The bigger the better? Approaching Generative AI by size

As the Generative AI (GenAI) hype continues, we’re seeing an uptick of real-world, enterprise-grade solutions in industries from healthcare and finance, to retail and media. As the technology matures, we’re also learning more about its potential, shortcomings, and barriers to entry—ethical considerations, accuracy, hallucinations, and more. But beyond industry, however, there are factors that play into the success or failure of Generative AI projects. 

One of those lesser talked about factors is company size. A recently conducted survey by Gradient Flow explores the state of GenAI in healthcare, an industry that’s been on the pulse of the technology since inception. Among other findings, the results show clear discrepancies in the way companies of differing sizes approach adoption, implementation, and priorities of AI. 

From budget allocations to model preferences and testing methodologies, the survey unearths the areas that matter most to large, medium, and small companies, respectively. Understanding these nuances can lead to more targeted solutions, higher adoption rates, and less false hope around the transformative power of GenAI. 

Large companies

Large organizations, defined as those with over 5,000 employees, had substantial resources and a strategic focus on evaluation and implementation of GenAI projects. These companies had the highest percentage of respondents actively evaluating GenAI use cases (26%), reflecting their capacity to invest in extensive research and development initiatives.

They also have the means to back it up. GenAI budget increases were significant, with 12% of respondents reporting an increase of more than 300% compared to the previous year. This financial commitment indicates a strong belief in the potential of GenAI to enhance various facets of healthcare delivery and administration. The emphasis on substantial budget hikes also suggests that large companies are not just experimenting with GenAI but are seriously investing in its integration and scaling.

With a clear preference (54%) for healthcare-specific, task-oriented language models, as opposed to general purpose large language models (LLMs), the desire for precise, reliable outputs in critical healthcare applications is clear. Despite more resources and use cases, the preference is to outsource expertise. The high adoption rate of proprietary LLMs through SaaS APIs (cloud-based) in these organizations indicates a preference to rely on third party vendors to drive the AI strategy and implementation. 

Not surprisingly, fairness and private data leakage were top priorities cited when it comes to testing and evaluation of GenAI models, likely due to the high-compliance environment of healthcare and potential reputational damage. The complexity and scale of operations in large organizations necessitate robust testing frameworks to mitigate these risks and remain compliant with industry regulations. 

Medium companies

Medium-sized companies—501 to 5,000 employees—were characterized by agility and a strong focus on GenAI experimentation. These organizations have the highest percentage (24%) of respondents engaged in experimenting and developing AI models, showcasing their ability to adapt quickly and innovate without the bureaucratic constraints that affect their larger counterparts.

Similar to large companies, GenAI budgets are healthy and on the rise just a size down, with 36% of respondents reporting a budget increase of 50-100%. This significant investment points to a growing recognition of GenAI’s potential to drive operational efficiency and innovation in healthcare.

Healthcare-specific task-oriented models were also highly favored, with more than half (57%) of respondents using these models. However, testing priorities differed slightly, with a stronger emphasis on explainability and hallucinations/disinformation. This suggests that medium-sized organizations are keenly aware of the importance of maintaining trust and accuracy in AI-driven healthcare solutions, especially as they scale their AI initiatives.

Small companies

Small companies, with less than 500 employees, represented nearly a third of the survey respondents and exhibited a pragmatic approach to GenAI adoption. Resource constraints and the need for immediate, tangible benefits are likely to have shaped their more cautious, results-focused approach. 

As such, small companies had the highest percentage (39%) of respondents not actively considering GenAI as a business solution. This hesitancy can be attributed to budget limitations and the high costs associated with deploying advanced AI solutions. However, of the small companies that are using GenAI, budget increases of 10-50% were reported, showing momentum. 

Patient-facing applications are a priority for small companies, with high adoption rates for use cases like answering patient questions (38%) and medical chatbots (36%). These applications offer immediate, visible benefits, enhancing patient engagement and communication without requiring extensive resources or infrastructure.

Small companies place a higher emphasis on supervised fine-tuning (39%) and de-biasing tools and techniques, showcasing a commitment to ensuring model accuracy and fairness. Testing priorities also include bias and freshness, indicating a focus on maintaining relevant and unbiased AI outputs that can adapt to changing healthcare needs.

Universal challenges and benefits 

While company size brings different GenAI challenges, some are universally acknowledged. Accuracy, security, and privacy risks are top concerns across the board, reflecting the high stakes associated with AI in healthcare. The potential for legal and reputational risks is being felt by all organizations, highlighting the need for robust mitigation strategies.

It’s not all bad news, though. The survey showed a shared belief in the transformative potential of GenAI, particularly in patient-facing applications like transcribing doctor-patient conversations and providing medical chatbots. While the pathways to adoption differ markedly, healthcare organizations of all sizes are meeting GenAI where they are. The 2024 Generative AI in Healthcare Survey brings to light the strategies and priorities influencing healthcare. While large companies drive extensive evaluation and investment, small- and medium-sized companies bring agility and pragmatism to the forefront. Each approach reflects the unique challenges and opportunities inherent to different scales of operation. Understanding this will not only help create better, more tailored AI solutions, but make the technology more accessible to all.