What's Next for AI in 2025?
2025 is set to be a pivotal year for AI, with shifts in administration, evolving regulations, and monumental investment projects already underway. Agentic AI, generative AI, experiential AI, and small language models all stand to reshape work, education, and governance. Safeguards and training will be needed, new laws will be tested, and breakthroughs in critical areas like healthcare, drug discovery, and climate are forecasted to make significant impact.
So, what should you watch for in AI this year? Read what our leaders are keeping an eye on.
Does your team need help navigating and meeting the demands of the fast-moving AI landscape? Talk with our team!
What’s next for AI in 2025?
Stakeholder expectations will drive interest in experiential AI.
“I think we’ll see a lot of the hype around AI come to an end, and then a realization that there’s a gap between AI’s amazing promises and disappointing deployments. Stakeholders now are expecting real ROI on real tasks. I think this will include a maturation of Experiential AI (AI with human in the loop) as AI will be thought of as a helper tool that can learn. This will mean an uptake of much smaller models that are specialized because learning and fine-tuning is too hard on massive Large Language Models (LLMs). The biggest worry I have is the continuing enablement of sophisticated misinformation and deep fakes, which will take us to a world where we cannot trust anything we see or read in our mobile online world. The bad actors have a huge advantage over the cybersecurity and infosec defenders, and this will get much worse in 2025 as attacks become more sophisticated and adaptive."
- Usama Fayyad, executive director, Institute for Experiential AI; professor of the practice, Khoury College of Computer Sciences
AI will join the healthcare team.
“In 2025, I'm seeing AI transform healthcare delivery through its ability to analyze vast amounts of patient data and identify unique patterns that help clinicians make more informed decisions about prevention and treatment. What's particularly encouraging is how AI has become a powerful complement to human medical expertise, enhancing rather than replacing the critical thinking and emotional intelligence that only healthcare professionals can provide. At Bouvé College, we're closely monitoring how these tools are improving workflow efficiency and reducing administrative burden, and are incorporating them into our experiential learning models, so that students are given a competitive advantage that makes them highly desirable in the workforce and savvy users of technology and data.”
- Carmen Sceppa, dean, Bouvé College of Health Sciences
Legal questions will get thornier.
“The best way to change the future is to build it, as Alan Kay said. I will be watching, 1) the enforcement of the EU AI Act, which from next month imposes a minimal AI literacy on all companies; 2) the court cases of The New York Times, New York Daily News, Getty Images, and others against companies that build generative AI models violating the copyright in their training data; and 3) the initiatives of the Trump administration related to AI.”
- Ricardo Baeza-Yates, director of research, Institute for Experiential AI; professor of the practice, Khoury College of Computer Sciences
Businesses will face new risks with the expanded deployment of enterprise AI.
“I think one of the most significant AI trends this year will be the expanded deployment of enterprise AI, LLMs, and agentic AI systems within business contexts, and the complex ethical challenges they present. A critical question is how organizations will navigate issues such as decision-making opacity, unintended bias, and the risks associated with over-reliance on autonomous systems. Adoption of robust and comprehensive Responsible AI strategies that ensure fairness and accountability will likely determine which organizations succeed in gaining public trust. And this applies not only to private companies but also to higher education institutions, healthcare organizations, and governmental agencies, if they intend to be successful in their AI adoption and transformation.”
- Cansu Canca, director of Responsible AI Practice, Institute for Experiential AI; research associate professor at the Department of Philosophy and Religion
Learn about the Responsible AI Practice.
Agentic AI will raise questions about human relevance.
“In 2025, we will experience the rapid rise of agentic AI. As interest in this technology grows, so will our curiosity about what it means to be human and how we define our relevance. The accelerating combination of generative AI with robotics and social media will prompt us to rethink what we consider original thoughts and meaningful interactions. The creation of impressive video content will become more accessible, accompanied by an increase in the prevalence of deepfakes, or the licensing of individuals' likenesses. Keep an eye on net job creation/losses as the pressures to create ROI for the enterprise mount.”
- Tim Weidinger, director of business development, Institute for Experiential AI
The face of work will begin to change.
“I’m interested to see how various industries begin to grapple with how AI or GenAI will transform work within their domain. As the gap between knowing what AI is and using it begins to narrow, we are going to see more creative ways that people begin to use AI to not only improve efficiency and administrative aspects of work but also more complex aspects. I think by end of 2025, we will hear more about upskilling. Here at Northeastern, I expect more and more students will have had exposure to AI and GenAI in their coop."
