
How Small Language Models Deliver Big Business Benefits
There seems to be no limit to what artificial intelligence (AI) can help people do. But the tens of billions, even trillions of parameters used to train large language models (LLMs) can be overkill for many business scenarios.
Enter the small language model (SLM). SLMs are trained on relatively small amounts of specific data—fewer than 10 billion parameters or so. Because of their small size and fine-tuning, SLMs require less processing power and lower memory. This means they're faster, use less energy, can run on small devices, and may not require a public cloud connection.
Like LLMs, SLMs can understand natural language prompts and respond with natural language replies. They are built using streamlined versions of the artificial neural networks found in LLMs. But SLMs are trained on focused datasets, making them very efficient at tasks like analyzing customer feedback, generating product descriptions, or handling specialized industry jargon.
'LLMs are like a starship. It's very powerful and can go far, far away, but if you're doing something very tactical and specific, that starship is way too powerful,' says Neil Sahota, CEO of research firm ASCILabs and an AI advisor to the United Nations. 'If speed and costs are concerns, SLMs are the better way to go.'
The sweet spot for SLMs tends to be narrow tasks in high-volume niche applications or in low-power environments, such as on smartphones or Internet of Things (IoT) gadgets. They are also useful when data privacy is crucial, or internet access is sparse. For example:
Field service engineers don't always have high-bandwidth internet access. With an SLM on their device, they could use generative AI to query their field service manual. Low computational requirements and local processing make this possible.
Sales representatives might need to access a generative AI model containing sensitive data at a client site to provide tailored recommendations. An SLM could provide those results without the lag and potential privacy concerns that often come with using a mobile device.
Clinicians could use an SLM to analyze patient data, extract relevant information, and generate diagnoses and treatment options. The fact that data never leaves the device is a huge benefit for privacy.
But don't expect a significant shift from LLMs to SLMs. Organizations are more likely to implement a portfolio of models, each selected to suit a specific scenario.
AI developers, in fact, often work through a pipeline of models. A query might first go to an LLM, then to an SLM for classification, then back to the LLM to extract the information and generate a response.
At larger organizations, an LLM could be used for complex tasks—like developing a long-term business strategy that considers macroeconomic policies and global effects—while multiple SLMs handle dozens of business-unit-specific tasks such as analyzing consumer feedback and social media posts to guide new product development.
And while SLMs may be a cost-effective alternative to LLMs, they still have limitations. They don't understand complex language well, they lose accuracy when doing complex tasks, and they have a narrow scope of knowledge.
There are other trade-offs. While SLMs generally don't cost a lot to run, costs could add up if multiple SLMs are in use. 'If you have five models deployed and they're each using GPUs and occupying space and electricity in the data center, that costs more versus having one huge model,' says Sean Kask, AI chief strategy officer at SAP. 'Sure, the LLM uses a lot of electricity, but it's being used for a lot of different things, and you can refine data for smaller, more specific queries through prompt engineering.'
What's more, SLMs present many of the same challenges as LLMs when it comes to governance and security. 'You still need a risk and regulatory framework,' says Jim Rowan, head of AI at Deloitte Consulting LLP. 'You need an AI policy because you don't want business units using data and AI models without your knowledge. And you still have to set up guardrails because SLMs hallucinate too,' he adds.
SLMs also aren't necessarily easier to manage than LLMs. Even though the big AI players offer versions of SLMs through a service model where they provide the underlying engine, 'you still need people who know what the right data is. You need domain experts and a data scientist who can develop a good training strategy for the model,' Sahota says.
Companies will need to ask important questions before incorporating SLMs into their AI strategy:
What business case are you solving for? If the dataset is very small, controlled, and available, such as HR documents or product descriptions, it makes great sense to use an SLM. 'But if it's a large stack of constantly changing data or there's lots of variability in it, such as current mortgage rates or daily geopolitical events, you probably want to go the LLM route,' Sahota says.
What kind of performance and accuracy are needed? SLMs can be very accurate about straightforward questions, like an inquiry into current benefits. But if an employee says 'I would like to pay a third mortgage; can I draw off my 401(k)?' they may get a more generic answer. An LLM might be better at handling this type of question, as it could include information on HR and tax standards for 401(k) use.
What are your growth needs? Businesses need to anticipate how big the SLM might get over time. 'If you're a retailer and you're going to toss tens of thousands of products into the model over the next few years, that's certainly an LLM,' Sahota says.
As the number and type of available AI models continue to grow, businesses will need to understand the range of what's available to create their AI model portfolio.
'Choice is very important to your strategy,' Kask says. 'Pick the model that's right for you and for your embedded use case.'
This story also appears on SAP.com.
