5 AI trends CIOs should watch in 2027

In 2027, CIOs are going to be looking for bigger gains from AI—and getting more selective about which projects they fund.

Interviews with technology leaders point to five priorities taking shape for the next year. CIOs plan to concentrate investment on a handful of business processes, as their competitors push them to rethink customer experience (CX). Other priorities include matching AI models to specific tasks, bringing finance and business leaders further into spending decisions, and building networks of AI champions to help employees change how they work.

“The job is to use AI to reimagine the work, not just automate the old way of doing it,” said Gill Haus, CIO for consumer and community banking at JPMorganChase.

For some companies, that redesign will involve giving autonomous AI agents responsibility for several steps in a workflow. For others, that might be a reach, given that 42% of finance leaders in Deloitte’s “Finance Trends 2027” report indicated that their ambitions for agentic AI exceeded their organizations’ current capabilities.

That readiness gap could shape the pace and scope of expansion, giving CIOs another reason to limit their AI projects to a few workflows they can support.

1. CIOs move from many AI pilots to a few big bets

Several CIOs are heading into 2027 with plans to concentrate their AI investments on a handful of business processes. After experimenting across individual tasks, they’re looking for larger gains by redesigning how work gets done from start to finish.

“[We’re] thinking about what are the five to 10 areas that we really have to focus on and think about end-to-end to drive value," said Mojgan Lefebvre, chief technology and operations officer at Travelers. "It’s so easy to get mesmerized by what I call ‘demo love.’”

Travelers’ experience with new-business submissions shows a potential payoff. Its Bond & Specialty business redesigned how it receives submissions, cutting the time to register them from two hours to two minutes, Lefebvre said.

Choosing where to pursue those larger improvements will require companies to narrow the field of experiments they support. Vrinda Menon, CTO for transformative AI at JPMorganChase’s private bank, urged that focus during a panel with Lefebvre at the GAI World 2026 conference in September.

“There are thousands of pilots, thousands of experiments, a million flowers blooming everywhere,” Menon said. "I think it’s an opportunity now to double down on two or three big bets for every organization."

For some CIOs, those bigger bets will involve giving AI agents some autonomy. But expanding AI's role will also require companies to build the systems to oversee it;  only 24% of respondents to Infotech’s 2027 Global IT Leadership Survey said they already had the architecture in place to monitor agentic workflows, according to Brian Jackson, research director at Infotech.

At Samsara, a fleet-management technology provider, CIO Stephen Franchetti plans to expand the use of agents across a small set of high-value workflows in 2027. A shared platform gives teams across the company, including those outside IT, a common place to build and run agents. Each business function is responsible for its outcomes and has a quantified roadmap.

That approach leaves room to discover new uses for AI. Franchetti said Samsara will continue running low-cost experiments to identify projects worth expanding while directing larger investments toward the workflows with the greatest potential impact.

Agentic AI expansion will also shape staffing plans at First Student, a school bus operator. Chief information and AI officer Sean McCormack said the company has seen a fivefold increase in software development throughput and an 80% reduction in cost since adopting AI agents for development. He plans to build on those gains in 2027, relying less on outsourcing and more on a new internal team to oversee agents throughout development.

“If 2026 was about automating different tasks and augmenting staff with generative AI," Jackson said, "2027 will be about orchestrating workflows and automating processes with agentic AI.”

2. AI competition pushes companies to rethink CX

As established companies face the prospect of new rivals using AI to deliver services faster or at a lower price, competitive pressures will likely push CX higher on the AI agenda for 2027.

AI-native startups are reaching $1 billion valuations in about three and a half years with roughly half the staff startups once needed, said Manish Jain, founder of Strategic Horizon Research.

At OneDigital, an insurance and financial services firm, co-founder and chief growth officer Mike Sullivan is preparing for that competition to reach the company’s existing clients.

“We’re approaching life these days like AI represents a systemic risk for our business, and someone’s going to come knocking at every single one of our clients’ doors saying, ‘We’ll do what they do, better, and for half the price,’” Sullivan said.

