Smaller Language Models (SLMs) are the future
In the world of artificial intelligence, sometimes smaller is better. Smaller language models (SLMs) are rapidly becoming the go-to solution for businesses looking to integrate AI into their operations without blowing their budgets or overloading their systems. Unlike their bulkier counterparts, large language models (LLMs), SLMs are designed to handle specific tasks with impressive efficiency. This is a perfect example of how scaling down can mean scaling up results.
So, what makes SLMs such a game-changer? Their smaller size—fewer than 10 billion parameters compared to hundreds of billions in LLMs—means they can be trained faster, deployed quicker, and customized more precisely for niche applications. For industries like healthcare, finance, and legal services, where accuracy and privacy are vital, SLMs offer solutions tailored to their specific needs without the bloat of unnecessary functionality.
SLMs are also cheaper to run and more environmentally friendly. Training OpenAI’s GPT-3 consumed 500 metric tons of carbon—an energy cost many companies simply can’t justify. By contrast, SLMs deliver comparable results in focused domains while consuming far less computational power. This efficiency makes them ideal for businesses looking to prioritize sustainability alongside innovation.
For executives worried about data privacy, here’s another reason to care about SLMs: they work exceptionally well on-premises or in private cloud environments, which keeps sensitive information secure. Gartner Research even predicts a 60% increase in SLM adoption by 2025, driven by rising software costs and the need for efficient, trustworthy models. A Harris Poll backs this up, revealing that 75% of IT decision-makers believe SLMs outperform LLMs on key metrics like speed, cost, accuracy, and ROI.
“Smaller models punch well above their weight and are rewriting the rules for AI adoption. For industries with high stakes and tight margins, they’re a competitive advantage.”
Predictive AI and going back to basics for bigger results
Let’s talk about predictive AI, the original workhorse of artificial intelligence. While generative AI has dominated the headlines lately with flashy tools that create content, predictive AI quietly delivers results that businesses can bank on. It’s about using data to see what’s coming. And for companies looking to make smarter decisions faster, it’s becoming a must-have.
What makes predictive AI so effective? It’s simple: it uses historical data and machine learning to forecast future outcomes. Whether you’re managing inventory, maintaining equipment, or personalizing customer experiences, predictive AI can tell you what’s likely to happen and when. These insights turn guesswork into strategy, letting businesses cut costs, reduce downtime, and maximize efficiency.
Generative AI has its place, no doubt. But it’s not always the right tool for the job. Building a flashy, large-scale model to predict maintenance schedules or optimize a supply chain is like using a rocket to cross the street—overkill. Predictive AI is faster, cheaper, and purpose-built for these kinds of practical applications. That’s why over 50% of AI use cases in 2025 are expected to rely on predictive models, according to Forrester.
There’s also a growing trend of combining predictive and generative AI to get the best of both worlds. For example, a predictive model might forecast demand for a product, while a generative model creates custom marketing campaigns based on that forecast. This synergy is where things get interesting, with combined use cases expected to rise from 28% today to 35% by 2025.
When it comes to solving real-world problems, sometimes the future lies in rediscovering what already works. Predictive AI delivers measurable results, and for executives focused on ROI, that’s hard to ignore.
Generative AI is a revolutionary tool with real challenges
Generative AI systems create text, images, videos, even music, at a level that’s nearly indistinguishable from human work. But while the potential is jaw-dropping, the path forward isn’t without hurdles. Generative AI is impressive, but it’s also expensive, energy-intensive, and not always practical for every task.
The power behind generative AI lies in large language models like GPT-4, which can write essays, debug code, or even compose poetry. By 2025, an estimated 750 million apps will integrate these models, a testament to their transformative impact on industries like media, customer service, and education. But there’s a cost to all this innovation. Training GPT-3, for instance, consumed as much carbon as driving 1.1 million miles.
For many businesses, the question isn’t whether generative AI works—it’s whether it’s worth it. LLMs often deliver far more capability than most companies need for straightforward tasks like automating workflows. Here’s where smaller, more efficient alternatives like SLMs shine, offering focused solutions that balance performance, cost, and sustainability.
Even so, generative AI is far from a passing trend. The market is expected to skyrocket from $1.59 billion in 2023 to a staggering $259.8 billion by 2030. This growth reflects its potential to unlock entirely new ways of working. The key for executives will be aligning this potential with business needs, avoiding unnecessary complexity, and focusing on tools that drive real value.
Multimodal AI is causing disruption across industries
Artificial intelligence is moving beyond text and numbers into a world where multiple types of data—like images, audio, and video—work together seamlessly. This is the core of multimodal AI, a technology that’s opening up entirely new possibilities for innovation and problem-solving. Think of it as giving AI the ability to see, hear, and understand the world the way humans do.
