Navigating the complexities of implementing and integrating GenAI
Sponsored - GENAI document implementation · Verdict

In This Article:

Interest in artificial intelligence (AI) and GenAI is booming, and businesses across multiple sectors are eager to use these technologies to personalize customer experiences, improve employee efficiency, streamline operations, or create new content, such as text, visuals, audio, and code. GenAI's scope spans a wide variety of use cases across all types of business operations and procedures.

In Spain, the GenAI market is projected to expand significantly. According to GlobalData’s GenAI forecast, the market opportunity in Spain is expected to surge from $19 million in 2022 to $425 million by 2027, reflecting a remarkable compound annual growth rate (CAGR) of 86%. This trajectory indicates a strong appetite for GenAI solutions among Spanish businesses, aligning with global trends in the wider adoption of AI technologies. However, several obstacles exist to effectively integrating GenAI into a company’s infrastructure, particularly when it comes to operationalizing the models to replace manual tasks and become stable processes. Document analysis is a key example.

"We already have different GenAI solutions for document analysis and automatic text generation, which can save companies a lot of resources and time while minimizing human errors, but only if we know how to effectively implement and integrate them into daily processes", says Rubén Granados, leader of the Data Science team at Telefonica Tech. "Implementation is not as straightforward as sometimes thought. It is not just about taking the trendy solution and launch requests. This will respond acceptably well in general, but for a GenAI solution to work in a specialized domain and for a specific task, it will be necessary to adjust and adapt it to the client's needs to obtain the best possible results."

Integrating GenAI

Implementing GenAI solutions into enterprise document analytics frameworks is a complex process and, for some companies, the implementation process takes more time and effort than they initially expected.The first obstacle that a company usually encounters is uncertainty about which large language models (LLMs) to use, a step that can hinder the decision-making process and delay implementation. There are major differences between generative models on the market, and it is necessary to understand them and choose the most suitable one for each type of use case based on the objectives pursued.

Working with open-source models can lower costs, but data security must be taken into account, and applications to specific use cases limited. A lack of internal expertise in deploying these types of models can also lead to mismatches, longer-than-expected implementation times, and unforeseen costs.