Hashtags

Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Yahoo
14 minutes ago
- Yahoo
Only one in six back reducing or removing coursework to avoid AI misuse
Only around one in six adults believe reducing or removing coursework is the best way for schools to avoid the misuse of AI, a survey has suggested. Public support on whether students should use artificial intelligence (AI) to improve their writing style in coursework is divided, according to a poll. A YouGov survey, commissioned by Cambridge University Press & Assessment, found 89% believed it was 'unacceptable' for pupils to use AI to entirely complete their school coursework. But the poll, of 2,221 adults in the UK, found nearly half (46%) believe it is acceptable for school pupils to use AI to improve the punctuation and grammar in their coursework, while 44% did not. Only 16% of UK adults believe reducing or removing coursework completed at home is the best way for schools to avoid student AI misuse. The chief executive of exam board OCR is calling for a co-ordinated national strategy on AI. It comes after the independent curriculum and assessment review said it would consider reducing the 'overall volume of assessment' at GCSE. But the interim report, published in March, said the review had heard about the 'risks' to standards and fairness concerning AI in relation to coursework. The review – chaired by education expert Professor Becky Francis – will publish its final recommendations in the autumn. The YouGov survey, which was carried out in June, suggests more than three in five (62%) of UK adults oppose teachers using AI to mark coursework, while 27% support it. But the majority (59%) support teachers using AI to complete their administrative tasks, such as lesson planning. Jill Duffy, chief executive of OCR, said: 'AI is already in our schools and is not going away. 'A co-ordinated national strategy, with funding to ensure no schools are left behind, will build public confidence in its transformational potential. 'The public is clear that coursework is too important to lose, even in the age of AI. 'It enables us to test different skills, and to reduce the intense volume of exams taken at 16. 'These findings should be seen as a challenge to all of us in education: find a way to adapt coursework so it is fit for the AI century.'
Yahoo
42 minutes ago
- Yahoo
GM, Redwood Materials sign deal to deploy energy-storage batteries
This story was originally published on Automotive Dive. To receive daily news and insights, subscribe to our free daily Automotive Dive newsletter. Dive Brief: General Motors and battery recycler Redwood Materials have signed a non-binding memorandum of understanding to accelerate the deployment of stationary energy storage systems built using both new modules and second-life batteries from the automaker's electric vehicles, according to a July 16 press release. The plans are part of a new business unit launched by Redwood Materials in June named Redwood Energy that's focused on assembling and deploying low-cost stationary energy-storage systems to help meet growing power demands of AI data centers and other applications. 'Electricity demand is climbing, and it's only going to accelerate,' Kurt Kelty, VP of batteries, propulsion, and sustainability at GM, said in the release. 'To meet that challenge, the U.S. needs energy storage solutions that can be deployed quickly, economically, and made right here at home.' Dive Insight: Redwood Materials plans to establish a domestic supply chain to manufacture batteries to support energy storage applications via its Redwood Energy unit. The company currently repurposes around 20 gigawatt hours of batteries annually, which is the equivalent of 250,000 EVs or roughly 90% of all lithium-ion batteries and battery materials currently recycled in North America, according to its website. GM is already providing the battery recycling company with used batteries to help power what Redwood calls 'the largest second-life battery development in the world and the largest microgrid in North America' at an installation in Sparks, Nevada, per the release. The electricity produced by the microgrid is being used by AI infrastructure company Crusoe. 'Electricity demand is accelerating at an unprecedented pace, driven by AI and the rapid electrification of everything from transportation to industry,' said JB Straubel, founder and CEO of Redwood Materials, in the release. 'Both GM's second-life EV batteries and new batteries can be deployed in Redwood's energy storage systems, delivering fast, flexible power solutions and strengthening America's energy and manufacturing independence.' According to GM's press release, U.S. electricity demand is being driven by the increasing power demands of AI data centers. By 2028, the percentage of the national grid's electricity used by data centers is expected to triple from 4.4% to around 12%. This creates the need for additional energy storage units to help offset power outages and reinforce the grid during peak demand or when power delivery is limited. In May 2024, Redwood Materials announced it would recycle cathode anode materials and production scrap from GM's Ultium Cells battery joint venture plants in Warren, Ohio, and Spring Hill, Tennessee. The recovered materials will be processed and used to produce new batteries. In addition to GM, Redwood Materials is working with Toyota, Ford Motor Co. and Volvo Cars on battery recycling initiatives in the U.S. The efforts are intended to make EV batteries more sustainable, as well as lessen reliance on imported critical raw materials, most of which are currently sourced from processors and refiners in China. Last September, Redwood Materials and BMW announced plans to recycle second-life batteries from the automaker's network of nearly 700 locations in the U.S., including dealerships and distribution centers to recover critical minerals, including nickel, cobalt, lithium and copper. Redwood aims to return 95% to 98% of these critical minerals back to the domestic battery supply chain. Redwood Materials was founded in 2017 by Straubel. He was the co-founder of Tesla and the company's former CTO. Straubel was nominated to Tesla's board of directors in April 2023. In February 2023, Redwood Materials received a $2 billion loan commitment from the Department of Energy to expand battery materials recycling to support the growing EV market in the U.S. GM and Redwood Materials expect to announce more details about their planned collaboration later this year, according to the release. Recommended Reading Ultium Cells, Redwood Materials to partner on EV battery recycling Error in retrieving data Sign in to access your portfolio Error in retrieving data Error in retrieving data Error in retrieving data Error in retrieving data


