Not all AI is equal: why specialized models are the real ROI driver in 2025
By Ed Crook, VP Strategy & Operations, DeepL
When it comes to AI adoption, Forrester summed things up pretty well in a recent report: If 2023 was the year of FOMO and 2024 was the year of pilots and early adoption, then 2025 will be the year of calculating the financial ROI of AI workforce initiatives.
Following the hype cycles of recent years, conversations about what’s possible with AI are now giving way to discussions about how organizations can actually deploy and scale the technology to deliver ROI-positive value. After a steep learning curve and a period of testing the technology on their own, decision makers and their teams are ready to go. So much so that, according to new research from DeepL, almost three-quarters (72%) of decision makers intend to invest in AI this year. The Netherlands has the highest commitment to AI spending, where 30% of businesses plan to integrate AI into their operations, followed closely by Germany (29%), Belgium (28%), France (26%) and the United States (25%). Those companies that do invest are likely to reap significant benefits.
But here’s the thing. While I agree that 2025 is go-time for widescale AI deployment, there’s something organizations should keep top of mind: the key to unlocking competitive advantage will come from deploying specialized AI solutions rather than the general-purpose AI they may be more accustomed to seeing and reading about. Across all regions, DeepL research found the top AI budget allocations this year are for business operations (28%) and AI for specialized tasks (25%).
Specialized AI delivers ROI for enterprises across industries
General-purpose AI, which takes a one-size-fits-all approach to a wide range of needs across industries and use cases, does have merit. It can create recipes, answer questions and generate just about any text or image a person or business can imagine. It’s also handy for translating the occasional menu while on vacation. But general models also have limitations, especially for business use cases, mostly because of the way the technology is trained. These models are typically trained on a broad and diverse dataset – the publicly available internet – and therefore tend to lack accuracy, quality and customization for specific tasks, all of which are critical for enterprise AI applications.
In contrast, specialized AI models are tailored to specific industries and use cases, trained on domain-specific data and optimized for specific tasks or industries. Because they are tailored to specific needs, they often prove a better fit for enterprise use cases. While general-purpose AI models offer broad capabilities, in the enterprise, where the stakes are higher and the requirements more specific, specialized models offer tailored solutions that can address complex, industry-specific challenges with tangible ROI. Benefits of deploying AI technology powered by specialized models include higher accuracy and efficiency; regulatory compliance, as these models are typically built to incorporate industry-specific regulations and compliance requirements; as well as an enhanced user experience with customization opportunities. In other words, if you’re a law firm looking to leverage AI to read and evaluate critical documents; a healthcare provider assessing medical scans; or a retailer relying on AI to interact directly with customers for customer service needs, it’s specialized AI you’d trust to do it right.
Also Read: Beyond Maps: Driving Next-Gen Navigation with AI and Voice Assistance
Language AI takes communication from challenge to opportunity
One notable area where specialized AI is making an impact is in language, translation and communication. It’s no coincidence. When talking with CEOs and leaders in large global organizations, a huge, recurring challenge they face is how to effectively manage thousands of customers, teams and suppliers around the world, all speaking different languages. While English is, for many, the global language of business, the reality is that only 20% of the world’s population speaks it fluently. This language gap can hinder everything from your ability to expand abroad (35%), to engage new customers in other countries (32%) and even to properly serve your current customers (24%). On a more practical, day-to-day note, more than half of C-level executives say they spend over an hour a day dealing with ineffective communication. As you can imagine, this challenge seen at the top naturally extends to all levels of the organization.
I’ve built my career at the intersection of Language and AI for this reason. I’ve seen the power of technology to make global businesses more efficient, effective and empathetic. Specialized AI is also helping many organizations to see language as a strategic asset. For example, Panasonic Connect is using the technology to remove language barriers hindering collaboration within their organization, not only speeding up the translation and communication process, but also achieving “breathtaking” results with natural, human sounding Japanese translations. iCrowdnewswire, a leader in international press release distribution, is also leveraging the technology – integrating it directly into their platform via the DeepL API – to translate between 45 and 55 million characters per day on behalf of global customers, turning a highly manual process into a seamless, scalable, reliable process and creating cost efficiencies. Communication, instead of being a challenge to overcome or work around, becomes a competitive advantage – a way to improve understanding, reach global audiences and encourage better internal and external collaboration and ultimately power global expansion.
It’s clear that the rewards of AI deployment will be significant. When we look back on this time in years to come, I believe that 2025 will be a moment we will point to as the tipping point at which for many businesses, AI moved from abstract ideas to real-world results. But while off-the-shelf AI solutions can make valuable and lasting commercial contributions, decision-makers need to be aware of both their potential and their limitations. It’s the “precision over versatility” of specialized AI that promises to deliver seismic gains for businesses. We’re already seeing this in the world of Language AI, and I can’t wait to see where else its impact will soon be felt.
Comments are closed.