Global AI investment is forecast to exceed $1 trillion in 2026, yet most companies cannot find material cost savings on their financial statements. The disconnect between massive spending and missing returns has CEOs and investors demanding answers, and for product development leaders, the explanation-and the solution-sits inside a concept called the J-curve.
The J-curve describes a pattern where results initially worsen after heavy upfront investment before leveling off and rising. Four years after ChatGPT launched, most organizations remain stuck in the trough. "We blew through our AI budget in a quarter, for the whole year," Uber CEO Dara Khosrowshahi said in June 2026. Alex Karp, CEO of Palantir, told CNBC his customer CEOs are livid because "they're paying [millions] for tokens that create no value."
McKinsey's 2026 survey found that 80% of respondents report personal productivity gains from AI, but only 37% see any EBIT impact, and just 6% qualify as AI high performers. The gap between individual efficiency and firm-level financial performance is where product development teams lose the thread.
Why companies get stuck in the trough
Most organizations treat AI as a tool for automating tasks rather than transforming the business. Research points to several specific failure patterns. Companies measure success by token usage instead of customer outcomes or financial performance. Teams save time but no leader decides how that freed capacity should create additional sales coverage, product releases, or cost removal.
AI often speeds up one step without touching the full workflow cycle. A proposal takes 20 minutes to draft but still waits two weeks for review. The operating model-roles, decision rights, incentives, and approval rules-remains designed for pre-AI work. Pilots multiply everywhere while accountability stays nowhere, with business-unit leaders owning no baseline, benefit target, or deadline for what the source calls "Conversion."
Conversion: the precursor to financial ROI
Conversion means recognizing AI efficiency gains and deliberately channeling them toward a measurable financial outcome. This includes more sales but also intangible values like higher quality products, improved customer service ratings, and faster throughput workflows. For AI for Product Development teams, the distinction matters. Localized task improvements must convert into firm-level performance that investors reward.
Before approving any AI use case, the sponsoring executive should state exactly how the firm will capture value. High-level metrics fall into five buckets: more throughput, higher quality, faster cycle times, lower cost-to-serve, and revenue growth. Without this upfront clarity, time saved dissolves without trace.
Redesign workflows and redeploy capacity
Agentic AI can build new websites from a single prompt or set up supply chain efficiencies in minutes. But if a pricing recommendation takes five minutes and final review still takes two weeks, the bottleneck has simply shifted. Product development leaders must map old workflows from triggering event to post-sale satisfaction score, identify each new constraint, and redesign approval processes so humans enter only when data security or quality control requires it.
The real pivot from the bottom of the curve upward requires explicitly measuring and redeploying saved capacity. Job cuts alone will not pay for AI investments. Teams using pricey frontier models should adopt one primary, lower-cost, open-source model for in-house work, then target one or more value-creating destinations: more sales coverage, faster testing cycles, fewer defects, more product demos, or stronger compliance controls. No major efficiency gain should remain ownerless and unaccountable to the P&L. Product managers who follow an AI Learning Path for Product Managers can build the skills to track and convert these gains systematically.
Why this matters for product development
The bottom of the AI J-curve is not where companies fail because models are expensive or hallucinate. It is where companies fail because management treats a general-purpose technology as a plug-and-play task helper instead of an enterprise-wide operating model change. For product development teams, this means the work shifts from performing tasks to managing multi-step agentic AI workflows. The firms that turn the curve redesign the workflow, redeploy earned capacity, and hold business-unit leaders accountable for measurable value-creating outcomes. The question is not whether your AI can generate ROI-it is whether your team has a plan to convert speed into revenue before the budget runs out.
Your membership also unlocks:
