Business leaders evaluating artificial intelligence (AI) solutions should note that large-scale generalist models, such as Openai's ChatGPT, always deliver better results than specialized models. analysis It was published this week by Harvard Business Review.
Large-scale language models (LLM), such as ChatGpt; Claude and Gemini A typical task requires tasks that require you to retrieve information from a variety of sources, such as customer service, not as a customized AI model for a particular industry for domain-specific tasks.
tHe may seem obvious, but he actually pans out. That's the experience of the report author Amberni gumfounding CEO of startup Basics.aiand John GlaserExecutive in Residence Harvard Medical School Former CEO of Siemens Health Services.
The authors have built a generator AI system for health insurance companies to determine whether doctors' recommended treatments are covered by patient health insurance. They discovered that choosing the right GEN AI model is important for achieving optimal business outcomes.
The key differences to understand between the Generalist and Specialist models are: Specialized AI models know not only what data to retrieve, but also “how that information works within a decision framework for a particular domain.”
For example, a specialized AI model should decide whether health insurance companies should approve coverage for treatment plans for patients with lung cancer.
The author also wrote that influencing treatment or decisions regarding insurance coverage should take into account patient conditions such as end-stage renal disease and recent hospice referrals.
The generalist AI model attacking the same problem looks for historical patterns of insurers approving coverage for other patients with similar symptoms. However, “This pattern-matching approach misses the fundamental clinical and policy logics that drive these decisions, especially in more complicated cases like the ones mentioned above,” they said.
Please refer: From buzzwords to final lines: Understanding AI model types
Groundbreaking ideas
For the author, the “aha” moment is “Why do you want to make AI think like a computer when you need to think like a doctor?”
The author then trains GEN AI agents to “follow the clinician's reading – Understand the structure of the chart, move from section to subsection, identify appropriate findings of the context.”
According to the author, this understanding will help other companies choose the right AI model to deploy for their use cases, regardless of the industry.
According to the PYMNTS Intelligence Report, business leaders are coordinating Gen AI Systems to meet their strategic orders. Those working in the product technology industry are much more likely to use Gen AI to design and generate ideas for products. However, service companies use GEN AI for more specific purposes, such as creating better strategic positioning and generating insights faster.
Furthermore, while most business executives highly valued AI generals in terms of their effectiveness in completing specific jobs, they acknowledged that “broad” human surveillance was still needed. This means that “enthusiasm may reflect how quickly the adopters have started to find useful Gen AI in an accurate way, rather than being prepared to actually be present.” May 2025 Report.
The disconnect between expectations and reality underscores the importance of choosing the right AI model for your job. The author said companies should ask AI vendors the following questions to avoid common pitfalls:
- Can the system transparently explain that inference? Generalist AI models often provide shallow or opaque answers. The specialized system must show how criteria can be used in decision making and cite criteria or precedents.
- Are vendors working closely with domain experts to adapt to changing norms? Rich collaboration embeds the model with an evolving professional framework rather than an outdated set of rules.
- Will the solution grow across domains? Avoid siloed point solutions by choosing an architecture that can integrate professional logic and expand into new fields such as finance, legal, engineering and more.
read more:
AI models are mature with customization as the next focus
Report: Inferential that AI models fail if problems become too complicated
Microsoft plans to rank AI models safely
