We’ve hit a tipping point with artificial intelligence (AI) where boardroom discussions have shifted from debating efficacy to accelerating adoption. It’s an exciting time, especially considering the pace of change will never be this slow again. According to BCG, despite...
We’ve hit a tipping point with artificial intelligence (AI) where boardroom discussions have shifted from debating efficacy to accelerating adoption. It’s an exciting time, especially considering the pace of change will never be this slow again. According to BCG, despite global economic uncertainty, innovation rose as a top corporate priority in 2023, with 79% of companies ranking it among their top three goals.
But innovation for innovation’s sake is not a sound business strategy, and organizations that get caught up in the AI hoopla risk investing in hype, instead of solutions that create long-term value. Understanding the difference requires careful consideration of current capabilities and the patience to prioritize sustainable growth over short-term trends.
The Goldilocks Zone
Business history is littered with examples of companies whose strategic decisions at key moments have been consequential to their existence. For example, Amazon survived the dot-com bust by recognizing the importance of adjusting its accounting strategy and boosting reserves while other companies were burning through cash like there was no tomorrow. The point is, sound business decisions are more critical than ever during times of mass enthusiasm, and planning for tomorrow requires a keen ability to think through all potential scenarios.
Overall, there’s a general feeling of AI FOMO (“fear of missing out”) that has permeated leadership teams, further complicated by the reality that doing nothing (i.e. succumbing to “paralysis by analysis”) is also a real threat. (Just ask Kodak.) Here are 3 considerations for companies looking for that “Goldilocks Zone” of AI—not investing too fast or too slow, but finding the sweet spot of sustainable innovation.
1. Focus on Data Growth First
Like any machine, it’s important to understand its inner workings to derive where the value comes from. Meaning, AI isn’t a fully-formed product, rather its large language models (LLMs) rely on vast amounts of diverse data points to learn patterns, context, and linguistic nuances. The sheer size and complexity of LLMs require extensive training data to operate effectively across various domains and tasks. The quality and quantity of this data will greatly impact the performance of LLMs, and by extension, a company’s suite of AI tools.
Creating more robust data ecosystems is therefore a wise first investment for any company planning an AI transformation, and this data will serve as the foundation for LLMs as they grow and evolve. It’s in this evolution where high-quality data becomes even more critical. While studies have found that LLMs can be competent with minimal data, experts now say that “the impact of data quality and diversity on both alignment and other avenues of LLM training (pre-training, fine-tuning, steerability, etc.) is absolutely massive.”
2. Identify a Business Use Case
While AI certainly has the capacity for broad external applications, most companies are more focused on using the technology to optimize their internal processes. “Optimize” is the key word here, meaning that companies shouldn’t expect to just plug-and-play AI software to magically improve output. Rather, some of the most successful AI use cases involve analyzing data to reveal valuable insights into customer behavior, market trends, and potential risks. It’s also been proven effective at streamlining internal activities, including things like automating manual tasks to allocate employees’ time to higher-level activities.
In short, instead of wasting time figuring out which AI models to use, organizations need to focus on specific problems they need their AI to solve. (i.e. start with the needle you want to move, set up the KPI that you’d like to influence, and then work backward toward what AI tools will accomplish those objectives.) According to MIT’s Global Executive AI Survey, 90% of those who use AI to create new KPIs say they see their KPIs improve. “These AI-informed KPIs offer business benefits and demonstrate new capabilities: they often lead to more efficiency and greater financial benefit and are more detailed, time-sensitive, and aligned with organizational objectives.”
3. Build Bespoke AI Tools Using Open Source LLMs
To build, or to buy – that is the question. Building a customized AI solution can seem daunting, and many companies opt to purchase a license from an outside vendor with a proprietary LLM to avoid going down that path. However the license may restrict how the LLM can be used, and licensing fees can get very expensive over time. Alternatively, open-source LLMs are free and the underlying architecture is available for developers to access, build, and modify based on the specific company needs.
This open-source model ecosystem has gained in popularity as companies endeavor to keep sensitive information on their network and retain more control over their data. Open-source LLMs give companies this transparency and flexibility, along with the added benefits of reduced latency issues and increased performance. IBM and NASA recently teamed up to develop an open-source LLM trained on geospatial data to help scientists fight climate change, part of NASA’s decade-long Open-Source Science Initiative to build a more accessible, inclusive, and collaborative scientific community.
As with any open-source technology, there are risks associated with open-source LLMs, including potential security leaks/breaches, hallucinations/bias based on inaccurate or flawed information, and bad actors intentionally manipulating data. But open-source models are getting smarter and more secure over time, leading some experts to feel that open-source LLMs will soon reach the level of the best closed-source LLMs, justifying the investment in early adoption and time spent upskilling teams.
AI Adoption Will Be Multiple Quick Sprints in a Marathon
Based on recent figures, there are around 15,000 AI companies in the United States, more than double the amount in 2017. Worldwide, those numbers increase nearly fourfold. With this many vendors and hot new startups promoting their services, it’s no wonder that companies can struggle to decide where to invest their time and money. But by carefully assessing your needs and the risks/rewards presented by innovation, leaders will find the right mix of AI to propel their companies into a future of sustainable growth.
The post Solving the How & When: Infusing Business Strategy into AI Adoption appeared first on Unite.AI.