Gen AI: Headwinds into Tailwinds

If it’s certainty you’re looking for, you’ve come to the wrong economy. 

The recession that’s been “looming” for the last several quarters still hasn’t arrived. But it hasn’t stopped looming either, and businesses remain in a defensive crouch—scrutinizing budgets, keeping discretionary spending tight, and focusing on investments that promise a short-term return in value. 

There’s just one challenge: the tectonic shift in the business environment created by Generative AI (gen AI). 

There will be clear winners and losers in this new era defined by gen AI. Early adoption is key. It is time for companies to dispense with the “blue sky” thinking and resulting paralysis by analysis. If you’re going to win, you must act. 

Make three types of practical, measured, enabling investments that will position you to capitalize on gen AI with maximum optionality: data, training, and large language models (LLMs). 

Investing in AI without addressing data quality challenges is akin to constructing a house on an unstable foundation. AI systems heavily rely on high-quality data to generate accurate insights and help you make informed decisions. Neglecting data quality compromises the very essence of AI's potential. It undermines the trust in AI systems, erodes confidence among stakeholders, and jeopardizes desired outcomes. 

Next: skilling and training. A defining marvel of gen AI is that it can take instructions and deliver outputs in natural language, but that doesn’t eliminate the need for worker training. Far from it. To maximize gains and minimize risk, companies need workers grounded in basic AI principles—primarily model development and prompt engineering. At Cognizant we recently launched our “Synapse” initiative, which seeks to train one million global workers in cutting-edge technology skills, including these foundational principles by 2026. 

The final core investment: the LLMs that power gen AI applications. Here, the key is getting to value, fast, creating competitive advantage. There are generally three approaches to building LLMs: 

1. License an existing model and fine-tune it. Work with a global systems integrator (GSI), like Cognizant, to layer in the GSI’s sub-industry specific data and tailored models. Then, fine-tune it with your company’s proprietary data to securely create unique, competitive advantage that will set you apart from the competition. Cognizant is doing this right now in partnership with Google Cloud, modifying Google’s PaLM LLM to serve a host of clients in the payer sector of the healthcare industry. We’re harnessing all the power of PaLM’s original training data, enhanced with industry data from Cognizant, and proprietary business data from each client, to build applications that are both powerful and uniquely tailored to transforming administrative processes in areas such as appeals and grievances, and member and patient engagement for healthcare clients.

2. Build your own LLM from scratch. While this is potentially the most powerful and flexible option, it also requires the most time, money, skills, and data. Unless you’re a major tech company, along the lines of a Microsoft or Google, this approach will tie-up resources that could be better used in your core business delivering value to customers. 

3. Use an open, public model like GPT 4. While this is the fastest way to stand-up an LLM, it fails to leverage proprietary data to provide organizations a competitive advantage. It also requires data privacy and security due diligence. 

Gen AI’s arrival, in short, makes this a fitting moment to stop bracing for economic turbulence, and start turning uncertainty into opportunity. Commit to these core investments now, and you may one day look back on this uncertain macroeconomic moment as an inflection point, the start of huge, even radical growth for your business.