The 2026 software line item: budgeting for the AI premium
It is budget season, and every software line has an AI premium on top of it. Some of it buys capability. Most of it buys margin. Here is how to plan FY27 without funding the margin.
It is budget season, and every software line in your model has an AI premium sitting on top of it. Some of that premium buys real capability. Most of it buys margin you could cut without losing a single feature. The FY27 plan that works separates the two before it approves anything, because the vendors are counting on you budgeting them as one.
The macro numbers make the pressure concrete. Worldwide IT spending is growing about 14 percent in 2026, past $6 trillion. AI software specifically is growing 60 percent this year to roughly $453 billion, and Gartner has it growing another 41 percent in 2027. On that path, AI moves from about 41 percent of IT spending in 2026 toward half of all IT spend by 2027. Your software budget is being repriced around AI whether or not your usage justifies it. The planning question is which half of that repricing you fund.
"AI spend" is two line items pretending to be one
The single most useful thing a CFO can do in this cycle is stop budgeting "AI" as one number. It is two.
AI you own is capability running on your own model contract: custom agents on direct Anthropic or OpenAI billing, direct-API usage in your product, internal tools your team built. This is the productivity dividend. It is priced at raw model cost, which has fallen sharply, and it is yours to control. Grow this line deliberately.
AI you rent is the premium bundled into SaaS you already license: the mandatory AI tier on your platform renewal, the copilot half the team has not opened, the per-conversation agent priced well above the model underneath it. This is where the margin lives. Hold this line flat or down.
Budgeted together, the second line hides inside the first and the whole thing reads as "necessary AI investment." Split apart, the rent line is exposed as what it is: a price increase with a story.
The three buckets for the FY27 plan
- Differentiated direct AI: grow it. Where AI genuinely changes your product or your unit economics, spend on it, on your own contract, at model cost. This is the line that compounds in your favour, because the underlying cost keeps falling while the capability keeps rising.
- Forced-bundle AI premium: challenge every dollar. The AI uplift on 2026 renewals runs 20 to 37 percent, and some vendors go further. ServiceNow's Now Assist carries a 50 to 60 percent uplift on base pricing. Teams that treat the AI tier as a discretionary purchase requiring its own evaluation land 20 to 35 percent below teams that accept the bundle. Budget the challenge, not the acceptance.
- Rebuild candidates: fund a line for them. For the platforms where you use 15 to 25 percent of the product and the renewal keeps climbing, a fixed-price rebuild of the slice you use, on direct billing, replaces a rising rent line with a one-time build and a flat run. Put a rebuild line in the FY27 plan and point it at the top three climbers.
The move: reallocate, do not just absorb
The default plan absorbs the AI premium across every line and reports a double-digit software increase. The better plan reallocates. Hold the rented-AI premium flat by un-bundling and renegotiating, fund the direct-AI line from the savings, and use the rebuild line to convert your two or three worst rent-inflation offenders into flat cost. The total can land at or below last year while your actual AI capability goes up, because you stopped paying for the AI you were not using.
A worked planning shape
Take a $2M annual software budget. A typical stack audit finds 30 to 50 percent of it is cancellable or rationalisable at zero business impact, and the AI premium is landing hardest on the enterprise platforms inside it. A defensible FY27 shape: hold the two largest platform lines flat by un-bundling their AI tiers, cut $200K to $400K of genuine shelfware, redirect a third of that into direct-AI capability you own, and fund one rebuild of the highest-climbing slice with a payback under 12 months. The reported software number goes flat or down. The AI capability goes up. That is the opposite of what the renewal quotes assume you will do.
What not to touch
Scope honesty applies to budgets too. Do not cut security tooling, systems of record, or genuinely differentiated capabilities to make an AI-premium number look better. Do not rebuild a platform you use broadly and deeply just because its renewal went up. The discipline is not "cut AI." It is "fund the AI you own and interrogate the AI you rent," applied line by line. The three-question filter that works at the renewal table works at the budget table: is this priced from a higher-unit-cost era, what share do we use, and what is the rebuild math.
How to walk into the planning meeting ready
A two-week Strategy engagement produces the board-grade version of this: your software budget split into AI-you-own and AI-you-rent, the top rent-inflation offenders ranked, a rebuild candidate list with a number and a window on each, and the reallocation that holds the total flat. It is a written memo, not a deck, and it is built to survive the CFO and the board asking hard questions about every line.
Before that, get the shape yourself. Run the SaaS stack audit to find the cancellable spend, and the AI Agent ROI calculator to price any agent line against a direct-billing build. The budget that plans for the AI premium beats the budget that absorbs it, every time.
Read more: /strategy/ · /rebuild/ · Stack rationalisation brief · SaaS stack audit