AI Is Rising on CEO Agendas | Presented by BCG

Summary

The discussion centers on how CEOs can move AI from experimentation to measurable business impact. The speakers argue that adoption is advancing, but many companies are struggling with the organizational changes and investment choices needed to scale it.

AI adoption and business impact

The CEO Radar reviewed earnings calls at 4,800 companies worldwide. Its analysis suggests a broad adoption gap:

• About 5% of CEOs are at the cutting edge, embedding AI in core operations, reshaping how their businesses work, or developing new business models.
• Another 30–40% are making progress.
• About 60% are struggling to get onto that path.

More than half of companies are still focused on pilots or general productivity tools, rather than investing sufficiently in AI that could create a distinctive competitive advantage. According to the speakers, the companies most advanced in AI are growing faster, though no growth rate was provided.

CEOs should communicate their AI progress to employees, prospective talent, customers, and partners. Talent wants to know whether they will have the tools and opportunities to succeed; customers and suppliers want confidence that a company will remain innovative and competitive.

AI investment and pilot-selection playbook

• Start with business value, not technical novelty. Prioritize workflows where AI can produce a clear, measurable improvement or create a differentiated customer offering.
• Select fewer initiatives and deliver them. The leaders described as most advanced focus on a smaller number of important use cases rather than pursuing many disconnected experiments.
• Continue promising pilots when the underlying opportunity is significant. One client ended a three-month pilot with lukewarm results, but the tools had improved substantially during that period. If the workflow has high potential value, reassess it with newer tools before abandoning it.
• Stop pilots that improve the wrong process. A pilot can make a process much faster without changing the overall business outcome, particularly if that process is not a bottleneck. Continuing such projects can waste money and create “zombie” initiatives.

A useful test is whether solving the problem improves a key outcome, such as winning business, serving customers, or materially changing cost or productivity. If not, even a compelling demo may be a distraction.

Scaling AI requires organizational change

Moving beyond pilots requires more than adopting technology. Companies must integrate AI into processes, leadership practices, skills, and job design.

Executive alignment alone is insufficient. For example, HR may need to revise job descriptions and create new roles to hire people who can connect AI agents and workflows. Leaders should also invest in the underlying data and technology infrastructure needed to scale use cases.

AI’s current value is often best realized in specific, end-to-end workflows with a human reviewing or acting on decisions, rather than fully autonomous chains of agents.

Agentic AI: current limitations

Companies are deploying agentic AI in bounded workflows, generally with a human involved. The speakers cautioned against assuming agents can reliably coordinate long sequences of tasks on their own. If each of 10 steps is only 80% accurate, the combined process can be below 50% accurate.

Reaching more autonomous, multi-agent processes will require better accuracy, data, and infrastructure.

Examples discussed

• Insurance underwriting: A leading insurer described rethinking underwriting to make it faster and more effective. In commercial underwriting, speed matters because getting a proposal to a client quickly can influence whether the company wins the business.
• Mining operations in Western Australia: A mining company uses predictive and generative AI to support autonomous operations across a large area. A control center staffed by about 10 people helps make decisions such as when to load different ships and how to run the broader process.
• L’Oréal Beauty Genius: L’Oréal discussed an AI-enabled beauty platform that offers recommendations and supports order fulfillment. The service had rolled out in the United States and other markets during the year discussed.

The examples were presented as operational business uses, not merely small-scale technology demonstrations.

Jobs and workforce implications

In the short term, the practitioners described AI initiatives as primarily focused on growth opportunities, rather than plans to reduce headcount. Efficiency can become more valuable when it is reinvested in new customer value propositions and business growth.

Some operations, such as autonomous mining, can displace labor compared with previous ways of working. The speakers expect job roles to change and workers to need reskilling. They were optimistic about net job creation over the longer term, while emphasizing that the new jobs will differ from today’s.

Employees who embrace AI may become more capable and relevant, but companies need to support the transition.

Leadership and operating environment

Resilience and adaptability were identified as essential CEO attributes amid repeated shocks. Effective leaders anticipate what they can, respond quickly, and adapt or reimagine the business when conditions change.

Strong CEOs also return to basic questions:

• What problem are we solving?
• For which customer and in which market?
• What value will be created?

Tariff uncertainty remained among the top one or two topics on Q3 earnings calls, although mentions declined from Q2. The discussion had shifted from “What will happen?” to how companies would respond.

Tariff disruption prompted companies to revisit how they make money, where they are relevant, and what value they deliver. It also pushed them to build stronger geopolitical and supply-chain capabilities, including visibility into risks deeper in suppliers’ networks.

The speakers argued that a decline in AI-related stock valuations would likely affect markets more directly than companies’ practical AI adoption. Continued corporate investment will depend more on demonstrated business results than on market valuations.

Metrics and timing mentioned

• 4,800 companies: Earnings calls reviewed by the CEO Radar.
• AI adoption: Approximately 5% leading, 30–40% progressing, and 60% struggling to get started or advance.
• More than half: Companies still focused on pilots or general productivity rather than scaling AI for competitive advantage.
• Agent accuracy: Current accuracy cited at 80–90% in some cases; accuracy compounds across multi-step workflows.
• Three months: Example of a pilot that was stopped despite a high-value opportunity and rapidly improving tools.
• Q2 to Q3: Tariff mentions declined, while discussion shifted toward how to navigate the issue.

Presenters and sources

• Presenters: Rich Lesser, BCG Global Chair; Vlad Lukić, BCG Global Leader of Tech and Digital Advantage; and Edward Adams, Bloomberg Media Studios host.
• Source: BCG’s CEO Radar, based on analysis of earnings calls at 4,800 companies worldwide.