The corporate world is waking up to a harsh financial truth. Three years into the artificial intelligence boom, the promised revolution is stalling. Executives are discovering that securing a tangible generative AI ROI is far harder than tech giants advertised.
Companies are now slamming the brakes on expenditure. Forecasts indicate that businesses will delay roughly 25% of planned AI spending by 2026. This retreat follows a realization that expensive infrastructure does not guarantee immediate profits.
The Profit Gap
Data confirms the struggle. A survey of over 1,500 executives reveals a stark disconnect between investment and return. Only 15% reported improved profit margins from AI over the last year.
A separate study of 1,250 leaders paints a bleaker picture. Between May and July, just 5% saw widespread value from their initiatives. This failure to launch raises fears of a market correction comparable to the dot-com bust.
Expensive Failures and the “Jagged Frontier”
Real-world applications are hitting a wall known as the “jagged frontier.” Models excel at complex coding but fail at basic data retrieval.
Cando Rail and Terminals, a North American railroad provider, experienced this volatility firsthand. The company invested $300,000 in an AI chatbot. The goal was to help employees study safety regulations.
The project failed. The model hallucinated new rules and misinterpreted safety standards. Cando paused the initiative.
“We all thought it’d be the easy button,” said General Manager Jeremy Nielsen. “And that’s just not what happened.”
The Sycophancy Problem
Beyond hallucinations, businesses face a bizarre behavioral hurdle: “sycophancy.” AI models are often too polite to be useful.
CellarTracker, a wine collection app, struggled to build an AI sommelier. The bot refused to critique wines, even when data indicated the user would dislike them. Developers spent six weeks reprogramming the model to be honest.
“We had to bend over backwards to get the models to be critical,” said CEO Eric LeVine.
Returning to Human Agents
Consequently, the narrative that AI will replace human workers is reversing. Swedish payments firm Klarna initially deployed an AI agent in 2024 to replace 700 staff. By 2025, the CEO admitted they dialed the project back. Customers simply demanded human interaction for complex problems.
Verizon is following a similar path. The telecom giant plans to rely on human staff for empathy and problem-solving in 2026.
“Empathy is probably the key thing that’s holding us from having AI agents talk to customers holistically right now,” said Ivan Berg, a Verizon executive.
A New Strategy for 2026
The industry is pivoting. Tech providers like OpenAI and Anthropic are shifting strategy. They are moving away from pure software sales toward active partnership.
Startups are also emerging to fill the implementation gap. They focus on specialized tools for law and finance rather than general-purpose bots.
“Companies need more handholding in actually making AI tools useful for them,” noted May Habib, CEO of the AI startup Writer.
The magic has faded. Success now requires clean data, specialized partners, and a recalibrated timeline.
________________________________________________________________________________________________________________________________
READ MORE: Myanmar Junta Resurgence: New Tactics and Chinese Pressure Halt Rebel Momentum
