AI ROI stalls: workflow overhaul emerges as key fix

By mid-to-late 2026, reports from solo operators to enterprises converged on the same diagnosis: AI adoption was stalling not because models were weak, but because workflows were undefined, fragmented, or poorly integrated into daily operations. Affiliate Blogging Academy captured the small-business version bluntly, saying that over 90 days it rebuilt the operational spine of its business using AI tools by clearing obstacles to doing the work well, while IT Brief New Zealand found SMEs still struggling to turn AI hype into outcomes because they lacked the time, expertise, and practical guidance to fit AI into operations.

Enterprise evidence pointed to the same bottleneck at larger scale. According to Chloe Sikarskie, the Stage 2 to Stage 3 transition, from fragmented AI tools to integrated, measurable workflows, is where most enterprise AI programs stall, and McKinsey’s 2025 research across 1,993 organizations found that workflow redesign had the strongest correlation with EBIT impact. Only 6% of companies qualify as AI high performers, generating more than 5% EBIT impact from AI, and Sikarskie said the organizations generating the strongest returns from AI in 2026 treated workflow redesign as the primary investment, a pattern reinforced by CIO reporting and by fintech onboarding and Pythian case studies centered on orchestration, governance, and connected systems.

The late-2026 lesson was that AI can compress exception handling, but it cannot rescue a bad process. Supply Chain Now noted agentic systems can collapse “the last mile of exceptions,” yet “you still have to go understand the systems” and ask “What’s wrong? What’s broken?”; NBER’s cited finding that AI boosted support-agent productivity 14% on average, while “the technology helped people perform work, not eliminate the need to understand it,” points to the same constraint: automating fragmented handoffs speeds activity, but not necessarily outcomes.

That is why early volume wins so often turned into workflow debt. MIT Sloan Management Review Middle East reported that Klarna’s AI “handled 2.3 million conversations… accounting for two-thirds of Klarna’s customer service chats… It did the work of 700 full-time employees and reduced repeat inquiries by 25%,” yet CEO Sebastian Siemiatkowski later admitted, “We probably overindexed a little bit on that, and then in the last six months we have been trying to course correct,” after cutting costs too aggressively with AI led to service quality falling; the durable pattern, echoed by InfoWorld and ET Edge Insights, is that scale requires integration, governance, and clear ownership because “just 5% of integrated enterprise GenAI pilots were extracting millions of dollars in value,” and even though “88% of organizations use AI but only 7% have scaled it (McKinsey).”

By 2026, the financial case for AI remained concentrated in a small minority. Diginomica, citing BCG’s “The Widening AI Value Gap,” reported that only 5% of companies in a 1,250-firm global study were achieving AI value at scale, while 60% saw no material value; PwC’s latest Global CEO Survey was similarly bleak, with only 12% of CEOs reporting both cost savings and revenue benefits and 56% seeing neither. Harvard Business Review Analytic Service likewise found only 16% of 385 decision-makers reporting a high degree of measurable value, while 69% described impact as merely moderate or slight.

The sharper pattern was that meaningful returns appeared only when companies changed how work was done, not when they simply added AI tools. Venture Curator summarized the convergence bluntly: only around 5% of companies are generating meaningful returns, including McKinsey’s finding that just 5.5% report significant EBIT impact, and it identified the key predictor as workflow redesign before deployment: it was whether they redesigned their workflows before introducing AI. McKinsey found that top-performing companies were nearly 3x more likely to redesign workflows before automation, with 55% redesigning workflows first versus only 20% of everyone else. Briefglance’s coverage of Klarus showed the same gap in the mid-market: 73% had begun deploying AI, but only 10% had successfully scaled all initiatives.