I was in San Francisco last week to meet with leaders from Leadtail, Proofpoint, Tray.ai, Logitech, Observe.AI, and Acceldata to discuss what they can expect from CIOs in 2025. I shared insight from previous posts, such as three tactics for supporting innovation teams and my top investment priority for 2025 beyond gen AI.

Of course, our discussions revolved around gen AI, LLMs, and agentic AI: their hype, possibilities, and realities. What are my predictions? I recently shared these gen AI technologies that innovation teams should review, and I have an article coming soon on CIO.com on digital transformation ins and outs.

In short, I believe CIOs, Digital Trailblazers, and innovation teams will be much more pragmatic in 2025 by seeking gen AI use cases that deliver short-term value and more seismic longer-term impacts, which is why agentic AI holds so much promise.

But for this post, I captured 20 predictions from experts on gen AI, with some really important insights and advice from them. I organized them into five areas:

• Organizations will focus on pragmatic AI as LLM hype fades
• Agentic AI will improve today’s operations and decision-making
• Security, data privacy, and regulations will drive AI and data architectures
• CIOs will go back to AI basics around data and IT operations
• Winning AI organizations will drive change management

Below are the 20 predictions.

Organizations will focus on pragmatic AI as LLM hype fades

1.     AI moonshots dwindle, and AI POCs have weeks to prove value

 “Brands need to stop spending millions on AI experiments that never make it out of the lab. Next year, brands will demand to see CX automation outcomes before committing any investment to AI,” says Dave Singer, global VP of go-to-market strategy at Verint. “They’ll seek out and invest in solutions that deliver real results in weeks, not months or years.”

2.     LLM’s potential will max out and miss on AGI’s foundations

 “Despite the excitement around AI, many will begin to realize the limitations of large language models,” says Jeremy Burton, CEO  at Observe.inc. “While impressive at summarizing, translating, and regurgitating well-known information, these models are clearly not the foundation for artificial general intelligence (AGI). LLMs have arguably siphoned investment dollars away from approaches that may have had a greater chance of success. It’s not that LLMs are useless; it is quite the opposite: they are a better way for humans to interact with all kinds of software, devices, and systems, and they will fundamentally change many industries. But the AI super-intelligence, which interacts and reasons about its knowledge, surroundings, and people in the same way humans, is still many years away.”

3.     Domain and industry models will rise in importance

“In the rapidly evolving AI landscape, a significant paradigm shift is taking place: the rise of domain-specific LLMs,” says May Habib, CEO of Writer. “AI-Native applications are already touching every corner of the enterprise, but domain-specific models are setting the stage for the next wave of AI innovation because they offer unparalleled precision within their designated fields.”

4.     Small language models will reduce costs and improve accuracy

“In the new year, there will be a greater focus on training smaller language models due to their cost-effectiveness, accessibility, and ease of deployment,” says Uday Kamath, chief analytics officer at Smarsh. “Domain-specific language models specialize in particular industries or fields, like the medical or legal field. These models can be finely tuned with domain-specific data and terminology for these complex and often heavily regulated industries where precision is necessary. This targeted approach reduces the likelihood of errors and hallucinations that general-purpose models may produce when faced with specialized content.”

5.     AI agents require precision and will augment humans, not replace

 “In 2025, generative AI adoption will largely be driven by AI agents that assist humans with tasks, and AI replacing humans is a long way from happening,” says Arnab Mishra, CEO of Xactly. “The reality for enterprises is that most business workflows require high degrees of precision and today, generative AI tools are not sufficiently precise. For organizations, even a small difference in accuracy can result in significant business consequences.”

6.     Top AI agent vendors will support outcome-based pricing models

“As more businesses begin to test and rely on AI agents to support their business–like using them to handle customer service questions–software makers that supply the agents will get paid based on successful outcomes, not just a monthly or usage-based fee for their software,” says Scott Woody, CEO of Metronome. “This change will continue the upending of pricing for many fast-growing software companies, as their customers demand the ability to pay for successful outcomes only and not broad, perhaps ineffective, AI solutions.”

Agentic AI will improve today’s operations and decision-making

7.     Gen AI graduates from content generation to decision-making

Ravi Ithal, GVP and CTO of Proofpoint DSPM, Normalyze, says IDC predicts that by 2025, 30% of major brands will be generating at least 50% of their ad copy using GenAI, but the real power will be in AI-driven business decisions, not just content. “Generative AI will move beyond content generation to become the decision-making engine behind countless business processes in everything from HR to marketing.”

8.     Agentic AI will have quick wins in support and operations

“Agents will revolutionize customer experience through hyper-personalized interactions and predictive support while reducing operational risks via real-time threat detection and response,” says Karthik SJ, GM of AI at LogicMonitor. “Organizations implementing these technologies can expect significant improvements in incident response times and customer satisfaction metrics.”

9.     Improve CX with AI and human collaboration

 “The future of customer experience will be defined by how seamlessly AI and human teams can work together,” says Assaf Melochna, president and co-founder of Aquant. “We can scale support operations without sacrificing quality or personalization by simplifying multi-agent AI systems. Streamlined collaboration enables faster, more effective responding to customer needs, especially during high-demand periods, ultimately enhancing the overall customer experience.”

