Reimagine and Redefine the Enterprise of the Future with Generative AI Technologies

3AI May 16, 2023

Featured Article:

Author: Jayachandran Ramachandran, Senior Vice President – Artificial Intelligence Labs Course5 Intelligence

There are multiple inflection points in the history of the computation industry since the invention of computers in 1945. Some of the seminal moments are the advent of mainframe technology, personal computing, graphical user interface, Internet, mobile tech, cloud tech, artificial intelligence, and machine learning—and now we have Generative AI. Each of these moments has been transformative in nature and redefined the way we do work and run businesses. Once again, Generative AI is expected to create multiple opportunities for us. While we witnessed a significant acceleration in digital adoption during the Covid pandemic, Generative AI has the potential to further accentuate it multifold. 

Here’s why. 

Applicability to all business functions across the enterprise 

Generative AI has scope for application across the value chain for all business functions in an organization. It’s capabilities for automated content generation, search and summarization, conversational bots, translation, software code generation, visual input processing, reasoning, creativity, etc. have usefulness across Sales, Marketing, Operations, IT, Human Resources, Legal, and Finance. Organizations can reimagine all their business processes with enhanced AI capabilities and bring in more automation for improved operations and better customer experience. Do more with less! 

Natural language is the new democratic programming language 

Generative AI gives people without technology expertise the power to explore Data, Analytics, Insights, and AI. The barrier to entry is substantially reduced by having a natural language interface, enabling users to easily consume APIs and associated services. For example, anyone can write a software program by interacting with co-pilots in natural language or with pseudo code. This is a game changer as traversing the learning curve becomes faster than ever before. Software development is getting into a new orbit with enhanced velocity and acceleration. Since co-pilots are interactive, transparent, and offer a high level of control, developers would be more open to embracing co-pilots compared to current low-code/no-code platforms. 

Human roles and responsibilities will change 

Generative AI capabilities are driving enterprises to relook at people’s existing roles and responsibilities. While a good part of content creation can be automated, there is a strong need to provide oversight through human reviews to monitor and manage made-up/hallucinated content from Large Language Models (LLMs). For example, more reviewers than creators will be needed. New roles such as Prompt Engineer and Responsible AI Specialist/AI Ethicist will emerge.  

Going back to an interactive and conversational way of learning 

Unlike current ways of learning, during the Vedic era, Gurukul system of learning was prevalent. Learning through student teacher interaction in a conversational mode was a key characteristic of that education system. The ability of LLMs to mimic the immense and widespread knowledge of the teacher can potentially provide highly personalized learning to learners, commensurate with their abilities and in an interactive/conversational mode. Back to Gurukul days! 

Different ways of adopting and implementing Generative AI like LLMs  

Here are some key approaches for incorporating Generative AI technology and its capabilities in enterprises: 

• Out-of-box services (e.g.: ChatGPT) – Here, the capability will be limited to the current knowledge on which the LLM is trained. It is unlikely to have specific knowledge on the proprietary data of the organization. 
• Prompt engineering – This approach leverages Context Input and Generative AI capability. Users can use various prompts to enable zero-shot, or few-shot model learning, and get contextual and relevant results. The quality of the results would be in line with the level of prompt engineering that users can perform. 
• Fine-tuned custom models – This involves high domain adaptation by training an LLM on task-specific data to get the best results.  
• Hybrid Models – This involves combining the capabilities of traditional data science/ machine learning techniques and rendering the last-mile narrative insights through Generative AI capabilities. 

While applying any of the above approaches, it is important to ensure that there is no data leak so that proprietary data is protected.  

Establishing a Generative AI CoE is a promising idea to explore and exploit its full potential 

Over the years, many organizations have set up Analytics, Data Engineering and AI Centers of Excellence (CoEs) that drive enterprise-wide initiatives. Adoption of Generative AI too would require acquiring new skills, changing current roles and responsibilities, creating new capabilities, enhancing existing capabilities, making changes to business processes, etc. It is important to take into consideration the people aspect, especially the workplace culture. There is a lot of initial excitement and urgency to experiment but soon it is followed by nervousness and anxiety about the future which could lead to pushback from people in impacted roles. Since the change is far-reaching and transformational in nature, it would be wise to incubate a Generative AI/ LLM COE to bring focus and attention to the technology, so that end-user adoption is without friction. 

Risk Management and Governance are key 

Generative AI will induce new types of risks. A robust Risk Management framework and Governance mechanism is required. There is a dire need for embedding Responsible AI practices to ensure systems are valid and reliable, controllable, safe, fair, secure, resilient, accountable, transparent, explainable, interpretable, privacy-enhanced and trustworthy. There is also a need to strengthen the review mechanisms to identify potential risks and manage them throughout the AI system lifecycle.  

Experiment, validate, track success metrics, and deliver value 

Generative AI offers opportunities for reducing cost, improving productivity, and identifying new streams of revenue. Over the years, organizations have built many AI models and they are in business-as-usual mode. While Generative AI would provide enhanced capabilities, there are questions to be answered such as: What happens to the current investments? How to justify sunsetting existing models? Are they going to be sunk investments while embarking on a new journey with Generative AI? Would there be management buy-in? A clear strategy must be defined to identify use cases where generative AI has applicability with a clear path to create additional value commensurate with new investments. Baselining current KPIs/success metrics, monitoring and measuring progress will be important to track value realization and success. 

The experiments so far have revealed many risks in the form of privacy issues, security vulnerability, hallucinated content, etc. The ability of the models to produce creative content also plateaus after a few iterations since LLMs have limited real-world context. The technology’s cost will also become a significant factor in adoption. While the technology matures with subsequent versions of LLMs, these issues would get addressed over time. The best way to adopt Generative AI with all its current imperfections is through a “human + machine” model, thereby optimizing outcomes through mutual augmentation of each other’s capabilities. With the power of Generative AI, we are at an inflection point to reimagine and redefine the enterprise of the future. 

Title picture: freepik.com