Startup Snapshot

Go all in or warm up on the sidelines? How business leaders should decide on LLM investment

When deciding how much to invest in LLMs today, business leaders must take into account both sides of the coin. This article highlights the main points to take into consideration.

Startup Snapshot - Ilya Venger

Ilya Venger

How business leaders should decide on LLM investment - Startup Snapshot

In April last year, I had a conversation with a head of strategy for a mid-sized European bank. He asked me a direct question: “How much should we be investing in LLM and are we missing out if we don’t move quickly?”

In a somewhat controversial stance, I advised against succumbing to the pervasive fear of missing out (FOMO) and suggested a more measured entry into the realm of Large Language Models, particularly for mid-cap enterprises not at the forefront of technology.

This article delves into the complexities of that advice, exploring the evolving landscape of technological capabilities, the dynamics of consumer expectations, and the critical need to comprehend the true potential and limitations of AI. The considerations surrounding AI adoption are highly nuanced, and a number of risks and rewards should be considered before business leaders take their first steps into GenAI.


Hold your horses and take it slow

My initial stance is that mid-cap organizations should warm up on the sidelines before building products in earnest. The rationale lies in a number of points, each of which must be considered:

  • We are on an exponential capability curve of technology. The cost of development and running solutions is going down. This is due not only to better tools, but also to cost optimization of the models. Someone who starts building later, but leverages best practices and lower costs could get to product-market-fit earlier.
  • Many business stakeholders have ‘magical thinking’ about AI, while not truly understanding its true capabilities. A great example of this is stock picking by ChatGPT. This cannot be achieved through a chatbot alone, but people still can’t let go of this use-case. Many still don’t fully understand AI limitations and potential.
  • Consumers are also still in a learning phase. Their expectations are going to change andbuilders should be skating towards where the puck is going to be, but no-one knows where it is going to be. This adds further significant risk.
  • I vividly remember some business cases for digital apps with investments of tens of millions of dollars a decade ago. In a high-interest environment, the opportunity cost or risk-free alternative is high and needs to be taken into account. Given the risks above, the risk-adjusted-returns require a truly exceptional business case.


Don’t miss the innovation train

Despite the many reasons for taking it slow and not succumbing to FOMO, it is clear that doing nothing is not an option. Tales of Nokia and Blockbuster are being told in every MBA classroom. Shareholders and the market expect some action, especially when it comes to buzz words and innovative new technologies.

Here are some initial thoughts on what a strategically minded business leader must do to not miss the train.

  1. Ensure people both on both the business and the technology side understand AI capabilities and limitations. Learn through experimentation – organize hackathons, innovation workshops and share knowledge. Add emerging external AI tools into your processes. This is crucial for hands-on experience.
  2. Be strategic. Evaluate current processes and build a prioritization plan. Understand what unique advantage AI would bring to your business and what capabilities are missing in current solutions. Spend the time to define the portfolio of future change. Scout the market to not miss the moment when these capabilities appear.
  3. There’s nothing wrong with ensuring your experiments are externalized for PR and marketing value. This is going to both energize your people internally as well as manage market perceptions.
  4. Invest in foundations: particularly in data. Organizing, cleaning and documenting your data is going to be critical. Models without data are nothing.


So, it’s not just FOMO. There are real risks of inaction, and potential for outsized rewards. Strong strategic thinking, product management and execution – all still apply and large businesses must think about their investments strategically, but they have a higher chance to succeed and more to lose tha mid-sized companies.


About the Author:

Ilya Venger is a Principal Product Lead for Industry AI at Microsoft, where he focuses on defining the foundations that enable customers and partners to create GenAI-driven Copilots. Before joining Microsoft, Ilya led data architecture for the Group CTO at UBS and spent nearly a decade as a strategy consultant, advising Fortune 100 executives on digital transformation and proposition development. He holds a PhD in Systems Biology from the Weizmann Institute of Science.

Browse all our reports

Check out our data deep-dive

Rothschild 45
Tel Aviv, Israel