CEOs have funded pilots, hired talent, and announced roadmaps. BCG’s 2024 research asks a simpler question: where is the value? The answer is concentrated. Only a small share of companies have moved from experimentation into systems that change how the business runs.
Four maturity buckets
BCG groups companies into four categories: doing little (25%), still focused on proofs of concept (49%), scaling some value (22%), and operating value engines that create substantial value (4%). The first two buckets cover three-quarters of the sample. Pilots are common. Production value is not.
The 26% that have moved beyond PoC combine the scaling and value-engine groups. That is the practical bar for “we are getting something out of AI,” and it is still a minority.
Leaders put AI on the core work
Leaders are not defined by sector or by being digital natives. More than half in the top two categories are incumbents. What separates them is focus: they deploy AI against core business processes and support functions, not only side projects.
BCG reports that leaders generate 62% of AI value from core processes. They also set large targets, concentrate investment on a few high-priority opportunities, and put people and process change ahead of model shopping.
Ambition without a production path
Many organizations treat a successful demo as progress. BCG’s distribution shows why that is a weak signal. A PoC that never owns a recurring workflow, an owner, and an acceptance metric stays in the middle buckets indefinitely.
Leaders expect meaningfully higher cost reduction and revenue growth from AI than other firms, and they treat RoI as something to measure in the operating plan—not in a slide deck after the pilot.
What to change first
BCG’s playbook starts with a top-level commitment, a short list of high-value initiatives, and the minimum infrastructure those initiatives need. It then baselines capability gaps and keeps governance on people and process over algorithms.
For teams deciding where to spend the next quarter, the research points to a blunt filter: if the work is not a core process with a named owner and a path past PoC, it is unlikely to move the company out of the stalled majority.