On this page
Back to top
On this pageinpage
Model choice
Should you buy or build your generative artificial intelligence (GenAI) models?
That’s one of the most pressing questions business leaders addressed at the World Economic Forum’s Annual Meeting in Davos, as I found out in a debate on GenAI this week.
It’s front of the mind for industry leaders as the technology is advancing rapidly and requires substantial investment—particularly when you’re looking to build tools like Large Language Models (LLMs). When approaching this decision, many feel like they are entering a candy store, where the technological possibilities are at once enticing and the choice overwhelming.
Here are some key focus areas to inform your discussion:
Existing solutions
Making the most of existing solutions and utilising them effectively.
Our current strategy is to consume existing solutions, especially LLMs, where there are several options and the cost of building from scratch can seem prohibitive, particularly for smaller businesses. Buying readymade AI models allows in-house technical experts to focus on deploying them rapidly to add value to the enterprise.
We anticipate that over time we will see a combination of small and large models emerge. But for now, the emphasis will continue to be on exploiting available resources and incorporating them effectively.
Cloud plus AI
Combining the power of cloud platforms and emerging AI technologies.
We see a significant opportunity to utilise the power of cloud platforms as a secure way into AI. Cloud technologies can provide a means to explore the right guardrails, uphold the highest security levels, and ensure compliance with applicable rules and regulations. Organisations can then use their experience in cloud development to get ahead in this area.
Initial findings of the TCS AI for Business Study revealed that 46% of executives plan to create their own LLMs. It’s a complex challenge for any organisation and requires enormous network resources to optimise the models for cost, quality, security, and privacy. Working with a cloud provider can help streamline this process by ensuring IP security and freeing your network
Decision-making
Data and AI can transform traditional decision-making processes.
GenAI advancements are supercharging the way we handle unstructured data in every aspect of our lives. AI and alternative data will revolutionise traditional decision-making processes.
The panel I attended discussed the many practical applications of GenAI. One example is in economics, where, traditionally, governments release data based on surveys of manufacturers. But now AI can interpret satellite data and maps showing the state of supply chains in real time.
We can also introduce other data sets that AI has assessed, like what’s being purchased, for example, ultimately providing the shopping mall, or whichever business, with a clearer picture of economic activity.
There are obvious business applications to this strand of GenAI, as well as insights for policymaking and accelerating progress towards the sustainable development goals. Essentially, the more data we have, the faster and more effective our decision-making will be.
AI-first approach
Developing an AI-first architecture to be in lockstep with the GenAI revolution.
For business leaders on the precipice of the GenAI revolution, and trying to decide whether to build or buy, it’s important to understand that the journey will likely have many twists and turns. We at TCS believe in an AI-first architecture, where GenAI and its models are layered on top of existing enterprise IT systems and ultimately augment them.
That’s not to say the task is easy, but for those who keep an open mind, work with the right partners and are prepared to overcome the challenges, the rewards will be transformational.
