Here we have the second instalment of the interesting talk about Artificial Intelligence that we had with Arian Ribell, CTO of Catalog Player and responsible for the implementation of AI in all the company's products.
[Here is the first instalment “Artificial Intelligence will be the ‘core’ of Catalog Player”]
Let's look at specific examples: for example, how does AI help us at the content level?
Point number 1 was to solve a problem for companies with a lot of content and that need to consult it quickly and conveniently. Searches that can logically be done in different languages. Let's say for example that we have 15,000 PDFs, or hundreds of manuals, or more than a million and a half tickets from customers…. It is clear that when searching for content it is often impossible. Sometimes we are not successful because we work with a search engine that does not understand what I am asking it, or because the person who wrote some content has made spelling mistakes. When we search for something in this way, the most common thing is that we do not find what we want.
And does AI solve it?
Right now I can already tell you that in all this AI works spectacularly. For example, the last case we have done recently was that of a client in the pharma sector, for which we have indexed around 100,000 records, most of which are PDFs. When we connected and named the client's AI, the results were very good, and I myself (who do not have in-depth knowledge of the products, sector and use) started asking it unusual contexts and then compared them with the real PDFs and content, and the result was almost perfect.
Did the result surprise even you?
Totally. It is that this marks a before and after when it comes to accessing that content. You no longer even have to ask it exactly what product you are looking for. As it is conversational, you can now ask it directly, for example «can you recommend a product for my hair?», or «what product do you recommend if I am a bit tired or sleepy today?».
Do you also see usefulness at the B2B level?
Yes, a lot, because what in B2C an end customer would ask you, in this case the salesperson themselves can ask. For example, “I have a customer who is requesting information about a product, can you tell me if it is recommendable for this customer profile?”, or a new salesperson who does not yet know the product very well can ask for help about what their potential customers are asking them. All this type of search works really well.
Looking at the functional diagram of Catalog Player's Sales Assistant, which parts of the implementation of Artificial Intelligence would you highlight most right now?
The search and recommendation part is impressive. The analysis part is the one we are finishing training and it is very important to structure the information well to avoid AI “hallucinations”, which fortunately are increasingly less frequent. That is why we are working with prompts and increasingly detailed contexts to narrow down as much as possible the type of response or format it should return. The part of Prompt Engineering that we are working on intensely is key. To avoid as far as possible that it makes mistakes, especially focusing on the data analysis part. And I insist, we are just finishing improving the analysis part, in the search and recommendation part it already does it well.
I can see you are very excited about the Sales Assistant project.
Yes, we are in a very important transformation process of our platform! Little by little we are building a new product to solve searches, data analysis, activity and functional graphs for the automation of planning and visit reporting tasks, whether by text or voice. Seeing how AI transcribes your voice request into a request to your webservices is surprising.
“We are experiencing a very important transformation process of our platform”
Which parts will reach interested customers first?
As I have already mentioned, firstly everything related to content. We still need to finish improving some automations at the product level but it is already a reality and works very well. Secondly, the CRM, which is underway and we will shortly launch the first Beta version. And thirdly, the KPIs and sales which we are also activating progressively.
It seems the change is unstoppable.
Yes, we are pivoting our entire product towards AI. AI is transversal because it is connected to any of these modules and so the way of interacting with our platform will no longer be only through apps and clicks, but conversationally, in order to carry out or generate reports, graphs and display specific components in widgets format depending on the type of action the AI must perform.
[TO BE CONTINUED WITH THE THIRD AND FINAL INSTALMENT…]