Price is only one way a rival can win customers. At GAI World, an attendee cited neobanks before asking Travelers' Lefebvre how much she worries about new entrants in insurance. She acknowledged that digital insurers could offer strong digital experiences. Such competition gives Travelers another reason to keep improving its technology.

“We appreciate the competition because it keeps us on our toes and helps us realize that we'd better be as good a technology company as we are an insurance company,” Lefebvre said.

For CIOs planning their response in 2027, improving CX will also mean examining the work behind it. A startup can design a service around AI from the beginning; an established company might have complex procedures built around approvals, handoffs and systems that can slow delivery -- even after AI speeds up individual tasks.

Menon described how her team at JPMorganChase is redesigning workflows in business areas such as lending. The team examines what employees do at each step and which approvals they need, then determines which steps can be eliminated.

“In a bank like ours, some of those processes are from the 1990s, from the early 2000s,” Menon said.

Her point was that applying today’s technology to those processes without rethinking the underlying work misses the opportunity. Her team is examining how the process should work from start to finish, including whether some tasks are still necessary.

One workflow had required seven people to navigate 40 screens across seven systems, Menon said. After the bank redesigned and automated the process, the work went from taking seven days to a few minutes.

For CIOs deciding which workflows to redesign, one consideration is where customers would see an improvement. Forrester’s “Accelerate Your AI Voyage” report, published in April 2026 and based on a survey of 1,500 AI decision-makers, found that 52% of high AI adopters focused on CX, compared with 44% of low adopters.

One opportunity is to rethink the moments when customers first seek help. At Travelers, that included the initial call to report a loss. Lefebvre said more than 85,000 claim calls were eligible to go through its generative AI assistant. The assistant changes how customers interact with the insurer at the start of a claim, making it an example of AI reaching directly into service delivery.

Lefebvre said, over the next year, organizations should focus on improving the capabilities they control, while also taking the threat of new competitors seriously.

“If you’re a business leader, you have to always worry about competition, and you should never be arrogant enough to say you cannot be displaced,” Lefebvre said.

3. CIOs get more selective about AI models

As CIOs bring AI agents into more business workflows in 2027, large language model selection will become a more frequent decision within those processes. Leaders must decide which steps need a frontier model and where a narrower model tuned for a specific task can do the work. Differences in accuracy, speed and cost can add up when a workflow runs thousands of times or more.

At JPMorganChase, Menon pointed to tasks such as classifying information or figuring out the nature of a customer request. A smaller model can be tuned to recognize the categories and the language a business uses for those tasks.

“It’s become almost an imperative that we fine-tune models," Menon said. "We don’t need frontier intelligence for all products.”

Beyond fine-tuning, Menon said agents need access to the know-how employees build up over years of doing the work. Companies have to make that expertise available to agents as they're working on a particular task. For example, an agent working on a lending request needs access to relevant lending knowledge and practices.

Travelers’ experience shows how company-specific expertise can improve a model’s performance. The insurer built TravelersLLM, a proprietary insurance model, by fine-tuning an open-weight model. Travelers trained it on company documents and has used it in underwriting.

Lefebvre said data scientists and subject matter experts helped run thousands of tests. “The responses when it came to Travelers-specific tasks and questions are more accurate, ... 90% cheaper and five times faster,” she said.

Travelers still uses frontier models for work that calls for broader reasoning. TravelersLLM gives the insurer another option for tasks that depend more on its own data and insurance expertise. As AI expands, such flexibility lets the company choose a model based on the demands of each task.

CIOs are also planning for those choices to change as models improve. At First Student, McCormack wants the company’s AI tools to work with whichever model best fits the task over time. Model capabilities can shift from week to week, he said, and switching models should be possible without rebuilding the tools that use them.

“Rather than build everything around one model, we’re developing some of our own AI capabilities so we can switch between models as needed,” McCormack said.

