Despite the interest they are awakening, Artificial Intelligence and Machine Learning present different challenges. To talk about this, and respond to the challenge of implementing AI and ML, Byte TI organized a webinar that included the participation of Matías Sosa, OVHcloud Product Marketing Manager; Jose Ventura, IBM Data&AI Sales Specialist; Rafael Valdes, SAP Solution Advisory Manager Business Technology Platform; Carlos Polo, Director Line of Business Advanced Tech at Seidor and José González, Salesforce Senior Manager Solution Engineering who analyzed where the main challenges are.
Since the emergence of ChatGPT, artificial intelligence has sparked unprecedented interest among all organizations. Although its development is still in the early stages, much of what is currently offered on the market are Machine Learning solutions. The combination of both technologies offers companies advantages such as task automation, process optimization and the ability to make faster and more accurate decisions. But it cannot be denied that artificial intelligence has become an essential tool in the business portfolio. In this context, organizations are increasingly integrating AI solutions to optimize their strategies and improve operational efficiency.
An example of this is OVHcloud. Matías Sosa, his Product Marketing Manager, explains that the company is integrating Generative AI, “being a cloud company, we offer a complete set of platform-as-a-service solutions for companies. In terms of advantages, we work above all with open source solutions to improve all the reversibility that clients can bring and take their solutions, their models and offering, of course, important calculation capabilities, which is precisely what today it is being sued in the AI section.”
Much of what is currently offered in the AI market are Machine Learning solutions
Following this point, José Ventura, Data&AI Sales Specialist at IBM, emphasizes the company, “we base ourselves on the needs of organizations when using this technology. We try to cover the end-to-end of the AI lifecycle. We have an open source model. And in the end what we are looking for is the possibility of choosing the best model in each of the cases, well, in the end for each use case there is no good model for everything. You have to play with that.”
For his part, Rafael Valdés, Solution Advisory Manager Business Technology Platform at SAP, emphasizes that for the company this technology is not something new, “for almost 25, 30 years we have already applied AI to our business processes. We know, almost better than anyone else on the market: software like ours that is capable or is a leader in covering all business processes or business areas in a company, what SAP is centralized in is offering what we call our piece central AI, which is Joule, which is going to be our Copilot Cross across all the solutions.”
While Carlos Polo, Director Line of Business Advanced Tech at Seidor and José González, Senior Manager Solution Engineering at Salesforce, specify that their commitment to AI is directed towards the HR department, “we have made a generative AI model to simply through a Teams channel, being able to solve problems online and offload administrative work, in this case, to the HR department. In addition, we have done 50 three-hour training sessions, 50 workshops, with IBEX 35 and nearby companies. Because one of the things we saw is that there were very real expectations of AI.”
The customer experience
Through technologies such as chatbots, data analysis and automated recommendations, organizations can anticipate and meet their customers’ needs more accurately, as is the case with Salesforce. González explains that the company seeks to bring AI closer to business users to increase their productivity and efficiency. “This is done on three levels: first, through “out of the box” functionalities that users can activate directly, such as real-time assistance to contact center agents. Second, by providing tools for companies to define their own AI use cases securely. Finally, through conversational CRM, which includes a Copilot capable of understanding and executing tasks on the Salesforce platform.”
The same thing happens with retail, “in all sectors, we are observing that companies need help to respond in real time to customers. Preparing a campaign ahead of time is not the same as reacting quickly to possible queries. The ability to react quickly and efficiently improves the customer experience, which in turn increases brand loyalty and sales. And without a doubt, AI plays a crucial role in this sense,” says González.
In aviation, or in public service positions, Polo explains that having a call center that speaks 256 languages would be unsustainable in terms of costs. However, a generative AI that understands non-standard languages, such as Basque or Chinese, can intercept and translate calls, offering responses in the customer’s language. He also mentions how AI can help in bureaucratic processes, such as reviewing 2,000 applications for a public job offer. “Instead of manually reviewing each file, a generative AI can filter out candidates who do not meet the requirements, facilitating the work of officials and optimizing the process,” he summarizes.
Challenges in implementation
Matías Sosa from OVHcloud and José Ventura from IBM analyzed in detail the main challenges in implementing AI. Sosa highlights several critical problems: “Sometimes many organizations face different challenges related to the collection, cleaning, data organization and, of course, the privacy and regulatory compliance part. Security in AI models and data right now is crucial and must be addressed.”
José Ventura adds that “productivizing AI is not easy, that is, it is a challenge and we have to see how to address it,” he comments, pointing out the complexity of integrating AI into existing systems. Ventura highlights the heritage of the DevOps culture in the management of AI workflows, now called MLOps or AIOps: “It is all the operations necessary to automate all the flows of artificial intelligence so that it can once again be productive in the company ”.
From a different perspective, Rafael Valdés focuses on the speed of technological change and the need for deep understanding by IT departments. “What we found is that many times these departments are encountering technologies that advance very quickly, that are very dynamic and that change every day. And it is difficult for them to really have a deep understanding of what can or cannot be done,” he explains.
Protect data and ensure integration
Sosa highlights the importance of data anonymity to protect privacy: “On many occasions what is being done is to anonymize the data, even creating pseudonyms for it to guarantee and protect the people or companies behind that data.” . Furthermore, he highlights OVH’s commitment to compliance with regulations such as GDPR and the need to continually validate data and models to avoid bias and errors.
José Ventura, from IBM, emphasizes the importance of data security before any analysis: “Data is the greatest value that a company has, so we must ensure that this data is not used to train other models.” Ventura also mentions that IBM technology is designed for seamless integration with other systems, using open standards such as API-res.
From an openness perspective, Rafael Valdés comments that, “our generative AI offers a completely open architecture that can integrate with third-party LLM or third-party technologies.” Valdés assures that the context of the data will not leave the client’s business and that SAP has committees dedicated to continually review the security and ethics of AI.
For his part, Carlos Polo reinforces the importance of using non-confidential and masked data to guarantee security. “Data masking is important because it prevents this data from being linked to a person or a specific legal entity.” Polo also highlights the availability of on-premises and open source solutions that allow greater control over data.
To conclude, José González highlights the company’s philosophy regarding trust and data protection. “For us, trust is Salesforce’s number one value since it was born. Salesforce ensures that customer data is not used to train future models without permission and that any sensitive data is masked before being sent to the LLMs with which they have agreements.”
