I believe innovation teams, CIOs, CTOs, and CDOs must have a short-term value focus on their 2025 priorities. “Smarter efficiencies that impact customers” would be my recommended headline for your 2025 vision statements.

We discussed 2025 digital transformation priorities at a recent Coffee With StarCIO Digital Trailblazers. I shared research and my perspective on what every innovative CIO must focus on in their 2025 Budget beyond Just AI and covered one and only one critical leadership priority. Read the post for this recommendation.

In this post, I focus on innovation teams and what to focus on in AI that delivers “Smarter efficiencies that impact customers.”

“Develop a tighter, shorter, cycle around strategy because of rapid economic changes over the next 6-12 months,” recommends Liz Martinez, MD, at PMO Whisperer. “Leaders should review the tactics in their innovation labs every few months and realign priorities.”

Below are five AI priorities that innovative DevOps, data science, and product development teams should consider in 2025.

Gen AI Agents: From SaaS to productization

Appian, Atlassian, Cisco Collaboration, Forethought, IBM, Ivanti, Pega, Salesforce, SAP, ServiceNow, Workday, Zoho, and others launched service-oriented AI agents in 2024.

Going beyond AI agents in SaaS,  you’ll see many more companies experiment with developing AI agents from their proprietary workflows in 2025. Companies that were successful in developing data products and commercial AI services will consider productizing these AI agents.

“CIOs should be looking to use emerging AI technologies in pursuit of putting the most pertinent, contextually-aware information in the hands of users before they know they need it,” says Simon Margolis,  associate CTO of AI/ML at SADA. Smart AI agents that go beyond a chat experience and will provide insights the way humans do in existing systems like CRM, email, or other platforms. These generative agents will consider all of the information that a human agent would, considering practical knowledge as well as data from various platforms in pursuit of providing meaningful and actionable insights.”

Public cloud providers announced tools for developing AI agents in 2025:

• AWS Bedrock Agents enable generative AI applications to execute multistep tasks across company systems and data sources
• Google’s Vertex AI Agent Builder to build and deploy enterprise-ready generative AI experiences
• Microsoft Copilot Studio has Copilot Agents and announced the ability to create autonomous agents in public preview
• Salesforce Agent Builder to create and customize agents with an AI-powered, low-code builder

AI agents in SaaS aim to improve productivity and employee experiences. Going from buying to building AI agents requires innovation teams to consider customer and strategic outcomes. Developing AI agents will appeal to customers with proprietary customer or employee-facing applications where AI-enabled workflows will be innovative and potentially differentiating capabilities.

Innovation teams should also seek examples of where AI agents provide value in IT operations and infrastructure.

Gary Sidhu, SVP of product engineering and technology at GTT, says CIOs should invest in observability technologies to enhance insights into their organization’s network. “With an observability agent, organizations gain real-time data and actionable insights, enhancing control, engagement, and visibility into their network. Organizations can be more proactive and agile, quickly identifying anomalies, reducing downtime, and improving user experience – positioning themselves for greater operational efficiency in today’s complex digital landscape.”

AI in marketing: Demonstrate value at scale

The 2024 marketing landscape supergraphic featured over 14,000 logos. Their report shows that the average medium-sized business (500-2500 employees) uses approximately 250 apps, while large enterprises average 650 apps. That’s a lot of technology, data, and integration for innovation teams to consider simplifying, integrating, and optimizing.

The report shows that businesses took steps to consolidate the number of apps from 2023 to 2024, and I suspect innovation teams will find other simplification approaches in 2025.

“CIOs must maximize information on user interface utility to serve their profit centers better, says Paul Boynton, COO and co-founder of CSI. “Corporate marketing may need data on competitors, customers, and point of sale contacts, and a single platform interface that provides efficient, cost-effective access to this data will eliminate the need to source information from multiple platforms.”

Innovation teams can partner with marketing departments by focusing on data and AI governance.

I’ve been writing a lot about data governance, AI, and data science, and below are ten of my recent posts that I published on Drive and InfoWorld over the last six months.

