[Photo: Shutterstock]

As the overheating around AI cools, demands are also growing for it to prove practical value. Against that backdrop, Babson College professor Thomas Davenport (토머스 다벤포트) and Fortune 1000 adviser Randy Bean said in a 2026 AI outlook contribution to MIT Sloan Management Review that this year will be a year to level-set expectations to reality. They also presented five things companies should focus on regarding AI in 2026.

First, agent AI is not yet ready for deployment in real-world operations.

They said, "Hallucination errors persist, and vulnerabilities to hacking such as prompt injection are slowing adoption. Companies will keep human oversight systems for agent AI."

But the two forecast that within 5 years AI agents will handle a significant share of large-scale business processes. They urged that, at the current stage, AI agents should start with use cases that can be reused across the organisation.

Second, the AI bubble could deflate. The two said, "A structure that prioritises user growth over profit resembles the dotcom bubble. Technology tends to be overvalued in the short term and its long-term transformative impact undervalued. Companies should make the most of the AI technologies they already have while checking investment effectiveness."

Third, generative AI must shift from a personal tool to a corporate resource. The two said AI at most companies is currently used only as an individual employee productivity tool. "If it is not applied to company-level workflows such as new product development and improving customer experience, it is difficult to aggregate results and measure value," they said.

Fourth, AI organisational reporting lines are unclear.

The two noted that in a 2026 AI and data leadership executive benchmark survey, 38 percent of respondents said they had created a chief AI officer, or equivalent, position, but reporting lines are dispersed across business, technology and innovation leadership. "This dispersed structure leads to weak AI value creation," they said.

They pointed to JPMorgan as a company worth referencing on this. JPMorgan has a structure in which an AI executive is included on a 14-member operating committee and reports directly to the chief executive officer.

Fifth is building an "AI factory". An AI factory refers to an organisational capability to build AI systems quickly and efficiently by combining technology platforms, methodologies, data and existing algorithms. The two stressed, "It is not about filling data centres with GPUs, but internal organisational capability."

Keyword

#MIT Sloan Management Review #Babson College #Fortune 1000 #JPMorgan #CAIO
Copyright © DigitalToday. All rights reserved. Unauthorized reproduction and redistribution are prohibited.