In identifying these themes, this writer has predicted trends that governments and organizations alike can leverage to strategize responsible AI development

As the world tries to make sense of the impact that AI will exert in 2025, three major interconnected themes are emerging: authenticity, applied value, and autonomous agents. Why?

First, the rise of AI-generated content is creating an authenticity crisis for large language models (LLMs). As AI-generated content proliferates online, the creation of future models imbibes the risk of being trained on such AI-generated data rather than authentic human-created content. Simultaneously, businesses are restricting access to high-quality information, and legal battles over data usage are increasing. This shift threatens the quality and trustworthiness of AI models. As authentic data becomes scarcer, proving data provenance will become essential, and corporate data may be the next target for AI training.

Second, according to the person who made the predictive observations — Kelly Forbes, co-founder and Executive Director, AI Asia Pacific Institute and Council Member, Qlik — people will start seeing beyond the AI hype and expect to benefit from applied value. The focus must shift towards practical applications that deliver clear business value. Without this, some AI projects risk being abandoned. Organizations will need to balance market trends with internal needs to deploy AI effectively.

 Kelly Forbes, co-founder and Executive Director, AI Asia Pacific Institute and Council Member, Qlik

Three intricately-knit themes
Third, AI-driven agents will increasingly be granted autonomy to execute tasks independently and adapt to feedback. However, their success will depend on data authenticity and securing applied value, forming the foundation for effective deployment to solve complex problems beyond current human capacities. Multi-agent architectures will emerge, requiring interoperability to maximize reach and value. This shift will force a rethink of application development: some tasks may no longer require traditional apps, as agents can fetch answers directly. Businesses will need to balance building, buying, and adapting intelligent applications that evolve dynamically, learning from data to deliver predictive and personalized experiences.

According to Forbes, the three themes do not exist in siloes: they interact with each other. Without authenticity, no value can be realized. Without demonstrable value, the resources required to deploy agents and unlock their huge potential will not be made available. Nestled within these three themes are specific trends, which, if followed, can drive a positive impact and shape how businesses invest in, deploy, and benefit from AI in 2025. 

Thematic AI trends to watch for

Within each of the themes, Forbes has predicted several trends:

Authenticity

  • Data has to be trustable: Data quality is crucial but increasingly hard to prove. Efforts such as the EU AI Act can help, but more will be needed. The current focus is on how a model was created and trained, but we need to be able to signal whether a model can be trusted. An AI Trust Score can act as a filter through which all data should go, establishing provenance, lineage, and ultimately, overall trust. Data profiling markers will become important, particularly discoverability, accuracy, consumability, timeliness, security, and diversity.
  • Data has to be interoperable: Open table formats are emerging as industry standards, enabling flexible data organization, reducing costs, improving governance.
  • Dark data is waiting to be tapped: Organizations will use AI to analyze vast amounts of unstructured and unused data for insights. Businesses will be racing to tap into this hidden resource.
  • More AI/data marketplaces will spring up: High-quality, private data is becoming increasingly valuable, leading firms to treat it as a product. With sufficient demand, there will be more marketplaces offering vetted, ethical, and quality-controlled data and AI assets.

Applied value

  • Increased scrutiny of AI co-pilots: While such agents boost efficiency, their value is being questioned, with some providing low-quality insights at high costs. Better use-case understanding, proactive anomaly detection, and deeper problem-solving will be needed.
  • Cost governance will be key: AI operations are becoming increasingly expensive, with rising costs for generative AI prompts and energy consumption expected to surpass other IT initiatives by 2027. Businesses will need to manage costs through query optimization, training-inference separation, and smaller, more efficient models.
  • Optimal framing of AI will matter more: Advances in RAG, knowledge graphs, and larger context windows will improve AI’s ability to provide tailored responses with the right context. Accuracy will depend on matching the right AI approach (graph, vector, relational) to the right data.
  • Conversational AI will drive data access: Traditional analytics tools reach only 25–30% of users, but in 2025 generative AI-powered chat interfaces will democratize insights, simplifying access to data and automation while reducing complexity.

Agents

  • Multi-agent architectures will emerge: Just as with cloud environments and AI models, multiple interoperable agent systems will coexist, specializing in tasks such as data integration, automation, and analytics. Human oversight remains essential for governance.
  • Process intelligence and automation will matter more: Poorly designed workflows hinder efficiency, even when automated. Process mining and analytics will optimize workflows, enabling agents to interact seamlessly while maintaining clarity and structure.
  • Real-time data will always be essential: Agents will need up-to-date information to make accurate decisions. Advances in real-time data ingestion and hybrid processing will  be reshaping system architectures.
  • Applications will be redefined: Agents may replace or enhance traditional applications by fetching answers directly. Organizations will need a mix of buying, building, and adapting intelligent applications that evolve with user needs.