• How to support accurate revenue forecasting with data science and dataops
• 6 important AI and data governance non-negotiables
• 5 ways data scientists can prepare now for genAI transformation
• What you need to know about AI governance
• The definitive guide to data pipelines
• 3 important genAI insights for every innovative digital transformation leader
• 7 reasons analytics and ML fail to meet business objectives
• 7 steps to improve analytics for data-driven organizations
• How to choose a data analytics platform
• 12 helpful gen AI strategies for winning with data science

“In 2025, data will lead the way,” concluded Joe Puglisi, fractional CIO at 10Xnewo, during the Coffee Hour.

Knowledge management: AI’s killer app in skills-based industries

Data is a subset of information, which is a subset of knowledge. LLMs and RAGS need knowledge repositories to fuel accurate answers and power AI agents. Knowledge management is critical to retaining and training new talent in knowledge-heavy industries like construction and manufacturing.

“We will see a big change around the notion of tech estates and debt,” says Joanne Friedman, PhD and CEO of Connektedminds. “Tech debt is going to be recognized as not necessarily being debt in terms of systems, but the data and the knowledge that was created in those somewhat siloed systems needs to be brought back into what is truly a knowledge environment where AI can use that data to help inform better decision-making.”

So, while DevOps focuses on tech debt and data governance on data debt, Joanne and I both believe that 2025’s focus should be on knowledge debt. Innovation teams should begin to address knowledge debt by converting document repositories into wikis, knowledge stores, and vector databases that simplify usage and can fuel gen AI capabilities.

Gen AI in Devices: From voice to vision

Will gen AI and LLMs be limited to typing in prompts? Or will virtual agents like Siri and Alexa integrate with LLMs and usher in a new, voice-enabled gen AI experience? Will B2C voice-enabled gen AI consumer experiences drive voice-powered applications in the enterprise?

I recently spoke with  Shaown Nandi, director of technology at AWS, about when we’ll see LLMs integrated with voice assistants. “Very close, very close,” was his response. He then acknowledged scale, cost, and understanding dialects as some of the hurdles bringing this capability to production.

Voice is one emerging form factor for LLM integration.  For example, Webex AI Agent is voice-driven, listens to customer issues, and provides natural-voice responses to their problems and questions. Integrating voice with LLMs, upgrading to high-fidelity video conferencing, and adding smarter call center technologies are three ways to drive amazing employee experiences with gen AI.

Looking beyond voice, video, and computer vision at the edge is a second innovation that teams should consider researching in 2025.

 “CIOs should strongly consider vision AI at the edge for valuable business insights at a fraction of the time, money, and resources to deploy at a scale for true transformation,” says Eita Yanagisawa, Sr GM, AITRIOS at Sony Semiconductor Solutions. “Plus, embedded vision AI processing eliminates the need for an excess of video data, reducing overall data footprints and associated cloud costs.”

While voice and video may not be production use cases for many innovation teams, they should be on the research radar and possibly a POC priority, depending on how fast tech companies release new capabilities in 2025.

Green FinOps: Deliver savings ahead of new consumption

What will pay for all the emerging and practical innovations in 2025? More IT departments must look for ways to trim costs and find efficiencies. Cloud efficiencies and FinOps should be a priority for many organizations with expensive public cloud costs that want to find easy savings.

“Costs are under a microscope, and FinOps is in, says John Patrick Luethe, a regular contributor to the StarCIO Coffee Hour. “Companies are investing in FinOps tools to decrease the costs to serve and looking at where to host their workloads between public clouds and private clouds in their data centers.”

FinOps is also a sustainability program, as many platforms report carbon consumption and recommend reduction opportunities.

“CIOs must invest in green data center technology in 2025,” says Anant Adya, EVP at Infosys Cobalt. “Globally, data centers account for about 3% of the total energy use, equivalent to about 200 million tons of carbon emissions, which is expected to grow by over 30% every year. As CIOs strive towards energy sustainability, it is critical to consider greener data centers.”

The challenge is that AI is expensive in terms of cost, computing, and power. Innovation teams that create full-cost transparencies around the AI opportunities they pursue help bring the conversation around value to the forefront.

“AI is going to continue to be a priority, but hopefully, it’s going to be more realism rather than hype,” says Martin Davis, managing partner at DUNELM Associates.

Innovation teams will be under a lot of pressure in 2025 to do more than separate realism from hype. Delivering value will be an important priority, and top innovation teams will aim for short-term value, mid-term strategies, and longer-term visions.