How can we put generative AI to work?

Aug 22, 2024

finance
IT
operations
artificial intelligence & RPA

We all know generative AI or genAI has far-reaching implications for arts and entertainment, and even in healthcare and education the impact is palpable. But what about day-to-day business challenges? Can this omnipresent technology really make life easier for you and your stakeholders, or is it all just marketing speak? Our in-house AI experts Sven Arnauts and Wouter Labeeuw discuss the challenges – and opportunities – of making genAI work for you and your organization.

“When it comes to deploying generative AI, there are roughly two kinds of organizations and leadership teams,” Sven starts. “Those with a clear challenge who are curious about genAI’s ability to fix it, and those who are captivated by its allure and acquire it eagerly without reflecting about useful application for them. When it comes to leadership teams, there are some who believe genAI technology is a silver bullet that will solve all their problems, while others fail to see its potential. That’s why it’s so important to raise awareness.”

Demistifying genAI

Broadly speaking, genAI is used for:

unlocking knowledge and information. Users can converse with their organization’s knowledge base through natural language to get an immediate answer (instead of just being pointed to a specific file).
generating texts, documents images and videos. Users can ask the AI-tool to generate new content, e.g., executive summaries, emails, policy briefs, etc., based on historical files and information from the knowledge base, with simple natural-language prompts.

“Here, the AI model relies on a knowledge base to retrieve and infer answers or generate content,” says Wouter. “With advanced technology, it’s possible to talk directly to data as well. This means users are able to communicate directly with the organization’s vast repositories of data – structured and unstructured alike.”

While the difference might seem subtle, the consequences are far-reaching. Wouter: “The ability to talk to data directly democratizes data analytics. In other words, it enables a wide range of non-technical users to extract actionable insights, using nothing but simple prompts. The potential impact on innovation and efficiency across different domains is huge.”

Under the hood: full-text and vectorial indexation

While genAI models are extremely proficient at retrieving and inferring information from high volumes of data and documents, their success still depends on the indexation of all that information.

This can happen in several ways, for example:

Full-text indexation is how traditional search engines query documents and pages: by creating a searchable catalog of all the words it contains. It’s the perfect approach for precise, keyword-specific searches.
Vectorial indexation converts text into numerical vectors using embedding-models (OpenAI Ada, Google BERT, Mistral Embeddings, etc.). Each word or phrase is a point in a multidimensional space, and the distance and direction between these points represent their semantic relationships. This approach is particularly effective when semantic meaning and intent are more important than specific keywords – for example in natural language queries.

“Vectorial indexation is crucial for genAI models and chatbots to work,” Wouter continues. “It allows large language models (LLMs) like GPT to capture nuances, context, and even user sentiment. This enables a more accurate and contextually relevant understanding of the prompt.”

Vectorial indexation is a complex task that requires a deep understanding of the data, the objectives, and the LLMs involved. “Since different models have different strengths and weaknesses, they may produce different results as well. So, for vectorial indexation, you need more than just a basic knowledge.”

Democratizing NLP capabilities

GenAI is taking over tasks that were traditionally addressed by NLP. “Large, pretrained models like GPT 3.5 and 4.0 can understand and generate natural language at an unprecedented level of sophistication,” says Wouter. “As a result, they can now handle challenges that were once the unique domain of NLP, including translating, writing, summarizing, generating code, and more. Plus genAI has made these capabilities accessible to a wider, non-technical audience as well.”

GenAI in your business: 5 use cases

Beyond the hype, genAI can be a workhorse for efficiency and innovation even in the most traditional sectors. Here are a few examples.

1. Streamlining operations

One of the immediate benefits genAI brings to the table is automating routine tasks. Sven: "From sorting emails to managing inventory, genAI can take over tasks that are time-consuming and prone to human error. This frees up your employees to focus on more complex and creative tasks."

2. Personalizing customer interactions

And what about personalizing customer interactions? "GenAI can analyze customer data and use it to create highly personalized experiences," says Wouter. "Here, the point is not to take the human touch out of the equation, but to augment it to be even more effective."

3. Risk assessment and management

GenAI's ability to sift through large datasets can significantly improve risk assessment as well. Sven: "The ability to identify patterns and predict outcomes with high accuracy is invaluable for financial forecasting or to spot potential operational bottlenecks before they turn into major problems."

4. Enhancing collaboration

GenAI can also streamline the way your teams collaborate. Wouter: "Imagine a virtual assistant that schedules your meetings, gives you a summary of the key points from previous interactions, and suggests agenda items based on ongoing projects.”

5. The marketing edge

Marketeers love GenAI. And it’s not hard to see why. "Content generation, market analysis, and customer segmentation are just the beginning,” says Sven. “GenAI can also predict trends, enabling marketing departments to be proactive rather than reactive."

We start every workshop with a simple question: who are your stakeholders, and what is driving them crazy?

Sven Arnauts, senior manager Data & AI at delaware

To genAI or not to genAI

“The key to making genAI work for you is to start small and scale responsibly,” both Sven and Wouter conclude. “The first, most important step is to identify specific pain points and apply genAI solutions incrementally where they make sense. It's a transformation journey that, if navigated thoughtfully, can lead to significant competitive advantages.”

Through ideation and inspiration workshops, Sven and Wouter try to build a better understanding of genAI and guide companies towards concrete use cases and experiments. “We take care not to steer organizations towards genAI just because it’s a hot topic. In fact, we start every workshop with a simple question: who are your stakeholders, and what is driving them crazy? Sometimes, that turns out to be something genAI can solve, but it can also be a data platform or even a data governance issue. That’s why we always bring an industry expert along who understands the organization’s unique challenges and speaks their language. In the end, our goal is to offer the best possible solution.”