From Security to Generative AI: IBM Experts Share Predictions for 2025
As the end of the year approaches, IBM has selected a set of predictions that could be on the agenda for 2025.
As the end of the year approaches, IBM has selected a set of predictions that could be on the agenda for 2025. These are bets ranging from security to generative AI, based on the opinion of the company's executives and their extensive knowledge of the market and scenarios.
AI Agents are Here and Now
"It will be essential to build barriers for secure and powerful autonomy: As AI agents become a main topic in 2025, marking a fundamental shift from traditional AI solutions to proactive agents and teams, questions will also arise about the responsibility and control of these increasingly autonomous systems. This will attract greater attention to the barriers, processes, and tools of how we manage agents in order to generate trust for this new and powerful wave of AI capabilities. It will also increase the need to enhance the skills of employees across all disciplines and leadership levels so they can responsibly develop, use, and supervise solutions."
u2014Ritika Gunnar, General Manager of Data and Artificial Intelligence, IBM
Follow us on: 
The role of "individual contributors" will evolve
We will all be agent managers: We are entering a new chapter on how employees perform their work with the emergence of AI agents. Unlike assistants, agents have the ability to generate plans based on a notice and perform tasks independently. They are more effective when they focus on specialized tasks and work alongside other agents on complex requests. As AI agents become more common, companies will need to reassess their work processes and create new types of teams for humans to supervise groups of autonomous AI agents.
- Jill Goldstein, Global Managing Partner, Talent Transformation and Human Resources, IBM Consulting
Open Source AI Will Drive Business Adoption
Despite increasing pressure, many companies still struggle to demonstrate the measurable return of their AI investments, and high licensing fees for proprietary models are a significant factor. By 2025, open-source artificial intelligence solutions will emerge as a dominant force to close this gap. Thanks to community-driven development and open-source models, they are quickly equaling the best patented offerings in strength and the proliferation of open AI solutions for specific industries and tasks will make it easier than ever for organizations to apply them to a wide range of innovative use cases, without fees or API call costs. With its friendlier cost structure, greater transparency and auditability, and support for multi-cloud architectures, we expect open-source AI to be fundamental in helping organizations scale beyond experimentation and start obtaining returns next year.
u2014Bill Higgins, Watsonx Platform Engineering and Open Innovation, IBM Research
Automation Becomes a Requirement for AI
2025 will be the year of AI initiatives, when technology-driven automation has reached a tipping point from something desirable to a requirement. In short: automation is necessary to solve the complexity of AI. Organizations can now confidently move forward and scale their initiatives using automation, moving from spending time managing and maintaining AI applications and IT environments to proactively detecting and solving issues. Automating these tasks will be essential to gain a competitive edge. Next year, you can't have a conversation about AI without talking about automation, and vice versa: you can't have a conversation about automation without talking about AI.
u2014Bill Lobig, Vice President of Product Management, IBM Automation
The acceleration of purpose-adapted AI will increase the performance and security of the mainframe
By 2025, we will see companies adopt an AI-compatible approach using specific hardware, especially on mainframes that handle large volumes of transactional data. These hardware accelerators, which can be delivered on chips and external cards, allow the use of traditional AI models alongside LLM language models based on encoders at the user's choice, enhancing large-scale and real-time data analysis and information for sectors such as banking and insurance. Because this approach allows AI workloads to stay on-premises, it also improves the security management, resilience, and compliance process for mainframe operators in regulated industries, allowing them to unlock new levels of efficiency and insight, setting the new standard for predictive outcomes.
- Tina Tarquinio, Vice President of Product Management, IBM Z and LinuxONE
As organizations begin the transition to post-quantum cryptography next year, agility will be crucial to ensure systems are prepared for continuous transformation, particularly as the US National Institute of Standards and Technology (NIST) continues to expand its toolkit of post-quantum cryptography standards. The initial standards from NIST were a signal to the world that now is the time to begin the journey towards quantum security. But equally important is the need for cryptographic agility, ensuring systems can quickly adapt to new cryptographic mechanisms and algorithms in response to changing threats, technological advancements, and vulnerabilities, ideally leveraging automation to streamline and accelerate the process.
"-Ray Harishankar, member of IBM, IBM Quantum Safe"
The Emergence of Shadow AI
In recent years, companies have been dealing with Shadow IT: the use of unapproved cloud infrastructure and SaaS applications without the consent of IT teams, potentially opening the door to data breaches or non-compliance. Now, businesses are facing a new challenge on the horizon: Shadow AI. Shadow AI has the potential to be an even greater risk than Shadow IT because it not only impacts security but also other natures. The democratization of AI technology with ChatGPT and OpenAI has broadened the scope of employees who have the potential to put confidential information in a public AI tool. By 2025, it's crucial for businesses to act strategically to gain visibility and maintain control over their employees' use of AI. With policies on AI use and the right hybrid infrastructure, businesses can be better positioned to manage confidential data and the use of applications.
u2014Nataraj Nagaratnam, Chief Technology Officer of Cloud Security at IBM
Multimodal AI, especially for processing complex documents, will grow significantly within the company
Multimodal AI is prepared to generate substantial value for businesses by enabling them to unlock more value from their data. Multimodal AI models are capable of processing and analyzing all types of complex documents with enriched content embedded in the form of images, tables, and graphs. These models are also evolving to support other modalities, such as audio and images, opening up countless new possibilities for knowledge. As a result, organizations will have to start bringing order and method to the way they handle all this unstructured multimodal data to prepare them for enterprise AI. This will put pressure on existing infrastructure, including higher storage requirements and robust management solutions.
u2014Sriram Raghavan, Vice President of IBM Research for AI
Companies will combine AI and automation technologies to achieve sustainability goals for 2030
Companies have bold sustainability objectives for 2030, but they also have a more complex infrastructure and more data sources than when those objectives were first announced years ago. By 2025, organizations with sustainability ambitions and objectives must implement AI-driven automation capabilities, including observability, resource management, and application lifecycle management. These capabilities can help reduce pressure on data centers, including managing energy consumption and improving the performance and life cycle of assets, which can ultimately help advance overall sustainability goals.
u2014"Kendra DeKeyrel, Vice President of ESG and Leader of Asset Management Products at IBM