Businesses are increasingly generating financial returns from artificial intelligence projects but remain unable to expand many of those experiments across their organisations, exposing a widening gap between AI pilots and full-scale enterprise adoption.
A study by management and technology consultancy BearingPoint found that almost three-quarters of companies surveyed reported positive financial returns from AI initiatives.
Yet only 13 per cent were considered to be successfully on track with their overall artificial intelligence strategies.
Less than one-third of organisations have been able to effectively scale their AI projects.
Successful Experiments Don’t Guarantee Transformation
The findings highlight one of the technology industry’s emerging challenges.
Creating an AI demonstration or introducing a chatbot into one department can be relatively straightforward.
Transforming an entire company around AI is considerably harder.
Businesses frequently operate decades-old software, fragmented databases and specialised internal systems that were never designed to interact with modern generative AI.
Integrating new models into that infrastructure can require significant spending on cloud computing, cybersecurity, data preparation and software engineering.
Regulatory requirements create another obstacle, particularly for companies operating in highly regulated sectors or across multiple jurisdictions.
BearingPoint identified regulation and integration with legacy information-technology systems among the main factors limiting broader adoption.
AI Economics Come Under Greater Scrutiny
The findings arrive as companies face growing pressure to demonstrate that enormous spending on artificial intelligence produces measurable economic returns.
Technology suppliers have invested hundreds of billions of dollars in chips, data centres and cloud infrastructure.
Their customers, however, ultimately need AI systems to either generate additional revenue, reduce costs or improve productivity enough to justify continued investment.
Positive returns from early projects are therefore encouraging.
But the inability of many organisations to move from individual pilots to company-wide implementation suggests that the next stage of AI adoption could be determined as much by organisational change and infrastructure as by improvements in the models themselves.
Next Battle Is Implementation
The early phase of generative AI was dominated by questions about which company could build the most capable model.
For businesses, the next phase is increasingly about implementation.
Companies will need to determine where AI produces genuine productivity improvements, how employees should work alongside the technology and which processes should remain under direct human control.
They must also modernise the underlying technology systems needed to support AI at scale.
The BearingPoint findings suggest companies are already seeing evidence that artificial intelligence can produce financial value.
The bigger challenge is turning those isolated successes into sustainable organisation-wide transformation.
