The narrative is well-established. AI is eating software. Foundation models have become broadly capable enough, and cheap enough, to take on tasks that previously required dedicated software. AI-assisted code generation is lowering the cost of building internal tools that previously would have been bought. The implication, drawn by investors and analysts across the spectrum, is that the budget feeding AI must be coming from somewhere, and the somewhere is the incumbent enterprise software stack.
For investors, three possibilities are worth testing against the data:
Additive. AI is net-new spend on top of existing stacks, funded by IT budget expansion.
Reallocated. AI is funded by squeezing services, contractor spend, and SaaS vendors, not yet by replacing flagship platforms.
Displacive. AI is actively replacing core enterprise software line items at the category level.
This piece tests which dynamic is actually playing out, drawing on three ETR data sources: the quarterly Technology Spending Intentions Survey (TSIS), which polls roughly 9,000 enterprise technology leaders on adoption, spending intent, and replacement activity across about 400 vendors and 29 sectors; the AI Product Series, a semi-annual survey covering enterprise AI behavior; and ETR's Shared Accounts methodology, which filters survey respondents using a specified vendor and surfaces what they are doing across the rest of their stack. For this analysis, the AI-forward cohort is defined as TSIS respondents using OpenAI, Anthropic, or both.
Investors keep asking whether AI is killing enterprise software. ETR's data through April 2026 says not yet, but the funding mechanics have already turned. The evidence below shows where the pressure is concentrated, which incumbents are absorbing it, and the two checkpoints that would signal displacement has arrived.
ETR polls enterprise technology leaders on year-over-year IT budget changes every quarter. The 2024-over-2023 read settled at 3.9% growth, 2025-over-2024 at 3.7%, and 2026-over-2025 prints 3.6% as of April 2026.
The intra-year pattern is also consistent. Forecasts start optimistic in the planning cycle. October and January prints typically run 4 to 5%, then mark down through the year. The 2025-over-2024 read peaked at 5.3% in January 2025 and printed 3.7% by January 2026. The 2026-over-2025 read is tracking the same shape, peaking at 4.6% in January 2026 before printing 3.6% three months later. Mid-single-digit growth is the operating reality. AI is being funded inside that envelope.
Two metrics matter when reading TSIS sector data. The first, Net Score, measures spending momentum within the citing base. It is a directional read on how aggressively buyers are increasing or decreasing spend. The second, Pervasion, measures stack presence, the share of respondents citing the sector at all. Think of Net Score as intensity and Pervasion as footprint. Read together, they reveal what is actually happening across the enterprise software landscape.
In April 2026, nearly every software sector shows year-over-year Net Score compression, including ML/AI itself. The decline is broad, not concentrated where the displacement thesis would predict. At the same time, every sector shows year-over-year Pervasion expansion. Stack presence is rising across the board. ML/AI is expanding 12.8 points year-over-year, more than twice the rate of the next-fastest sector.
ML/AI is the singular accretion outlier at +6.7 combined delta. RPA is a displacement outlier at -6.6, with spending intent collapsing as agentic AI substitutes for traditional automation workflows, even as Pervasion grows.
The middle of the table tells the story. On the accretion side, Information Security, Database, Cloud, Infrastructure-Software, ECM, Analytics, Contact Center, and UEM all post small positive combined deltas, accreting AI workloads rather than being replaced. On the pressure side, Enterprise Apps, Productivity, Marketing, Container Platforms, and Diversified Apps post small negative deltas. Pressured but not displaced. This is what reallocation looks like at the sector level. Broad pressure absorbed across the stack, with one clear winner and one clear loser.
The displacement question turns on funding mechanics. If AI is paid for out of net-new budget pools, the incumbent stack is safe. If AI is funded by squeezing existing IT spend, the incumbent stack is under pressure.
ETR's AI Product Series asks the question directly. Where does the budget for developing AI come from? Respondents select from broader IT budget (the operating wallet that funds renewals and infrastructure), innovation or digital transformation budget (typically ring-fenced strategic capital), R&D budget, and non-IT department budget.