4. Finance and business leaders gain influence over AI spending

In 2027, decisions about which AI projects to fund and what they must deliver will increasingly be shared among CIOs, finance and business leaders. Much of that spending already falls outside IT. Respondents to Boston Consulting Group’s 2026 Applied AI Index indicated that 80% of AI spending sits outside the enterprise IT budget.

Finance leaders are helping decide where that money goes. In Deloitte’s “Finance Trends 2027” report, 54% of surveyed finance leaders said they lead decisions about where to invest in enterprise AI and technology. Decision-making has evolved into a partnership among finance, IT and the business leaders whose teams will use the technology, said Ed Hardy, Deloitte’s U.S. finance services leader.

That partnership becomes more consequential as companies expand AI beyond pilots. A consumption-priced tool can look affordable in a trial but then can cost much more as it's put into production. Finance and IT teams need to estimate what a workflow will cost at scale, while business teams must define what improvement would justify that expense.

Defining the expected benefit also raises the question of who's responsible for delivering it. Sedgwick, a claims management company, is putting greater weight on accountability when evaluating AI investments.

“We’re placing greater emphasis on understanding who owns the expected business value and how we’ll measure the outcome before committing resources,” said Sean Safieh, CIO of global platforms and digital solutions at Sedgwick.

That scrutiny also needs to continue after a workflow goes live. Tracking the cost of completing work can help leaders judge whether expanding AI use makes financial sense. For example, for a customer support workflow that could mean measuring the cost of resolving a request alongside the quality of the service.

At Samsara, Franchetti said the company is shifting its AI budgeting as it decides which workflows to expand in 2027. “We’re investing in the shared platform and a focused set of high-value workflows," he said, "shifting from budgeting for licenses to tracking cost per unit of work so money follows results.”

5. Companies formalize networks of AI champions

Nearly four years after ChatGPT’s public debut, IT leaders are still trying to get more employees to use AI regularly in their jobs. To make AI part of their workday, employees often have to change routines they’ve relied on for years.

“The biggest obstacle is people and process, not technology, because people don’t change how they work by mandate,” Franchetti said.

As companies expand their use of AI, some IT leaders are making peer learning a more formal part of rollouts. The U.S. Department of the Navy wants to connect experienced AI users so that what one team learns can help others, said Justin Fanelli, Navy CTO.

“The secret goal is to have a champions network where AI superusers—across different groups, departments, teams—have a place to share notes, [and] collaborate with each other,” Fanelli said on the GAI World panel with Lefebvre and Menon.

Those exchanges can help employees work through practical questions about using AI. For example, a team that has tried it on a recurring task can share where it saved time, where the output still needed a person’s review and what colleagues should consider before trying a similar approach.

Samsara is also organizing a peer network across its workforce. Its champions help identify where AI fits within different business functions. It gives employees a role in shaping AI adoption as executives push its expansion.

“We pair top-down executive commitment with a bottom-up AI champions network drawn from every function, since the people closest to the work know best where AI fits,” Franchetti said.

For CIOs, the challenge will be to get that knowledge beyond the employees who are already eager to experiment. Champions can help colleagues see how AI applies to their own jobs and work through the difficulties that keep them from using it regularly.

Key takeaways

In 2027, CIOs are set to concentrate AI investment on fewer, larger bets while rethinking the processes that shape CX. They’re becoming more selective about which models handle the work, sharing spending decisions with finance and business leaders, and organizing AI champions to help employees adopt the tools. Together, these changes put more emphasis on making AI work consistently across the business as AI pilots move into production.

CIOs unsure of where to begin can choose one business process where AI could make a measurable difference. Business leaders can define the improvements customers or employees should see, IT can assess the models and systems needed, and finance can help determine whether the costs justify expansion. Involving employees early can also reveal what support they’ll need to make the new process part of their everyday work.

Tim Murphy is a reporter covering IT strategy for TechTarget, with a focus on IT leadership, governance, workforce skills and AI strategy. His enterprise technology coverage has earned multiple Azbee Awards.