In its various iterations, artificial intelligence (AI) has been the subject of much hype across industries due to its potential to solve a multitude of business challenges. But despite the excitement AI has generated, some sectors — and/or individual organizations — haven’t managed to optimize this rapidly evolving technology to the extent originally hoped.
For instance, although global AI in manufacturing is expected to surge from $3.2 billion in 2023 to $64.63 billion by 2031 — a CAGR of 45.6% — SupplyChainBrain’s Robert Bowman notes that only 20% of generative AI (GenAI) projects launched in that sector are viewed as a success. Some of the challenges faced there and elsewhere include cost, lack of interoperability with legacy systems, poorly defined use cases, and the tendency for companies to bite off more than they can chew.
AI hype vs reality
Cited in Bowman’s article, Matthew Hart, founder and chief executive officer of Soter Analytics, says that in addition to the “underlying problem” of legacy applications that don’t play well with evolving AI innovations, GenAI isn’t quite “living up to the hype.”
“GenAI continues to suffer from ‘hallucinations’ and the dispensing of advice that’s just plain wrong,” Bowman writes, adding that the technology has a “long way to go,” before it can effectively sort through the massive amounts of information available to create accurate answers on demand.
Additional issues? Unrealistic expectations and a tendency to “turn this into a huge process” that undermines success, Hart says — underscoring the importance of specificity when it comes to applying AI: “When you break it down to a small, specific problem, you can build very high-quality solutions.”
In this context, specialized AI has been getting a lot of attention lately to help organizations make the most of this technology by zeroing in on specific needs.
What is specialized AI?
According to AI provider UiPath, specialized AI refers to “artificial intelligence systems that are designed and trained to excel at specific tasks or domains. Unlike general-purpose generative AI tools that aim to replicate human intelligence across a wide range of capabilities, specialized AI focuses on mastering a particular skill or solving a specific problem.”
Within the broader AI landscape, UiPath says specialized AI is a “crucial” piece of the larger AI puzzle: “It’s a practical application of AI that’s already making waves in the real world, solving specific problems and transforming industries.”
In this context, the company says specialized AI focuses on solving industry- and company-specific problems and is built to “take full advantage of a company’s private, internal data and business context” to do so — which is why it’s often considered part of the organization’s proprietary intellectual property (IP).
Described as “precision instruments designed for a specific job,” specialized AI models are also typically much less “computationally intensive” than general GenAI tools.
In the following video, Matt Hicks, CEO of software company Red Hat discusses the “importance of small, specialized AI models.”
Embed 2:11 min video
Embed code:
<iframe width=”560″ height=”315″ src=”https://www.youtube.com/embed/98UeougY0zQ?si=yWjyNUihBA6S-vh5″ title=”YouTube video player” frameborder=”0″ allow=”accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share” referrerpolicy=”strict-origin-when-cross-origin” allowfullscreen></iframe>
Specialized AI models
In a recent post for AIMultiple Research, principal analyst Cem Dilmegani describes three business-related categories of specialized AI systems:
- Horizontal AI: Specialized AI systems that are “focused on specific business functions” and “can be applicable across various industries.” Dilmegani says Horizontal AI models include three key characteristics: cross-industry applications, business-unit-specific applications, and general-purpose tools.
- Vertical AI: AI systems that are “designed to solve problems within a specific industry or domain,” are “highly specialized,” and niche focused. He says the three key characteristics of Vertical AI include industry-specific expertise, customized data, and specialized applications.
- Common AI: Refers to “language or action models covering many generic use cases.”
Other specialized AI models Dilmegani describes “cater to more focused, cross-disciplinary needs or unique environments” and include:
- Edge AI — which are “optimized to run on edge devices (like IoT sensors or smartphones) rather than centralized data centers. …”
- Multimodal AI — which “…can process and integrate data from multiple modalities such as text, images, audio, and video, providing more comprehensive and context-rich outputs.”