At its heart, multimodal AI integrates different data types into a unified model, letting businesses tackle challenges that were previously out of reach. For instance, in healthcare, AI can now combine medical imaging, patient histories, and lab results to improve diagnoses and treatment plans. In financial services, it can analyze text from customer interactions alongside speech data to optimize customer support. And in industries like automotive, multimodal AI is enhancing autonomous driving by integrating inputs from cameras, LiDAR, and GPS systems.
Why does this matter? Multimodal AI makes those capabilities more accurate and practical. A single modality—like text or image alone—can only offer part of the story. Through combining modalities, AI can uncover patterns and correlations that would otherwise go unnoticed. This creates smarter, more adaptive systems.
The technology is still evolving, but it’s already making waves. Tools like OpenAI’s GPT-4 are paving the way by processing both text and images, while others are exploring video generation and multi-input learning. As businesses look to innovate, those that embrace multimodal AI will gain an edge in areas ranging from customer engagement to operational efficiency.
Leadership, data quality, and upskilling
The truth about AI adoption? It’s about people, processes, and leadership. As companies race to integrate AI into their operations, many are realizing that success depends on more than cutting-edge tools. It requires clean data, skilled teams, and leadership that knows how to connect technical capabilities with business goals.
First, let’s talk about data. AI is only as good as the information it’s trained on. If your data is messy, outdated, or incomplete, even the best models will fall short. A Capital One survey revealed that 70% of technologists spend hours every day fixing data issues. That’s valuable time wasted on problems that could be solved with better data management practices. By 2025, 30% of enterprise CIOs are expected to bring Chief Data Officers (CDOs) into their teams to address these challenges and ensure data quality becomes a strategic priority.
Next, there’s the issue of talent. AI isn’t a “plug and play” solution—it requires people who know how to implement and optimize it. Unfortunately, there’s a growing gap between what companies need and what the workforce is equipped to deliver. Upskilling employees in AI technologies has become a business-critical investment. Companies that prioritize training now will close this gap and gain a competitive advantage as AI becomes even more integral to operations.
Finally, leadership is key. C-suite executives must go beyond delegating AI projects to IT teams. They need to align AI initiatives with broader business strategies, making sure investments deliver measurable ROI. This is why roles like the CDO are becoming so important. These leaders act as a bridge between technical teams and business units, translating complex AI capabilities into tangible outcomes.
“For companies looking to make AI work for them, the formula is clear: clean your data, upskill your workforce, and empower leaders to think strategically about AI’s role in your business. Get these foundations right, and the rest will follow.”
Agentic AI brings immense promise, but not quite yet
Imagine AI agents capable of handling complex tasks autonomously, working across different systems, and adapting on the fly. This is the vision of Agentic AI—AI architectures designed to act like independent problem-solvers. It’s an exciting concept, but it’s not quite ready for prime time.
The potential here is huge. Agentic AI could transform industries by automating workflows that require decision-making, reasoning, and coordination between different AI models. For example, imagine an AI agent managing supply chain logistics, monitoring global disruptions, and making real-time adjustments—all without human intervention. That’s the promise of this technology.
But the reality? Agentic AI is still in its infancy. Building these agents requires aligning multiple models, integrating data from diverse sources, and customizing solutions for specific outcomes. These aren’t small challenges. Many companies that try to build agentic AI systems in-house end up frustrated, with Forrester predicting that 75% of these efforts will fail in 2025. The technology is also heavily reliant on retrieval-augmented generation (RAG)—a method for pulling in external knowledge—which remains complex and underdeveloped.
Despite these hurdles, optimism is high. Businesses are eager to explore agentic AI’s potential, even if most end up relying on consulting services or pre-integrated solutions from established vendors. And as the field matures, we can expect breakthroughs that will bring us closer to the vision of truly autonomous AI agents.
For now, the key for executives is to temper expectations. Agentic AI is worth exploring, but it’s not a quick fix. Focus on foundational AI technologies like predictive models and small language models, while keeping an eye on agentic AI developments. It’s a long game, but one that will pay off when the technology catches up to its potential.
Key takeaways for leadership
- Shift toward smaller language models (SLMs): SLMs are becoming the preferred solution for cost-efficient, task-specific AI applications. Leaders should explore integrating SLMs to improve operational efficiency and reduce resource consumption, especially in industries like healthcare, finance, and legal services.
- Predictive AI will dominate: As generative AI faces high costs and complex implementation, predictive AI will reclaim a significant share of use cases, particularly for operational needs like maintenance, demand forecasting, and supply chain optimization. Businesses should prioritize predictive AI investments for more immediate ROI.
- Multimodal AI as a competitive advantage: Multimodal AI, which integrates text, images, and other data types, is expanding capabilities in industries such as healthcare, finance, and autonomous driving. Executives should evaluate multimodal AI for innovations in customer experience and operational efficiencies.
- Data quality and leadership are critical: The success of AI initiatives hinges on clean data and strong leadership. Organizations should ensure data governance is a priority and invest in upskilling teams while empowering C-suite leaders, like Chief Data Officers, to drive AI strategies effectively.
