Driving Innovation and Unlocking Business Value with Gen AI

The global market for generative AI is projected to reach $668 billion in 2030, achieving a compound annual growth rate of 47.5 percent. This rapid growth is happening because enterprises are quickly learning how to apply generative AI to solve business problems such as serving customers more efficiently. At the same time, companies are learning about the risks of commercial gen AI tools such as ChatGPT. Increasingly, we’re seeing businesses embrace their own large language models (LLMs) to ensure that they meet their business needs and mitigate against the pitfalls of gen AI. 

Gen AI Is Solving Business Problems
When ChatGPT exploded on the scene in November 2022, businesses quickly began to experiment with the tool to do everything from write code to create content. What’s really remarkable is how soon businesses began to use generative AI to solve tangible business challenges in addition to experimenting. 

For example, retailers are adopting gen AI to assist with shopping in many ways. 

• Marketplace Mercari launched a shopping assistant tool named Merchat AI to suggest products. 
• Zalando recently announced the roll-out of a similar tool powered by ChatGPT, which assists shoppers in finding items through fashion terms or queries they pose. 
• Klarna announced a collaboration with OpenAI to present users with handpicked product recommendations. 
• Google recently announced a virtual try-on feature in the United States for select women’s tops from H&M, Anthropologie, Everlane and Loft. 
• Google is even using generative AI to virtually display items on a variety of real models, which helps shoppers visualize how the clothing would look on someone similar in size and stature to themselves.

In the consumer-packaged goods space, there is no shortage of brands applying AI to their businesses. 

• Coca-Cola, Colgate-Palmolive, Mars, and Nestle are using generative AI for everything from marketing to product ideation. 
• Nestle and other CPG firms are reportedly using a generative AI platform called Tastewise to validate new product ideas and generate market research reports. 
• Earlier this year, Coca-Cola launched marketing campaigns using generative AI, an example being the use of gen AI to create artwork and a “Create Real Magic” contest in which consumers could use AI to create original artwork using assets from the Coca-Cola archives.

Pharma companies are also using AI to automate and accelerate data processing. 

• A global pharmaceutical business is working with Centific to use generative AI to improve the way the company collects and analyzes data required to develop its portfolio of drugs. The collaboration is expected to produce results such as a significant reduction in manual annotation time by up to 70 percent, which enables more effective data-driven decisions. 

These are just some of the ways businesses are using generative AI to improve functions such as customer service and product development. TechTarget recently published a round-up of many more applications that demonstrate the power of gen AI.

Off-the-Shelf Large Language Models versus In-House Tools
As businesses embrace gen AI, they’ve typically been faced with the choice of using a commercially available large language model such as ChatGPT or developing their own. 

There are certainly advantages of using a commercial tool. The providers of these large language models do all the heavy lifting with technology development and management. But commercially available tools come with risks, many of which have made news headlines. They include:

• Security vulnerabilities: generative AI tools can be complex and contain security vulnerabilities. These vulnerabilities could be exploited by attackers to gain access to sensitive data or systems. For example, an attacker could trick an AI tool into generating code that contains malicious code.
• Intellectual property infringement: as noted, generative AI tools can be used to generate text, code, and other creative content. If this content infringes on someone else’s intellectual property rights, the user of the AI tool could be liable for damages. 
• Bias: generative AI tools are trained on data that is created by humans, and this data can contain bias. As a result, the output of generative AI tools can also be biased. This could lead to discrimination or other harmful outcomes.

For these reasons, businesses are taking a hard look at how willing they are to use commercial generative AI tools at all. Samsung famously banned employees from using ChatGPT over security risks. Indeed, because of privacy risks, Centific and the pharmaceutical client mentioned above do not use a commercially available LLM – opting instead to use the client’s model. But developing an in-house LLM also has downsides. For instance:

• Data management: Large language models require a massive amount of data to train. This data can be expensive to collect and prepare, and it can be difficult to find high-quality data that is free from bias. Not every business has the resources to do the data preparation and management needed.
• Computational resources: Training large language models requires significant computational resources. This can include powerful graphical processing units and specialized hardware like large language model chips. Training a large language model can also be very time-consuming, taking weeks or even months.
• Expertise: Developing a large language model requires expertise in natural language processing, machine learning, and deep learning. It is also helpful to have experience with distributed computing and large-scale systems.
• Cost: Developing a large language model can be expensive, especially when you factor in the cost of data, compute, and expertise.
• Interpretability: Large language models are complex and can be difficult to understand. This can make it difficult to explain how they work and to ensure that they are not making harmful decisions.

An Evolutionary Approach to Applying Gen AI
We believe that businesses can succeed by taking an evolutionary approach.

It’s totally fine to use a commercial large language model such as ChatGPT or Bard from Google as a starting off point for employees to learn how to use generative AI. But it’s important to do your homework first. Be aware of the risks and benefits. Understand how to use generative AI responsibly. Being restrictive with generative AI is not only unrealistic, it’s also harmful for business. Your employees need to be fluent in generative AI for your company to be competitive.

Eventually you will graduate to your own large language model to manage real-world problems as we have done as a company. This is a natural and desirable evolution as your business wrests control of your own data – and your customers’ data. Only an in-house large language model will offer you the control and safety you need in the long run.

But what about all the hard work required to train and manage an in-house large language model, or get an early start by leveraging Off-the-Shelf LLMs? This is where Centific can help. Our AI experts start by defining the scope of the business problem they want to solve, a proof of concept for a business unit, and a full-scale generative AI implementation. All of this includes configuring the right AI model applicable for a defined use case, reviewing the model against safe AI tenets, and much more. 

Learn about Centific's Generative AI Solutions and Services.