- Rupal Patel, affiliate faculty; professor, Khoury College of Computer Sciences and Bouvé College of Health Sciences
Governments will adopt AI to improve service and engage the public.
“As concerns about AI's risks dominate headlines, a quieter revolution will unfold as governments increasingly deploy AI to enhance civic engagement, improve service delivery, and make sense of public feedback, demonstrating how these tools can strengthen rather than undermine democracy. This shift from AI doomerism to institutional innovation will begin to reshape how we think about technology's role in democratic governance. 2025 is the year for democratic AI."
- Beth Simone Noveck, core faculty, Institute for Experiential AI; director, The Burnes Center for Social Change
AI agents will begin to work alongside humans in the workplace.
“I think AI agents, powered by LLMs, will dominate 2025. Building on theoretical advancements from 2024, these agents will integrate conversational AI with functionality-specific tools to automate human workflows. For example, call center applications will see LLMs engaging with customers to interpret caller intent, followed by automated actions like completing transactions with text-to-SQL functionalities. I think we’ll also see the proliferation of collaborative AI agents capable of working alongside humans and other autonomous systems. These agents can provide real-time decision-making and project management assistance.”
- Omer Alis, associate director of AI Solutions Hub, Institute for Experiential AI
Read more predictions from Omer.
Human-centered AI will be more important than ever.
“As AI becomes more integrated into daily life across the globe, the need for it to remain human-centered will become even more critical. AI offers the potential to enhance efficiency, enrich lives, and facilitate global communication. However, it is essential to prioritize equity, recognizing that AI systems still carry significant biases. By leveraging our human intelligence, we can continuously improve this technology, mitigate harm, and ensure it serves as a powerful tool to advance fairness and inclusivity.”
- Christie Chung, executive director, the Mills Institute
Looser regulations will deter liability-conscious healthcare companies.
"For me, the biggest thing to watch is on the regulatory side. The new administration will likely loosen many of the regulations that were initiated under the Biden administration. If successful, it will create an interesting landscape. Companies may be more aggressive in bringing AI technology to market. While this may spur more innovation, it may also lead to development of some technologies that don’t get properly vetted or go through appropriate safeguards by regulatory agencies such as the FDA. This may place even greater burden on consumers, particularly healthcare providers and healthcare organizations, to be accountable when adopting these technologies. I imagine that with looser regulations, healthcare providers and organizations will take a more conservative approach to adoption to mitigate potential harm to patients and exposure to liability. They will in turn become the key stopgap for adoption. A likely difference-maker for adoption will be healthcare provider education to develop AI literacies, which is the other major thing to watch.”
- Eugene Tunik, director of AI + Health, Institute for Experiential AI; associate dean of research and innovation, Bouvé College of Health Science
Learn about AI + Health.
Data limits and ROI pressures will select the industry’s winners and losers.
“Like many others, I'll be watching for diminishing performance increases with upcoming AI releases from the major players in the space. There's a concern that we're effectively running out of good data to throw at LLM training (also that the general approach used to train LLMs has hard limits). My prediction is that we will start to see increased pressure on AI delivering operational value to organizations, and winners and losers will start to emerge amongst big companies, in higher ed, and governments.”
- Sam Scarpino, director of AI + Life Sciences, Institute for Experiential AI; professor of the practice, Khoury College of Computer Sciences and Bouvé College of Health Sciences
Learn about AI + Life Sciences.
Global AI investments will start to show returns.
“I am very impressed with how effective computing is at the edge. In addition to LLMs moving to phones, I am impressed with what you can do with Network Attached Storage (NAS), or cloud computing in your house. These boxes cost less than a laptop. Disk space at the home is $10/TB for the lifetime of the disk, much less than in the cloud ($300/TB/year). Currently, you have to know too much to manage an NAS, but there is no reason why that can't be fixed, but it will take more than a year to happen. Another trend is globalization. China has been investing huge amounts in AI. Those investments will become more obvious in time. I'm also seeing more investments in languages in Southeast Asia and India. LLMs will move beyond English.”
- Kenneth Church, senior principal research scientist, Institute for Experiential AI; professor of the practice, Khoury College of Computer Sciences
AI-designed drugs will hit their stride.
