Making AI work requires careful budgets

The article argues that AI adoption requires disciplined budgeting and clear use-case selection, warning that broad experimentation can lead to 'pilot purgatory.' McKinsey’s Basel Kayyali and StackAI’s Jonathan Kleiman say companies should target a few end-to-end workflows tied to growth, productivity, or new business, and prioritize easy-to-build use cases with measurable cost or revenue impact. The piece is mainly strategic commentary for IT leaders rather than a company-specific market-moving event.

Analysis

The market implication is not “more AI spend,” but a forced reallocation from experimentation to workflow capture. That favors vendors that sit closest to measurable business processes and can show payback inside a single budget cycle, while punishing broad horizontal platforms that rely on endless pilots and executive enthusiasm. In practice, the winners are likely to be software layers that compress a specific operating expense line or accelerate a revenue workflow; the losers are consulting-heavy rollouts and generic copilots that become easy to defer when CFO scrutiny tightens.

Second-order, this shifts bargaining power toward finance and line-of-business owners versus centralized IT. Over the next 1-2 quarters, procurement will likely demand usage-based pricing, hard ROI gates, and shorter implementation milestones, which should pressure gross-margin-rich AI tooling that is still priced on promise rather than throughput. That is a negative for vendors with long sales cycles and diffuse use cases, but a positive for application-specific automation names that can prove payback in days or weeks.

The contrarian angle is that “pilot purgatory” may actually be a feature, not a bug, for the ecosystem: stalled enterprise pilots create a long-tailed funnel of latent demand that can convert abruptly once budgets are forced through a few approved domains. If that happens, the spend inflection could be back-end loaded into 2025, making the current caution around enterprise AI adoption look like a timing problem rather than a thesis break. The key catalyst is budget season: if management teams move from exploration to named workflow deployment, adoption quality matters more than headline AI penetration.

Tail risk is that CFOs use this discipline to cut AI breadth so aggressively that even high-quality vendors see near-term deceleration before productivity savings show up. That would hit sentiment in the next 1-2 earnings cycles, especially for companies monetizing assistant-style tools without clear cost takeout. The reversal catalyst would be a few high-visibility case studies showing sub-12-month payback, which would unlock follow-on spend much faster than model quality improvements alone.

Key Decisions for Investors

• Long application-layer AI vendors with clear ROI hooks vs. broad AI platform exposure over the next 3-6 months; favor names that sell into finance, support, or back-office automation where payback can be measured within a quarter.
• Short the basket of consulting / implementation-heavy AI beneficiaries on any strength for the next earnings season; these are most exposed to pilots slipping and budget gates tightening.
• Pair trade: long workflow automation / enterprise software with usage-driven monetization, short horizontal AI infrastructure or copilots where monetization depends on adoption breadth rather than immediate cost takeout.
• Use call spreads, not outright longs, on enterprise AI leaders into budget season; the thesis is a deferred conversion of pilots, so upside may arrive with lag while downside from budget scrutiny can show up immediately.