Adopting “AI” for Digital Transformation requires specificity

The term "AI" has become a ubiquitous, catchall term. The overuse and imprecise usage of the term AI – particularly in the context of digital transformation – threatens to turn it into yet another overrated and empty buzzword, as it lacks the specificity and precision to convey the different forms it can take and their potential positive impact on businesses.  
 
Some much-needed clarity is in order. What exactly are the different types of AI, and what are the distinct value propositions of each one – and more importantly, how can enterprises strategically, precisely, and purposefully adopt these technologies to meet their digital transformation goals?

A closer look at LLMs 

It's helpful if we break AI down into three general buckets: large language models (LLMs), generative AI, and predictive AI.

LLMs – which, as the name suggests, are trained on very large amounts of data – are very well suited for creating productivity gains within the organisation. From a digital transformation perspective, this might mean using LLMs to power chatbots for customer service or providing a new way to answer frequently asked questions when onboarding new employees.

While these are very specific use cases, there can also be more ad hoc usage of LLMs to achieve productivity gains – for example, harnessing them to distill large amounts of information into bulleted or summarised lists, or otherwise providing a quick "jump start" that helps take some of the grunt work out of common business activities.

Generative AI for content creation

While there's some overlap with LLMs (basically, LLMs are "the engine" of generative AI), it's best to think of generative AI as a tool for content creation.

Many professionals have already dipped their toes in the generative AI waters, so it will come as no surprise that good use cases for generative AI typically revolve around generating content that then can be edited and revised by a human to produce a polished final product. Whether it's creating marketing materials or white papers or company memos, generative AI provides a useful starting point for content, helping accelerate its development. 

Making smarter decisions with predictive AI

Predictive AI is a slightly different animal. When it comes to digital transformation, the main use case here is analysing large amounts of structured data in order to make better, more informed business decisions.

A specific example of this might involve trying to predict customer behaviour so that the enterprise can take appropriate action in response to what the data is predicting. For instance, running predictive AI against structured customer data can help identify at risk customers so that account teams and customer success teams know where to focus their efforts to reduce customer churn. Alternately, it can help identify optimal cross-sell or upsell opportunities in the customer base, letting sales teams home in on customers that are most likely to yield the best results. 

Real impact requires a thoughtful approach

Regardless of which "flavor" of AI an organisation opts to deploy, there are certain best practices that hold true across the board.

For starters, it is absolutely essential to clearly define the use case for that piece of AI and the expected outcome. Do you want to reduce customer churn by 15% using predictive AI? Reduce the amount of time required to produce an annual report by 20% with generative AI? Increase customer response time by 30% using an LLM-powered chatbot? Measuring actual outcomes against expected outcomes will ensure there's alignment between the AI being utilised and the business need being addressed.

Additionally, enterprises need to turn a careful eye toward the data that their AI is drawing upon. AI delivers best results when it has carefully curated, vetted, "gold standard" data to harness. For LLMs or generative AI, that data can be unstructured; predictive AI, by contrast, will require very structured data. In either case, what's most important is that the data is high quality and trustworthy. Having access to this pool of curated data automatically builds in guardrails around the AI, ensuring it doesn't use faulty information to produce its outputs.

You can't solve your business problems with "AI"

Specificity matters. You don't eat a meal with "silverware" – you eat a meal with a fork, or a knife, or a spoon. Likewise, you don't tackle a DIY home improvement project with "a tool" – you use a hammer, or a saw, or a screwdriver.

AI deserves the same level of specificity. Enterprises hoping to effect digital transformation within their organisation by adopting "AI" need to get more granular and focus not just on the specific business changes they're hoping to achieve but which specific form of AI is best suited to the task at hand.

By getting specific, enterprises can give themselves the best shot at successful digital transformation – while preventing "AI" from being more than just a vague and meaningless buzzword.