Building a better chatbot with text annotation services
True Lark, an AI communications platform, turned to CloudFactory to speed product development of their virtual business assistant, Sasha. The CloudFactory team took on data labeling operations, analyzing and tagging customer conversations.
Retail
California, USA
11-50
90%
of Work Day Freed up for Client’s Internal Team
Personal service businesses, like dental offices, spas, and exercise studios, need to have someone to handle questions and book appointments while the staff takes care of customers. And helping customers after hours is becoming more important. These businesses also need a way to engage and close leads coming through ads, social media, and marketing campaigns.
True Lark offers an AI-driven solution that can be trained to answer questions, book appointments, make purchases, etc. all around-the-clock. The solution is a virtual business assistant known as Sasha.
The company is the result of Srivatsan Laxman’s desire to put his AI expertise to work in a new way. “All I’ve done in my professional career is AI so I started to look for interesting problems where people haven’t yet successfully realized the full potential and power of AI. There’s a sufficiently large whitespace (in the personal service sector) that seems exciting and accessible.’’
CloudFactory assists True Lark by annotating massive volumes of customer conversations, enabling the company to deliver uniquely powerful automated conversational experiences. The partnership has given the True Lark AI team 90% of their time back, which they now are able to use to drive AI model innovation and rapid productization.
Chatbots aren’t new, but they are often limited in what kinds of questions they can answer, or tasks they can manage, and they often lead to rigid and frustrating user experiences before kicking the inquiry to an actual live person to help finish (or in some cases restart) the task.
“A lot of the chatbot conversations are very shallow in that they are a one-two punch. There's not a whole lot of context. The depth and the conversation is still, at best, one or two turns. You want the AI to hold the conversations for several turns,’’ Laxman explains.
By focusing on a couple of industries, True Lark felt it could bring that full range of necessary context. But to do that, it needed labeled data to build its models, and for that it needed help from CloudFactory.
To build models, True Lark needed to analyze and tag conversations to train the AI that makes Sasha work so seamlessly. It’s time consuming work. “If I spend my time or my team members’ time doing this, then we're missing out on some other critical aspects of product building. And that's when I started to look out for options like CloudFactory.’’
It’s hard for a company like ours to build and scale teams for large data annotation tasks. Especially because there can be significant churn in that workforce. People leave because the task is repetitive and it is hard to keep these teams motivated. It is much easier with CloudFactory.
Srivatsan Laxman
Founder and CEO
Laxman quickly realized that while True Lark could potentially build out a full-fledged data annotation team, he didn’t have the time to hire, train, and monitor the work being done daily.
“It’s hard for a company like ours to build and scale teams for large data annotation tasks. Especially because there can be significant churn in that workforce. People leave because the task is repetitive and it is hard to keep these teams motivated. It is much easier with CloudFactory.’’
True Lark’s CloudFactory team is trained to understand the subtleties in language required to accurately tag the text needed to train natural language based applications.
True Lark spent time early on explaining to their CloudFactory team what needed to be done, from annotating conversations to categorizing sentences. CloudFactory took it from there, reviewing large batches of text message conversations and applying labels based on the improvements True Lark is trying to make to Sasha.
Those improvements include noun recognition and time identification. CloudFactory reviews conversations to determine if a certain word is an intended appointment time, requested staff member, or specific service type. The team then determines if a client is proposing, accepting or rejecting a time, or making a conditional request (“I can do any time except”).
The result for True Lark: “I have a partner that I can go to for scaling up my labeling efforts. I can churn out new models because I can get the data prepared quickly.”
Along with giving back True Lark’s AI team 90% of their work day, he says he’s found an arrangement that works long term. He’s having success testing and training the model, reporting positive feedback to CloudFactory.
“We’ve hit on a working relationship that is sustainable. There is this intense commitment, on CloudFactory’s part, to making it work. We had a couple of situations where I wasn’t sure something was doable, but CloudFactory made it possible. That’s cool.”
We have 10+ years of experience helping our clients focus on what matters most. See what we can do to help your business.
Businesses of all kinds are scaling customer service with the help of robust, well-trained chatbots and virtual assistants.
Heretik wanted to make legal contract review less tedious so they created an AI-fueled solution. Learn how CloudFactory helped bring their software to life.
Customers expect timely support and responses from the brands they use, which often requires the action of real humans.
Fill out this form to speak to our team about how CloudFactory can help you reach your goals.