Fast Company
an hour ago
- Fast Company
AI For marketing: When to use it, when to lose it
Savvy marketing leaders recognize that AI is a powerful tool that can be used to reshape how teams operate. But even as AI tools rapidly improve, there are still limitations in today's technology that demand titration in how it's used. The best marketers are those who understand how far to push the AI envelope within their business and market context. And now is the time for marketing leaders to expand their existing skill set and develop their AI intuition—when to use it, when to lose it—to make the most of AI's transformative power and deliver the best marketing work possible. By combining broad AI fluency through adoption with a willingness to remain agile as the technology matures, teams can thrive and make the best use of AI for the biggest impact. With that in mind, here are AI-related 'do's and don'ts' every marketing leader should know. AI as a research assistant, not a business strategist If teams don't understand the context in which their business is operating, they won't be effective marketers. The best marketing leaders are absolute experts on their own companies, with knowledge of the details of their offerings, customers, competitors, and broad industry trends. When marketers understand their business this deeply, they can build programs that align with company priorities, connect with customers, and fuel their company's growth, making it easier to secure budget and executive buy-in. Do use AI as a cheat sheet to stay on top of the market and competitors, and to improve business intelligence across departments, like finance and product. AI can analyze mountains of internal and external data and content it would otherwise be impossible to comb through, like companies' annual reports, customer and competitor earnings, press releases, and newsletters. At Guild, my teams turn to AI for tasks such as market sizing, benchmarking against other marketing organizations for budgeting, and business planning and competitive analysis. AI can also be helpful with pricing research and gauging customer insights. Don't expect AI to replace the nuanced understanding of a business's priorities or its context. Ultimately, teams should create their own strategies, but using AI with business context will make those strategies better and more complete. AI can't read betweenthe lines or predict the future—only meaningful conversations, deep curiosity, and astute understanding of one's business can do that. Use AI as a production designer, so creatives can shine Campaign work in marketing can quickly become rote, tying designers and writers up with tasks that are far from strategic or highly creative. All variants of ad design or ad copy can be generated by AI. It can take a first pass at any blog content, social media graphics, or resizing assets for different platforms. Let AI bring ideas to life faster and more efficiently to free up creatives to focus on developing breakthrough campaign concepts and compelling brand stories. Do use AI to rapidly prototype visuals, generate multiple design variations, and handle time-consuming production tasks. AI excels at replicating creative direction so take advantage and use in moments where multiple iterations are needed. Don't rely on AI for creative inception or novel campaign ideas. This is where creatives shine. Use AI to handle the heavy lifting and transform the creative work into end product downstream so creative strategy and brand decisions are owned by creatives themselves. Use AI to accelerate pointed insights With AI, leaders can access an abundance of real-time data, turning marketing into a velocity engine for optimization and peak performance. The speed to insights around what's working and what isn't, along with the ability to move with agility, is where the real power of AI lies for marketing leaders. Whether that's analyzing customer data to spot engagement patterns, detailed segmentation in email open rates, or digging into a competitor's recent launch, AI helps leaders get to the root of 'why this matters' faster than ever before. Do use AI to quickly analyze performance data, competitor portfolios, or deep segmentation analysis. Be specific and give AI the full context in your prompts. Instead of asking 'How did this campaign perform?' try 'What is variance by audience segment in each of these ad headline click-through rates?' This is a case where AI can be a game-changer when giving pointed insights. Don't treat AI-generated insights as final answers. AI may be able to determine if certain creative elements performed well, but it can't determine if that approach aligns with your strategy and broader business objective. Just because AI quickly spots a pattern doesn't mean it requires changing the work itself. Successful marketers will need to develop the skills to quickly decipher what insights matter and why. The marketing leaders who will truly shine in the AI era will be those who master the art of balancing technology with their human instincts. In order to be successful, they'll need to leverage AI's power as it exists today, and be agile enough to adapt as the rate of improvement around that technology shifts. Today, that means letting AI handle much of the heavy lifting—research, data analysis, creative production work—to improve expertise and create leverage across the marketing organization. As AI technologies rapidly evolve and mature, the best leaders will dynamically change with it so that their organizations can deliver better, faster, and more novel results for the business.