What to put in the 2027 technology budget, and what to take out

Most technology budgets fail review for the same reason: they are a list of things IT wants, priced, with no way for a reader to tell which are obligations and which are choices. Here is a structure that survives the conversation.

Start from the run cost, not the wish list

The first number is what it costs to keep everything working with no new projects at all: licences, hosting, support contracts, the people who run it. That number is not optional, and putting it first changes the tone of the review. Everything after it is a decision the business gets to make.

If you cannot produce the run cost cleanly, that is the first thing to fix. It is also the number most likely to be challenged, so it needs to reconcile to the general ledger rather than to a spreadsheet only IT has seen.

Budget AI consumption as a variable cost

Assistants and agents are increasingly billed by usage rather than by seat. That puts a variable cost inside a budget structured for fixed ones, and it behaves differently: it rises with adoption, which is the outcome you are funding.

Give it its own line, forecast it from an expected volume rather than from a licence count, and set a ceiling with an alert. A budget that assumes flat AI spend will be wrong in whichever direction adoption goes.

Put the licence renewals where someone will see them

Renewals are where most of the surprise lives, and they are usually managed by whoever happens to hold the relationship. List them by date, with the cost, the notice period, and whether the price is contracted or open. Three of them will be worth a conversation with the vendor and the rest will not — but you cannot tell which without the list.

Name what is being retired

A budget with no subtractions reads as a wish list, and reviewers treat it as one. Every plan should name the applications, licences or contracts coming out, with dates and the saving attached.

This also protects the additions. “We are adding this and removing that” is a different conversation from “we are adding this.”

Separate run, grow and transform

Three numbers do more work in a review than forty lines of detail:

• Run — keeping what exists working.
• Grow — improving what exists so it costs less or does more.
• Transform — building something the business does not have today.

Most organizations discover their split is closer to 80/15/5 than they assumed, and that the transform number is smaller than the ambition attached to it. That is a useful thing to find in September rather than in June.

What to take out

Projects with no named business owner. If nobody outside IT will stand behind the benefit, it will not survive the first cut and should not consume the planning time.

Tooling consolidation with no driver. Worth doing when something forces it. Rarely worth funding on its own.

Contingency spread invisibly across lines. Reviewers find it, and it costs credibility on every other number. Put it in one place and label it.

Anything already funded and not delivering. A project that has not moved in two quarters is a decision, not a line item. Stop it, restart it, or give it what it actually needs.

How to present it

A CFO wants to know three things: what has to happen, what changes if we spend more, and what breaks if we spend less. If your budget answers those three in a page, the detail behind it will rarely be contested.