- Generative AI — which “…can create new content such as text, images, music, and even video by learning from existing data.”
- Explainable AI (XAI) — which are AI systems “designed to provide transparent and understandable decision-making processes. …”
Benefits of using specialized AI
UiPath says specialized AI can be a game-changer for businesses, since it helps:
- Enhance efficiency: “… by focusing on a specific task or domain, specialized AI solutions can outperform their general-purpose AI counterparts in terms of speed and accuracy. …”
- Increase accuracy: “… with supervised or semi-supervised training on relevant data sets, these AI models become incredibly accurate at their designated tasks. …”
- Provide enterprise-specific solutions: “… specialized AI models can be customized to address the unique challenges and opportunities of your specific enterprise and business operations. …”
- Reduce associated costs: “… compared to the costly development of generative AI and large language models, specialized AI solutions can be more budget friendly. …”
In addition to the benefits that AI solutions can offer overall — such as data-driven decision making, automation-related cost savings, an improved customer experience, and the ability to tap into new opportunities — specialized AI can also give businesses a unique competitive edge.
“First and foremost, it makes use of your company’s proprietary data — on customers, suppliers, product and network performance, and much more,” UiPath explains. “This exclusive data can provide the information and context that specialized AI models require to thrive. With better-performing models that no other competitor owns, you can outpace the market in operational efficiency, optimized decision making, and enhanced customer experiences. Plus, you’ll be well-positioned to adapt to market changes, capitalize on emerging trends, and maintain a leading position in your industry.”
Challenges of using specialized AI
Although specialized AI offers a robust list of benefits, UiPath says “it’s crucial to acknowledge and prepare for the challenges that accompany its adoption,” which include:
- Data dependencies: “Specialized AI models rely on large amounts of high-quality, domain-specific training data. …”
- Financial implications: “Developing and deploying specialized AI solutions can be a significant financial investment. …”
- Ethical considerations: “As specialized AI systems become more sophisticated and integrated into critical decision-making processes, ethical concerns come to the forefront. …”
- Integration hurdles: “Integrating specialized AI solutions into existing systems and workflows can be a complex undertaking. …”
For additional details about the benefits and challenges of adopting specialized AI — as well as a roadmap to deploy it — please see the UiPath resource.
Specialized AI and the supply chain
Recently, Supply Chain Dive hosted a webinar during which industry experts discussed “how specialized AI is transforming supply chain management and why traditional approaches no longer suffice in today’s dynamic environment,” according to a Loop summary of the event. The panel featured Matt McKinney (CEO and Co-founder of Loop), Kevin Donnelly (President and COO of Paravel Luggage), and Taylor Lucas (VP of Finance at Passport).
Underscoring the “unprecedented unpredictability” of the current supply chain landscape, the summary describes the pitfalls of traditional supply chain approaches:
- Disconnected data both internally and externally
- Manual systems that limit decision quality
- Outdated processes that can’t quickly adapt to changes
And describes the “three key barriers” of people, process, and systems.
“While people are often open to experimenting with new processes, they’re frequently held back by outdated systems,” the summary says, citing McKinney. “This creates a cycle where antiquated technology limits the advancement potential of both people and processes.”
But Loop says that when it comes to supply chain applications, a “key distinction emerged between general AI and specialized AI,” with McKinney explaining that “while general AI (like ChatGPT) works well for simple tasks, enterprise use cases require specialized AI that understands industry-specific language and contexts.”
In this context, the benefits of specialized AI in the supply chain are described as:
- Higher accuracy for mission-critical workflows
- Domain-specific training preventing costly errors
- Transparency and explainability in decision-making
- Ability to standardize and centralize messy data
“The future of supply chain management lies in specialized AI that can understand and adapt to industry-specific challenges,” Loop says. “Companies looking to stay competitive must evaluate their current systems and processes, understanding that the goal isn’t just to automate existing workflows but to reimagine them entirely through the lens of specialized AI capabilities.”To learn more, please watch the webinar.
