Kolsky Outlines Enterprise AI Budget Priorities

Enterprise AI budgets are likely to tighten in 2027 and face closer scrutiny over returns and value, according to Esteban Kolsky, Chief Distiller at Constellation Research. Speaking at Connected Enterprise 2026, Kolsky said cybersecurity is taking back part of AI funding, sometimes while that spending remains labeled as AI. Global shocks have repeatedly derailed investment plans, and a show of hands at the event indicated that technology budgets are rising while AI is no longer the top priority it was in 2024 and 2025. Kolsky said pushback against AI budgets began in October 2025, followed by the “tokenmaxxing” fad and efforts by enterprises to control funding.

Kolsky urged executives to assess AI investment across six areas: business value, data readiness, infrastructure readiness, governance, economics and talent. He said AI value follows a J-curve and tends to appear around year two or three, while investments may run for five to 10 years; he argued that executives will not find value by focusing only on return on investment. Agents need access to enterprise data, but Kolsky said some vendors restrict that access. He expects infrastructure spending to cover cloud, on-premises systems and technology neglected since the Covid-19 pandemic.

Kolsky said governance remains mostly discussion, and argued that chief financial officers will not accept token-based or outcome-driven consumption pricing. He called talent the most neglected budget category: experienced operators are needed to handle exceptions agents cannot manage, and layoffs risk discarding company knowledge needed to build models. He also said training employees to write prompts is not a talent strategy.

Enterprise AI budgets are likely to tighten in 2027 and face closer scrutiny over returns and value, according to Esteban Kolsky, Chief Distiller at Constellation Research. Speaking at Connected Enterprise 2026, Kolsky said cybersecurity is taking back part of AI funding, sometimes while that spending remains labeled as AI. Global shocks have repeatedly derailed investment plans, and a show of hands at the event indicated that technology budgets are rising while AI is no longer the top priority it was in 2024 and 2025. Kolsky said pushback against AI budgets began in October 2025, followed by the “tokenmaxxing” fad and efforts by enterprises to control funding.

Kolsky urged executives to assess AI investment across six areas: business value, data readiness, infrastructure readiness, governance, economics and talent. He said AI value follows a J-curve and tends to appear around year two or three, while investments may run for five to 10 years; he argued that executives will not find value by focusing only on return on investment. Agents need access to enterprise data, but Kolsky said some vendors restrict that access. He expects infrastructure spending to cover cloud, on-premises systems and technology neglected since the Covid-19 pandemic.

Kolsky said governance remains mostly discussion, and argued that chief financial officers will not accept token-based or outcome-driven consumption pricing. He called talent the most neglected budget category: experienced operators are needed to handle exceptions agents cannot manage, and layoffs risk discarding company knowledge needed to build models. He also said training employees to write prompts is not a talent strategy.