BCA Research says agentic AI is forward of its results

In a report of 23 October, BCA Research strategist Matt Dyson warned that investor interest in agentic AI grows faster than the revenue and productivity it gives. Agentic AI is AI that does tasks alone. To keep good sentiment through 2027, BCA Research says AI must give more revenue or a new app, such as a good personal assistant.

The gap

A 2026 McKinsey survey found that only 6% of the persons who answered connect a minimum of 5% of operating earnings to AI. BCA Research says AI with human oversight can do some tasks at about 75% lower cost than US employees.

The hidden costs

BCA Research says this saving leaves out the “AI validation tax”, which is the time staff use to check AI work. It is not sure that more use of tokens means more work that helps. Reasoning models use four to 15 times more tokens. Dyson says only 20% of freelance tasks by top models agree with the standard of a client.

The other side

In January, the MSCI US IT Services index fell 15% after Anthropic launched Claude Cowork plugins. BCA Research calls that reaction too strong. It says smaller businesses will probably keep established software.

Insight: Spending plans grow faster than proof of results

Gartner found that only 13% of 2,002 technology leaders see large AI results, but 75% of CEOs plan more technology spending. BCA Research adds that use of tokens can increase with no increase in output. Use numbers go up faster than proof of results. Because of this, buyers, and not vendors, must close the distance with their own measures. Count results, not tokens.