Broader IT budget reliance is rising across every cohort. All Respondents climbed from 56.7% to 61.7% in twelve months. Large Enterprises moved from 60.0% to 65.1%. Global 2000 went from 60.0% to 64.2%. AI is migrating from initiatives funded by separate strategic capital toward initiatives funded out of the same wallet that pays for renewals and infrastructure. This supports the reallocation signal.
Innovation budget reliance fell among All Respondents (52.7% to 49.7%) but held steady or rose among Large Enterprises and Global 2000. The largest enterprises are running both pools simultaneously. Smaller and mid-market buyers are further along the reallocation curve. The model holds while IT budgets grow at mid-single digits and AI's share has room to expand from the current 14% level. The pressure point arrives when growth normalizes, and 2026-over-2025 is already tracking below 2025-over-2024.
To see where AI-forward buyers diverge from everyone else, we filter the TSIS panel to respondents using OpenAI, Anthropic, or both, then report their spending intent on other vendors as a separate Net Score. Comparing that cohort against the broader survey across waves reveals where AI-forward buyers are accreting versus pulling away.
Four vendors illustrate the pattern. Palo Alto Networks demonstrates accretion among AI-forward organizations, while ServiceNow holds ground but sees pressure. UiPath and Monday.com show what workflow displacement looks like.
Security accretes. Palo Alto Networks' Shared Net Score sits at or above its Survey Net Score across most waves from January 2024 through April 2026, printing 35.4% vs. 32.2% in the latest survey. AI-forward buyers are consistently more aggressive on perimeter security, consistent with AI deployment expanding the attack surface.
Flagship platforms are pressured but not collapsing. ServiceNow's two lines track closely, with the cohort line modestly below the full survey line in recent waves (38.9% vs. 41.7% in April 2026). Absolute Net Score remains strong. Deep regulatory and integration moats are doing their job.
Workflow software shows the displacement signal. UiPath's lines move together at low double digits, with the cohort consistently below (10.8% vs. 11.5% in April 2026). Monday.com's shared cohort line drops toward zero in recent waves (0.6% vs. 3.6%), indicating AI-forward buyers are functionally indifferent. These are categories where AI agents and AI-assisted code generation directly substitute for the underlying use case.
Two metrics from ETR's AI Product Series capture the forward trajectory. The first is AI's share of overall IT budget. Respondents allocate AI spend to a bucketed range, and a weighted midpoint average across all respondents produces a single comparable number per wave. AI's share rose from 12.1% in March 2025 to 13.3% in September 2025 to 14.2% in March 2026, a two-point structural expansion in twelve months. AI is becoming permanent infrastructure inside the budget rather than experimentation at its edge.
The second is the 12-month forward spend trajectory, calculated using the same midpoint methodology. That number was 14.4% in March 2025, 13.6% in September 2025, and 13.0% in March 2026. The rate of incremental AI spend is decelerating, albeit modestly across three waves.
excluded from base. Open-ended tail buckets ("51% or more" in either direction for spend trajectory) capped at 60% to avoid tail bias. N=600 per wave.
AI's footprint inside the IT budget is structurally expanding at roughly two points per year. The rate of incremental AI spend is decelerating at roughly one point per year. The math has already turned. With IT budgets growing 3.6% and AI spend growing 13%, non-AI spend is under pressure. That is reallocation, not addition.
ETR data through April 2026 indicates AI spending in the enterprise has moved past the purely additive phase. Reallocation has begun, but displacement has not yet arrived. Pressure is concentrated in services, contractor spend, and areas of software, plus the narrow displacement signals visible in workflow software and RPA. Large cap platforms are pressured but not collapsing.
The forward question is, when does reallocation become displacement? July 2026 and September 2026 are the next ETR data checkpoints. If AI's growth rate remains elevated while footprint keeps climbing, the marginal AI dollar increasingly must come from elsewhere in IT spend.