AI will remain the biggest technology story over the year ahead, with many expert predictions focused on its transformative potential. Drawing on insights from leading vendors, analysts and journalists, here's a roundup of five thought-provoking AI predictions for 2025 and beyond. They highlight how AI will continue to push boundaries, presenting both opportunities and challenges for businesses and individuals alike.

1. AI is likely to "shrink"

One interesting prediction highlighted by Computing magazine is the rise of small language models (SLMs), which are expected to take some of the spotlight away from larger language models (LLMs) in 2025.

So far, much of the hype and excitement surrounding generative AI has revolved around the large models offered by OpenAI and Anthropic. These models are trained on petabytes of data, making them very versatile and able to respond to requests across a wide range of topics. However, building and training these massive models is a monumental task that is both computationally intensive and extremely costly. They also have a significant environmental impact, as they consume vast amounts of data center power.

This is why, according to Computing, we're likely to see a shift towards more focused, smaller models. These SLMs are less costly to train, more efficient, practical for niche applications - and more sustainable.

2. AI will help double the workforce

PwC predicts that the workforce could double in 2025 thanks to the rise of AI agents.

Rather than responding to individual prompts, AI agents function as compound systems that can work independently to tackle more complex tasks. According to PwC, this includes autonomously performing tasks such as "handling routine customer inquiries" or "turning human-provided design ideas into prototypes".

We'll see AI agents integrated into the workforce alongside human workers. Workflows will change dramatically, with humans setting tasks and interacting with AI agents like they currently do with independent creative teammates. However, business leaders must provide assurances that AI is intended to enhance people's value, not displace them.

Anthony Abbatiello, PwC Workforce Transformation Practice Leader, suggests that overall this will be good for business. "AI agents are set to revolutionize the workforce, blending human creativity with machine efficiency to unlock unprecedented levels of productivity and innovation."

3. Bosses will use AI to measure workplace sentiment

In its technology predictions for 2025 and beyond, Gartner discusses the potential of AI to perform sentiment analysis on interactions that happen in the workplace. It pictures employers using these insights to track overall sentiment among their workers - and stepping in to encourage behaviours that are most likely to produce a more motivated, engaged and productive workforce.

In fact, Gartner predicts that as early as 2028, 40% of large enterprises will use AI to "manipulate and measure employee mood and behaviours, all in the name of profit."

Using AI-driven behavioural technologies like this might hold the promise of increasing productivity, but how will individual employees feel about being monitored in this way? Gartner analyst Daryl Plummer says that "companies must balance efficiency gains with genuine care for employee well-being to avoid long-term damage to morale and loyalty" if they plan to use AI in this way.

4. Open-source will be key to helping organizations scale their use of AI

Over the next three years business executives expect their AI model portfolios to include 63% more open models, helping to bring down the cost of AI. That's according to research highlighted in the '5 Trends for 2025' report by the IBM Institute for Business Value.

Bill Higgins, leader of Watsonx Platform Engineering and Open Innovation at IBM Research, predicts that open-source AI will thrive in 2025. Despite the hype, many companies are still struggling to see measurable returns on their AI investments, largely due to the high costs of licensing proprietary models. Open-source AI models could bridge this gap by offering a more cost-effective and accessible alternative.

Open models have "friendlier" pricing, are more auditable and transparent and support multi-cloud infrastructures. The number of open models is increasing and they can be applied to many industry-specific use cases without fees or API call costs. All of this paves the way for open-source AI to come into its own in 2025.

5. GenAI coding will thrive, but won't replace developers anytime soon

Using AI to help developers with coding tasks has emerged as one of the earliest and most widely accepted use cases for generative AI. Forrester expects the growing use of genAI coding assistants to improve developer productivity. However, the analyst house predicts that in 2025 at least one organization will try to replace 50% of its developers with AI - but will fail.

That's because employers are ignoring all the other work that developers do aside from coding (working on designs, writing tests, fixing bugs, and meeting with stakeholders etc). Even when coding assistants are deployed you always need a human in the loop - to check that the suggested AI code is correct and performs efficiently - and to filter out poor code from going into production.

From smaller, more efficient models and its role in reshaping the workplace, to the rise of open-source models, AI is poised to evolve and continue redefining our working lives through 2025 and beyond.

This blog was first published on the IBM Community.