“I think over the next 12 calendar months you’re going to see a lot of work in actually developing new molecules using AI tools. For decades, we’ve been using programs to do small molecule-type iterations on drugs. I think we’re going to move aggressively over the next 12 months into the design and implementation of large molecules, so protein-based drugs—antibodies, for example. I think you’ll see a lot of that work accelerate over the next 12 months, and then I would say within the next five years, you’ll see a number of those products hitting the market.”
- Jared Auclair, dean, College of Professional Studies
The question will be, “What can we learn from AI?”
“2025 will be the year when the field of AI shifts from the question ‘what can machines learn from humans’ to the question ‘what can humans learn from machines.”
- David Bau, core faculty, Institute for Experiential AI; assistant professor, Khoury College of Computer Sciences
New and better interfaces will facilitate human-AI collaboration.
“My prediction for 2025 is focused on the interface side of things. The way we’re interacting with our computers, phones, and other applications is going to radically transform. Think about email. There are obviously already applications, such as Copilot, that help you write emails, but I think the way we’re interacting with these applications is going to radically change to a way that is more focused on interacting with AI. I do expect that there’s going to be more applications, and generative AI will create better videos and better images, but I think the interesting part is on the application side. That requires the change of an interface between human and AI, and I think 2025 will see a dramatic change there.”
- Casper Harteveld, core faculty, Institute for Experiential AI; associate dean of graduate programs and professor of game design,College of Arts, Media and Design (CAMD)
AI-generated educational courses will emerge—but with a predictable set of challenges.
"I think we're going to see a lot of progress made in the area of AI-generated courses. In 2025, these systems are going to continue to improve and facilitate personalized, modular, multimodal learning experiences. AI-generated learning experiences will be characterized by a few features: 1) prompt-based design, where a user can generate a learning experience with a high-level prompt; 2) AI-generated outputs like syllabi, interactive modules, assignments, and assessments; 3) adaptive learning paths; 4) multi-modal content like text, videos, animations, simulations; 5) localization or auto-translation into multiple languages. Am I looking forward to this? Yes, but I'm also worried about quality control, bias, inclusivity, fairness, intellectual property issues, alignment with human educators, and the potential for miseducation, diseducation, and maleducation."
- Joe Doiron, associate director, Education Programs, Institute for Experiential AI
Learn about our education programs.
Better, more useful foundation models in life sciences.
“More useful foundation models in life sciences. Not sure if we'll get any real, giant leaps in AI capability in 2025, but I think AI hype will continue to grow. Someone to watch out for is Yann LeCun!”
- Ramkumar Hariharan, senior scientist, Institute for Experiential AI
New and better foundation models for physiological data.
“We will start seeing better foundation models for neuroimaging and physiological data. These foundation models will in turn enable better predictive and exploratory fine-tuned models that can help us understand human behavior as well as predict the onset of neurological conditions such as delirium.”
- Zulqarnain Khan, research scientist, Institute for Experiential AI
The Institute for Experiential AI will continue to make inroads in health and life sciences.
“I look at 2025 as a year filled with dynamic and exciting opportunities. At the Institute for Experiential AI, we're evolving rapidly. Since entering our fourth year, we've been expanding and shaping our pipeline with strategic partnerships set to create lasting impact in health and life sciences for years to come. Specifically, in life sciences, we’re diving deeper into omics, particularly proteomics and transcriptomics, to explore new possibilities in personalized medicine. In health, we’re shaping the “hospital-at-home” model to redefine care delivery and make it more efficient and patient-centric. Additionally, education remains a strong pillar of our approach. Through expanded educational initiatives, we aim to provide the insights and tools our partners need to navigate this rapidly evolving landscape.”
- Maria Giovanna Trovato, global strategy & business development in healthcare and life science, Institute for Experiential AI
AI agents will make the case for AI more generally.
“Agents. I’m curious to see how companies rebrand and position their tools as agents. I think the concept of an AI agent will make it easier for customers to understand how to use AI. Personally, I'm looking forward to more agents. I would love to be able to talk to my phone and have it set up things instead of staring at a tiny screen all day trying to remember passwords to random sites.”
- Sean Martin, business development manager, Institute for Experiential AI
Multimodal, heterogeneous data will have its day.
“I predict that systems integrating multimodal, heterogeneous data will show more promise in high-stakes scenarios, with a focus on responsibility and social good.”
- Resmi Ramachandranpillai, postdoctoral research associate, Institute for Experiential AI
