Blog

Tharios: Artificial Intelligence for Community Management

6 Sep 2018 | Tech

This month, TradOnline would like to showcase a company undergoing heavy expansion. Their services may very well interest a number of our clients. Tharios offers a different look on a community management solution.

TradOnline: Hello, Christophe Bourgoin. To start, who are you and what do you do?

Christophe Bourgoin: I studied econometrics, and then I worked in the world of data – first in a bank, then in start-ups – on solving digital marketing and programmatic problems.

When my project before last finished, I decided to try me hand in an entrepreneurial adventure.

TradOnline: Tell us a bit about the Tharios solution.

Christophe Bourgoin: At the launch of Tharios, we offered our clients services for managing problems related to data. Then, quickly, problems related to publishing and gaining recognition on social networks emerged. So, we started working on this idea of generating texts. Thanks to Deep Learning, we found a solution to automate part of the tasks for creating content.

The need is real.

We have tested our solution with freelancers, start-ups, and big organizations. Their support motivated us to continue. And in order to create a truly innovative product, we gave up on the consultation part of Tharios.

Tharios has been incubating in Montpellier at the BIC (Business & Innovation Centre). There are 3 associate partners and around 10 people in total working on this project.

TradOnline: What do you mean by “generating texts”?

Christophe Bourgoin: Today, paid advertising on the Internet is showing worse results than in the past. The only big lever for gaining traffic is through SEO.

And with SEO, the written editorials, the content in some ways, is extremely important. But in order to have successful marketing through content strategy, you need time and ideas. And that is where Tharios steps in.

For the moment, we are concentrating on posts, i.e. that our text-generating bot only writes titles, descriptions/summaries for articles that our users decided to publish.

For example, one of our users finds an interesting article from Wired about AI on our platform, and he wants to share it with his network. Our text-creation bot will create a few propositions for posts on the three primary social platforms, LinkedIn, Twitter, and Facebook, adapting the titles and descriptions (format, semantics, etc.) while trying to optimize the text! The post written for LinkedIn will therefore not be the same as that written for Twitter for example.

We have restricted it, for the moment to writing posts, and not article, because the longer and more complicated the text is to write, the more difficult it is to obtain an effective assisted text.

Nakki Homepage
Home page of the platform created by Tharios

TradOnline: How does the Tharios solution work?

Christophe Bourgoin: Our solution is available as a SaaS platform. A virtual assistant is there to recommend content adapted to predefined criteria. The more you qualify these propositions, the more refined they will become until they correspond perfectly to your expectations.

Then, when you have chosen an article to share, a second virtual assistant with activate in order to share it on the networks you choose: Twitter, Facebook, LinkedIn, etc.

Of course, you will have the opportunity to modify these posts before they are published, because this is always left open for the client’s validation. But at least the post will have the right semantics.

And the last piece of our solution is performance monitoring of your publications as well as the analyses of your communities. This till will allow you to verify which prospects are the most important for you, if they are receptive to your publications, and to personalize your content.

TradOnline: Do you have any competitors?

Christophe Bourgoin: We don’t have any direct competitors. However, all media agencies or platforms with create posts at a low cost are our competitors.

TradOnline: How is your, quite different, solution viewed by your clients?

Christophe Bourgoin: We want to help brands that don’t have a content strategy to find a financially viable solution.

We are not looking to replace community managers, but to help them become more effective. Thanks to deep learning, we handle the tasks with little added value so that they can concentrate on the brand’s strategy, which is much more important.

However, it is clear that machines will always be objective and that they won’t have all your knowledge on a number of subjects. It will be up to you to add some anecdotes if this is part of your editorial strategy.

TradOnline: What are your objectives and your projects for 2018?

Christophe Bourgoin: We are currently testing a new approach with our tool. The idea is that the client enters the start of a sentence, and the bot will propose the end. We have put a lot of resources into R&D, notably for this solution.

We will also finalize our platform over the course of the summer to be able to offer a beta version to our clients by the end of the year.

And then, as R&D is expensive, we are also finalizing a fundraiser which should allow us to plan for a commercial release at the start of 2019.

TradOnline: What drives you personally every day?

Christophe Bourgoin: The group adventure keeps me going, and the fact that we are responding to a professional problem with an innovative solution, of feeling useful. We are not along and our solution is much awaited. It’s very motivating, all the more so because the companies who tested our product have been very delighted.

TradOnline: How are the communication and commercialization of your offering organized?

Christophe Bourgoin: We believe in digital marketing tools for ourselves as well! We think we’ll attract clients through our bot’s content production, through the use of data, technology, and our experts. 🙂

More seriously, our solution can help a large spectrum of organizations but for the moment we have decided to target freelancers, start-ups, and small businesses. These smaller organizations don’t have the same needs as large groups. They need to build their communities, and to help them, our bot will curate content. The advantage is that the bot will be more precise than you on a whole range of subjects for which you do not have knowledge. For me, it’s everything linked to the legal domain, for example.

As for pricing, we will be working on a subscription model. The price will depend, then, on the number of services that you choose: how much content per month, on which social networks, etc.

TradOnline: What kind of difficulties are you encountering on a daily basis?

Christophe Bourgoin: The principal difficulties that we encounter or technological. Given that Deep Learning is quite recent, we still have a lot to learn on the subject. Testing is complex, we are investing a lot into R&D, but it’s quite expensive.

And then, the infrastructure necessary for our tool needs to be quick and powerful.

That’s why we would like to raise some funds.

Thank you, Christophe, for presenting your project to us. The problems you face are similar to those we face at TradOnline as we are also experiencing a real revolution with the evolution of Machine Translations. Our concerns overlap, especially when it comes to analysing language. Our approach is the same: convert through teaching, but without overselling our solutions, because for Tharios, like TradOnline, Deep Learning solutions won’t replace humans.