13/12/2024

The evolution of AI in 2025: a personal reflection

Author: Matthew Candy

It has been two short, but busy years since the explosion of ChatGPT opened the worlds’ eyes to the transformative impact of AI – where every day spoken and written language became a universal programming language, and the power of AI was put into all our hands. With 2025 just around the corner, I wanted to reflect on what we can expect to see with the evolution of AI in what promises to be another eventful year.

The buzz around Agentic AI is ramping up, and it’s only going to get louder through 2025. But for enterprises, staying grounded in broader AI trends is key. I believe there will be a shifting role in AI and the need for safe, integrated, and user-centred AI strategies will drive real impact across industries.

1. Emerging AI advancements

Agentic AI is here—now, setting standards for safe and powerful autonomy will be a must: As agentic frameworks emerge as a predominate theme in 2025, marking a fundamental shift from traditional AI tools to proactive agents, so too will questions around accountability and control of these increasingly autonomous systems. This will lead to an increasing focus on standards, processes, and tools for how we govern them. As we continue to define the “rules of the road” for generative AI, we must also think about how businesses can deploy responsible, safe agentic AI workflows, and provide robust and scalable orchestration across processes, applications, models and technologies.

Model Selection must consider performance and cost: I have always believed that organisations will need to adopt a diverse portfolio of AI models to accelerate transformation, ensuring they are integrated and coordinated effectively. Smaller, highly performant models which can be trained and tuned for different domain and industry tasks should not be overlooked in favour of the bigger, more well-known models, but instead look at how this diversity serves your business needs whilst being highly cost competitive (both financial and carbon cost).

2. Safety, governance, and collaboration

The tides will turn more toward open-source AI models: Open-source AI models will continue to gain promising momentum for their transparency, flexibility, cost efficiency, and customisation options. They will help enterprises reduce vendor lock-in and encourage ongoing community-driven innovation, supporting trustworthy AI strategies for all.

Scaling AI without governance will hit roadblocks: As many organisations continue to successfully scale their AI efforts, there will be an increasing number of businesses realising the imperative to adopt AI governance solutions and frameworks crucial for mitigating risk, reducing bias, and adhering to the evolving regulatory landscape.

Ecosystem integrations will make or break open-source model growth in 2025: The increasing use of open-source models means that application platforms need to integrate easily with a range of models across the technology ecosystem, allowing for greater interoperability and adapting to new AI developments quickly.

3. Measurement, scalability, and ROI

AI growth will demand better measurement of ROI: Enterprises are increasingly investing in AI, but understanding the ROI, aligning AI with strategic goals and measuring and orchestrating value remains complex. With rising AI investments, companies will need mechanisms to measure AI’s ROI beyond productivity, tracking more granular KPIs like customer satisfaction, identifying the capacity to develop new products and services and employee impact and NPS.

Unified AI control centres will become the backbone for big-picture insights: Centralised control centres that unify isolated AI efforts across enterprise functions such as HR, procurement, customer service and marketing, will be key for providing a “macro view” that allows organisations to manage AI data governance, shared learning, and efficiencies.

A seamless UX layer will be essential to making AI tools intuitive and accessible: As AI applications proliferate, a cohesive user experience (UX) layer will be become increasingly desirable. A unified user experience, through dashboards or virtual assistants, can simplify interactions with AI, making tools more accessible and intuitive. Elevating employee work and interaction to natural language & chat will remove friction, and also isolate the experience layer from the back-end applications and processes.

Learn more on AI-first Marketing in our whitepaper

4. Human-AI interaction and workforce adoption

Knowledge workers will be augmented with AI Agents and software: Knowledge workers, particularly in industries such as my own in consulting, will be increasingly using software assets and methods that include AI agents and applications to deliver solutions at scale and accelerate time to value for customers if they want to stay relevant – these roles will be augmented with AI in order to accelerate time to outcomes.

Smooth AI adoption will come from empowering employees: I’ve said this all along and it remains true that effective workforce adoption of AI requires robust change management strategies and a deep focus on people. This means empowering employees to work alongside AI, offering training on new AI tools, and aligning AI implementations with day-to-day tasks. It’s also crucial that your employees trust your AI, because AI that people trust is AI that people use.

Conclusion

As AI continues to evolve and becomes more sophisticated, so too does our collective understanding of AI frameworks, safety, and collaboration. Let’s continue to engage in thoughtful dialogue, pioneer new solutions, and create innovative approaches to ensure our AI advancements align with the best interests of all humanity.

Share

About the author

Matthew Candy is Global Managing Partner for Generative AI within IBM Consulting, and is a member of IBM Consulting’s Global Leadership Team. Matthew helps IBM clients around the world embrace this new era of technology, combining AI & experience to deliver meaningful business outcomes beyond just improved productivity & efficiencies, but as an accelerant to growth. In addition, Matthew serves as the global managing client partner for one of IBM’s largest accounts in the consumer industry.

Matthew Candy

Global Managing Partner, Generative AI – IBM Consulting

Generative AI: From ambition to impact

In an era where innovation drives success, IBM iX empowers DACH businesses to harness the transformative potential of generative AI. We don’t just implement technology – we architect breakthrough solutions that address your unique challenges.

Learn more