With all the hype about Generative AI, it’s hard to know how enterprises are actually using this new technology. 

Let’s start with the basics. ‘Artificial Intelligence’ has been around since the ‘50s, when computer scientists began to create intelligent machines that can replicate or exceed human intelligence. Machine learning and deep learning followed. Gen AI, in simplistic terms, is a technology where an AI model can generate new content in text, image, audio or video format using a natural language interface. Behind the scenes a model tries to predict the next word or pixel, based on the large datasets on which it has been trained. 

Drilling down into the technicalities, Generative Pre-trained Transformers, commonly known as GPT, are a family of neural network models that uses the transformer architecture. GPT models enable the creation of human-like text and content (images, music, and more) in applications, and also offer the ability to answer questions in a conversational manner. 

Gen AI is based on ‘transformers’ – neural network architecture that relies on the parallel multi-head attention mechanism – combined with foundation models (which are unsupervised learning, large models). These foundation models are usually created for a specific purpose – one for code, one for research summaries, one for chat bots, and so on. The popular GPT-3 and DALL-E are models that are built on top of the foundation models. Since enterprises no longer have to build custom NLP models for each domain, they can accelerate the AI development lifecycle, and shrink their time to value from months to weeks. 

Adoption of Generative AI by Enterprises 

Adoption of Gen AI is off the charts, for all users. According to Similarweb, ChatGPT currently has over 100 million users. Organizations across most industries are using GPT models and Gen AI for Q&A bots, text summarization, content generation, and search.

Source: Exploding Topics

According to a recent Forbes Advisor Survey, almost all business owners (97%) believe that ChatGPT will help their business. Within organizations, 60% of those reporting AI adoption are using Gen AI. Enterprises also note the risks associated with Gen AI, with 30% listing inaccuracies as the most concerning issue, similar to the percentage of organizations concerned about security related issues. Yet only slightly over 20% of companies have risk policies in place for Gen AI, according to McKinsey. Even these companies may only be looking at specific scenarios and not at the broad implications of misinformation and inaccuracies. 

How companies are using generative AI

One of the impacts of ‘Gen AI’ versus ‘traditional AI’ is its impact on workers and roles. 

Whereas traditional AI could only be used by those with deep skills in technical areas like machine learning, data science, or robotics, Gen AI can be utilized by almost any user, although it still requires highly skilled people to build Large Language Models (LLMs) and train generative models. Surprisingly, less than a third of respondents in the McKinsey study say that their organizations have adopted AI in more than one business function, suggesting that AI use within organizations remains limited in scope, notwithstanding the exponential growth of Gen AI. 

Considerations in Adopting Gen AI 

Data Management 

Data management is the top technical inhibitor to AI/ML according to a WEKA 2023 trends report. Enterprise IT and data architectures are straining under the weight of ever-expanding AI workloads, with one of the primary sources being data management issues. These include managing the enormous volumes of data required to train AI models, and additional challenges arise from an increasingly complex ecosystem of deployment locations, use cases and stakeholders. Different data types pose challenges, with the unique requirements for structured data, unstructured text and streaming/real-time data. 

Ethical considerations, inaccuracies and what to do about them 

Despite its incredible potential, much of which we may not be able to see at the moment, ethical considerations regarding AI-generated content abound. AI-based decisions are susceptible to inaccuracies, and to embedded or inserted bias in the models they have been trained on, and in the other processes that shape their responses. Generative AI adoption within enterprises should ensure responsible AI frameworks and guidelines to protect the employees and the company from unintended consequences and harm. 

In tandem, organizations must implement robust data security and privacy measures that protect sensitive information and limit the company’s exposure. As Generative AI for enterprises continues to evolve into more sophisticated and versatile offerings, it may become more closely integrated with other AI techniques, leading to new horizons and capabilities. 

Generative-AI for enterprise applications

Organizations that want to use Gen AI to jump-start an AI-first approach, find that they need to adopt a more open and flexible mindset. According to Gartner’s GenAI Planning Workbook, only 10% of organizations have deployed AI techniques in more than a few business units. Gen AI shows promise of offering more correct responses than traditional searches since the GPT engine offers fast, comprehensive answers based on extremely broad sources of data. This saves employees time researching and generating content, enabling them to concentrate on personalization and enabling higher accuracy. 

An AI Clinical Intake Tool: A real life example of a Generative AI enterprise  implementation

The Sourasky Medical Center Emergency Department leverages Kahun, an evidence-based clinical reasoning tool for physicians. They incorporated Kahun’s ChatGPT chatbot into their existing AI system to create a triage tool. The bot obtains the patient’s symptoms and history in natural language, leaving more time for the medical professionals to focus on the treatment plan during visits, reducing fatigue and burnout among the clinical team.

Controlling retail costs with Gen AI 

Retailers are utilizing Gen AI to reduce losses from theft. Gen AI algorithms​​ help detect anomalies in patterns from customer preferences to fraudulent transactions — quickly and at scale, with fewer false positives. This results in lower customer service costs and improved customer experiences. These tools can also issue direct responses to customers at checkout to quickly resolve issues and reduce the need for manual interventions.

Last Words 

Enterprises have begun wading into the Gen AI waters, such as tweaking an LLM with enterprise data and domain knowledge, to create a sophisticated app that can support personalization with improved accuracy. Gen AI is better able to handle complex requests, route the requests to the correct applications, and continuously monitor for data quality and integrity.

Coming soon: A deep-dive into the technicalities of how enterprises can leverage their data with Gen AI to obtain true business value.