10.            Marketers will increase AI-enabled personalization but maintain privacy

“In 2025, innovation in customer experience will hinge on striking the right balance between personalization and privacy,” says Jacqueline Woods, CMO of Teradata. “Marketers will increasingly rely on first-party data—information customers have willingly shared and trust will be used responsibly—to deliver tailored experiences that feel helpful rather than intrusive. This shift will foster deeper customer trust and enable businesses to create personalized offers that resonate authentically. As AI evolves, the focus will be on leveraging trusted data to surprise and delight customers with relevant interactions, redefining customer loyalty and satisfaction in the digital age.”

Security, data privacy, and regulations will drive AI and data architectures

11. Software architecture complexity will challenge security posture control

“With AI and code generation becoming core to software development, we’re on the verge of unprecedented architectural complexity that will make traditional security posture control nearly impossible,” says Idan Plotnik, co-founder & CEO of Apiiro. “By 2025, new forms of malware and open-source codebase vulnerabilities will emerge, and attackers will leverage AI to craft advanced, evasive malware.”

12. Private LLMs will increase to support data privacy and AI guardrails

 “Generative AI’s honeymoon period will end in 2025 as CIOs grapple with increasing data privacy regulations and realize the risks of exposing proprietary enterprise data to LLMs,” says Claus Jepsen, chief product and technology officer at Unit4. “Moving past the hype of gen AI, tech leaders will go back to the basics by leveraging automation as a lower-stakes technology to improve user experience. I also expect to see more CIOs implementing guardrails such as AI governance boards to ensure compliance, ethics, and transparency as enterprises dabble with emerging technologies.”

13. AI security and data privacy will drive private cloud adoption

“As generative AI adoption matures, organizations will shift from initial excitement to a focused investment in secure, results-driven solutions, with data privacy and security taking center stage,” says Akhilesh Agarwal, COO of APEX Analytix. “This evolution will drive a shift away from public cloud reliance toward on-premises or private cloud deployments to protect sensitive data better and strengthen compliance. As a result, companies will consolidate investments in tailored, high-value AI solutions, favoring infrastructure choices that align with strict data governance, security requirements, and operational standards, creating a more reliable and mature AI landscape.”

14. CDOs will shift focus to increasing data security on unstructured data

 “Companies will need to have a full view of their unstructured data flows, authorizations, and permissions, as well as be able to filter out sensitive information like personal or confidential content,” says Jack Berkowitz, Chief Data Officer at Securiti. “They also need to be able to field context-sensitive controls or firewalls that prevent someone from circumventing or attacking the purpose of the GenAI tool.

15. CISOs will look to consolidate AI security platforms

“When looking at ways to obtain fast value for lower risk and better security, companies must cut quick, aggressive solutions that may meet immediate needs but, if not aligned with a broader security strategy, can create fragmented, unscalable solutions,” says Joe Crawford, head of the global technology office at Glassbox.  “Implementing quick solutions without a cohesive plan can create complex security layers that complicate management and increase risks of human error and misconfiguration.”

CIOs will go back to AI basics around data and IT operations

16. API-First and GenAI integrate analytics into every app

“In 2025, traditional BI tools will become obsolete, as API-first architectures and GenAI seamlessly embed real-time analytics into every application,” says Ariel Katz, CEO of Sisense. “Data insights will flow directly into CRMs, productivity platforms, and customer tools, empowering employees at all levels to make data-driven decisions instantly without technical expertise. Companies that embrace this shift will unlock unprecedented productivity and customer experiences, leaving static dashboards and siloed systems in the dust.”

17. Organizations will need AIOps well before autonomous AI agents

 “In 2025, we’ll see the rise of dedicated AIOps teams to manage AI operations, while the appeal of agentic AI won’t pan out as fast as some people are predicting,” says Eoin Hinchy, CEO of Tines. “AIOps teams will handle everything from model deployment to managing quotas and security for AI-driven workflows. However, the idea of fully autonomous agentic AI — AI that runs without human oversight—will remain a distant reality because real-world complexity will likely prevent us from going fully autonomous in the near future.”

18. Data engineers are key to private LLM accuracy

“As enterprises have started experimenting with AI, they’ve quickly realized that an AI system is only as valuable as the data feeding into it,” says Jeff Hollan, head of applications and developer platform at Snowflake. As a result, language models need proprietary data in order to deliver insights that actually matter for businesses — and data engineers are quickly becoming central to this effort. Data engineering teams will be expected to either repurpose their existing tooling to work for AI-centric data pipelines and workflows or to learn and implement new tooling specifically designed for this work.”

Winning AI organizations will drive change management

19. Greater focus on employee adoption and change management

“The reality is that the future of the workplace will continue to be heavily impacted by AI, which we have already seen in 2024,” says Rajeshwari Ganesan, distinguished technologist at Infosys. “It is paramount that leaders offer employees ways to learn and engage with AI tools positively. Providing collective experiences like AI training modules will equip employees with experiences needed to succeed in the rapidly changing AI landscape, upskill employees internally, and improve efficiency within the company.”

20. CIOs will invest more on learning programs for innovation teams

 “CIOs should first look at their teams and address cultural factors,” says Ronda Cilsick, CIO at Deltek. “While AI adoption will be important in 2025, these solutions cannot succeed without the people who implement and provide education and training to enable their use. By fostering a culture of innovation and involving employees in the adoption process, CIOs can secure buy-in and ensure that new technologies are effectively integrated to meet strategic goals.”

–

I selected these predictions because they offer reasonable advice to Digital Trailblazers, and 2025 should be an interesting year for ambitious leaders driving pragmatic AI capabilities.