aregfly - stock.adobe.com
Forrester Technology & Innovation Summit: Prepare for AI failures
The AI bandwagon is about to be derailed, and CIOs are going to have to pick up the pieces
IT leaders are set to get a boost in budget in 2026, but according to research from analyst Forrester, they are likely to feel greater pressure to deliver value and fix problems, particularly with artificial intelligence (AI) initiatives.
The analyst’s research suggests that enterprises will push back a quarter of their AI spending into 2027 due to a failure of many projects to deliver value. It predicted that in 2026, CEOs will pull more chief financial officers (CFOs) into AI deals.
Forrester predicted that fragmentation among AI providers will cause the majority of enterprises to build composable agent architectures, which will be needed to manage and orchestrate fractured AI agent deployments and enable complex multi-agent use cases.
“Companies will distribute their bets across agentic ecosystems and shift talent around as AI agents take over grunt work,” said Sudha Maheshwari, vice-president and research director at Forrester. “Savvy enterprises will invest in AI governance and AI fluency training to mitigate risk and slowly chart their AI voyage.”
With the increase in AI project failures, Forrester expects a quarter of CIOs will be asked to bail out business-led AI failures in their organisation. In a blog post, Forrester vice-president and research director Mark Moccia predicted that CEOs will turn to their technology chiefs to fix failed AI projects.
Forrester also reported that 39% of AI decision-makers say their CIO or chief technology officer leads AI technology strategy, and 21% lead AI business strategy. Moccia said these numbers are set to double as organisations realise that tech leaders are best positioned to marshal the teams needed for successful AI agents. He recommended that CIOs establish governance, curate data and knowledge assets, design user experiences, and manage output quality.
“CEOs cannot wait for a high-profile AI ethics or policy failure to force action,” said Moccia. “Tech leaders must strengthen governance and scenario planning now to avoid costly mistakes.”
Read more about AI project failures
- Why AI agent projects are stalling in production: A Confluent technology strategist explains why multi-agent workflows become brittle monoliths and how access to real-time, streaming data is key to the success of agentic AI deployments.
- Key hurdles in AI from proof of concept to production: We talk to DDN chief technology officer Sven Oehme about key technical and organisational hurdles when taking AI projects from test to production, in storage and building the technical team.
Forrester predicted that two-thirds of CIOs will need to justify budgets by linking tech spend to business value. While CIOs are driving business value with their tech strategy, Moccia noted that communicating this to CEOs and business leaders is often lost in translation.
He urged IT leaders to map tech services to the business capabilities that underpin strategy. This is hard and Forrester’s research shows that only about a third of enterprises have achieved this today.
Moccia warned that tech leaders will need to focus on value because technology spend is growing faster than inflation due to AI, cloud, and security. “CIOs will need to adopt IT finance frameworks such as technology business management, to map total tech spend to business capabilities, and practices like FinOps, to attribute variable costs such as cloud,” he said.
Moccia recommended that CIOs learn to speak fluent “business value”, since a CFO will want more than smoke and mirrors, and keep a governance toolkit ready to handle AI rescue plans when projects go wrong.
Forrester believes a new IT workforce is emerging, based on a mix of humans, AI bots and gig workers. Its research predicts that a third of CIOs will adopt gig worker protocols and agents to support multi-job IT. He said that managing this workforce mix will require an increased focus on IT teams’ leadership skills.
