION]”

Trurl chat: after a call about a broken returns form, Trurl rates it effective and notes the client's objection to the fee

Trurl handled the assessment very well. As you can see, it covered the conversation and summed up the customer service agent’s approach in this conversation.

Maybe it can advise us?

Rate conversations & help make better calls

Calls rating is an important part of any customer service operation. It allows businesses to evaluate the quality of their customer interactions and identify areas for improvement. However, calls rating can be a time-consuming and subjective process, especially when done manually. To improve the process, businesses can consider implementing a more automated system that uses AI and machine learning algorithms to analyze call data and provide more accurate ratings. Additionally, providing agents with real-time feedback and coaching can help them improve their performance and reduce the need for rating calls.

We trained Trurl to rate the conversations as well using our own call-center data with our team of experienced annotators.

Trurl chat: Trurl describes its ability to analyze conversations, including their context, sentiment and intent

Now let’s see:

Prompt:

“Can you advise how the agent could have handled this conversation better?”

Trurl's numbered list of five ways the agent could have handled the call better, from addressing the cost to showing empathy

It’s great, isn’t it?

Trurl was trained to handle both quick calls and longer conversations. Here is the example of a longer and more demanding call:

Prompt:

“Rate effectiveness of conversation on a scale from 0 to 5.
[CONVERSATION]”

Trurl rates a longer bank call 2 out of 5 and advises more empathy and a better solution, such as waiving the fees

Overall, by using AI-powered tools to evaluate conversations, businesses can automate the process of analyzing conversations and generate objective reports on the effectiveness of each conversation. This can help to save time and resources, and can provide valuable insights into the areas where agents are performing well and where there is room for improvement.

Prompt:

“How would you rate this conversation in terms of the agent’s empathy with the client on a scale of 1 to 10? Did the agent correctly resolve the issue? What was it?
conversation:
[CONVERSATION]”

Animated Trurl chat: a bank card call rated 8 out of 10 for agent empathy, with a new card and a fee refund as the resolution

And the last example for today:

Prompt:

“As a customer service quality controller, rate the following conversation. Was it an effective sales conversation? Did the client have objections during the conversation? Was the agent kind and understanding?
conversation:
[CONVERSATION]”

Trurl judges a call about long wait times as ineffective at resolving the issue, but finds the agent kind and understanding

Besides evaluation of conversations, Trurl can do way more. Let’s dive into other useful tasks.

Information extraction

Thanks to the ability to extract information, you can quickly and easily process this data and use it for other purposes, e.g. you can use our Trurl to quickly analyze files, documents or websites.

Animated Trurl chat: from JSON sales records, Trurl lists the products sold with their category and sale date

Summarization

Summarization is invaluable, condensing only key information for efficient communication, saving time and enhancing productivity.

Trurl is able to efficiently summarize the content of conversations:

Trurl summarizes a telecom call in which a customer cannot log in to their online account and is told to register again

We can continue the conversation, ask further questions referring to the previous message.

Trurl answers a follow-up on how the agent could have done better, judging the summarized call as handled well overall

Reason for calling

Have you ever wondered why your customers contact you? Analyzing conversations and getting to know the most common reasons will give you valuable information about what they most often have problems with and what their expectations are.

Trurl reads a bank call transcript and names the reason for contact: a 30 groszy fee charged for an internal transfer

Social media posts

Do you sometimes lack creativity? Let Trurl help you get your customers’ attention!

Trurl writes an emoji-filled social media post promoting summer dresses for women with 20% off
Trurl writes a social media post

And way more! Contact us and we will build a model for you trained securely on your infrastructure & your dataset.

…but who was Trurl?

„Trurl is a robotic engineer, known from Stanislaw Lem’s novels „Cyberiad” and” Tales of the Robots „called a „constructor” with almost godlike abilities. In one of the stories, he creates a machine called „Elektrybalt”, which by description resembles today’s GPT solutions. We named the first Polish GPT model after the illustrious writer and his immeasurable imagination condensed into the character Trurl.

Author: Patrycja Biryło

  • Vera Is Coming. A Modern AI Agent!

    Vera is an AI agent for sales and customer service teams. It analyzes phone calls, email and calendars, prepares summaries and offers for approval, and flags the cases where a sale or a support request needs attention.

  • Generating paraphrases

    We use paraphrases every day, but who among us has ever thought about using them? Paraphrases are used, among others, in search engines: they help the customer find the product more easily, suggest other ways to search.