A Framework for identifying Generative AI Use Cases

A Framework for identifying Generative AI Use Cases

Generative AI has captivated the attention of consumers all around the globe. The simple reason for this is that it has significantly enhanced the experience of human-machine interaction. It has elevated it to be interesting and engaging. This experience will now become the standard that consumers expect from enterprises during their interactions with any of their enterprise touchpoints. While this has triggered a wave of interest in enterprises there is always the challenge of where to start. This is a question that comes up in multiple client conversations.

I will present a framework to streamline the process of use case identification. Before we delve into that it is important to understand the strengths and limitations of Gen AI. Today foundation models trained on extremely large datasets are available for text, audio, video and images. The key strength of the models are the quality of content generated. These models are able to recognize the context and intent of the query and respond appropriately. This means today machines can engage in interesting conversations. This is also the weakness of the models since they also confidently respond to queries where they have no clue. The reason for this is due to the fact that Gen AI models are probabilistic in nature and not deterministic. The key take away is that these models should not be relied upon for use cases where consistent responses are the norm (example – Maths, logic etc).??The other aspect to keep in mind is that these models can provide different answers to the same query due to randomness introduced. This is also more likely to happen for same queries in different languages as the underlying datasets used for learning are different.

Having established some guardrails let us now consider the framework for use case identification. The two key determinants for use case identification in my view are the value generated through enhanced customer experience vs the ease of implementation. Let us consider each of them in detail.

Customer Experience

The impact to customer experience can be both direct as well as indirect. Let us consider the broad areas which have great potential for curating use cases for Generative AI

a.??????Knowledge Base

b.?????Content Creation

c.??????Productivity

d.?????Process Optimization

e.?????New ways of Working

The democratization of knowledge was kicked off by the invention of the printing press. Since then the internet & mobile revolutions have made it accessible on demand. Generative AI will further accelerate this trend and significantly deepen the personalization dimension.??Both knowledge and content can be served in a very personalized manner through Gen AI. Similarly productivity and optimization opportunities exist across the value chain of every enterprise. From customer acquisition, customer interactions, operations and support functions. Developer productivity is going to be accelerated with code assistants working side by side with developers. As with every new technology revolution new ways of working are bound to evolve. The intersection of Robotics & Gen AI is one interesting space to consider. It is also a pattern that with each tech advancement the barriers between customer and enterprises keep getting dismantled. This will open up opportunities for new business models to evolve.

Ease of Implementation

This is influenced by the following factors

a.??????Fitment

b.?????Skills

c.??????Costs

d.?????Risks

We discussed fitment to some extent when establishing guardrails for selection of use cases. Fitment has to account for both use case selection as well as the model selection. Skills are primarily around the ability to architect & design a scalable, future proof and cost efficient solution with MLOps in place for Day 2 operations.??Risk assessment for Generative AI needs to consider aspects of explainability, bias, inconsistency and copyrights of datasets used for content creation. It is imperative that organizations have a clearly defined risk assessment mat

These are exciting times and really looking forward to bringing to life some of these use cases in the coming months.