Last updated: August 22, 2026
By Kris Drouet, Engineering Executive, in partnership with KORE1
An engineering budget gets approved when every line names an owner, a unit of measure, and a date. Finance rarely rejects the work itself. It rejects requests it cannot audit, rank against a sales hire, or defend to a board.
Twenty-five years in and I have made that ask badly more times than I want to admit. I have also sat on the other side of it, listening to somebody else make theirs. Same meeting, different chair. The difference between the asks that survive and the asks that get pushed to next quarter has almost nothing to do with how important the work is. Almost nothing.
The Budget Meeting Is a Ranking Exercise
Here is the thing most engineering leaders get wrong before they write a single line item. You think you are being evaluated. You are being compared.
Your CFO is not deciding whether your platform work is worth doing. She is deciding whether it beats three sales reps, a warehouse lease renewal, and a marketing program that came with a modeled pipeline number attached. All four of those asks show up in the same spreadsheet, in the same week, and three of them arrive with a unit of measure she has seen before.
Yours arrives in story points.
That is the whole problem in one sentence. Not politics. Not finance being hostile to engineering, which in my experience is mostly a myth engineers tell each other. The ask is written in a language that cannot be sorted against the other asks, so it goes to the bottom of a list nobody ever gets back to. I have watched genuinely critical infrastructure work lose to an office furniture refresh. The furniture had a quote.
Everything below is about giving the engineering line a quote of its own. Headcount first, since engineering staffing is usually the largest single number on the page and the one finance probes hardest.
What Finance Is Actually Holding When You Walk In
You should know what mood the room is in before you get there, because it changed in the last twelve months and a lot of engineering leaders are still planning against the 2024 version of their CFO.
Deloitte surveyed 200 finance chiefs at companies above $1 billion in revenue between November 14 and December 7, 2025. In that Q4 2025 CFO Signals report, 87% said AI would be extremely or very important to their finance operations, half named digital transformation their top priority for 2026, and 49% said they planned to hire or promote internally specifically to manage costs. Confidence hit 6.6, the highest reading since late 2021.
Read those together and the picture is not a CFO who hates technology spend. It is a CFO who is optimistic, has money, and has already decided where a lot of it is going. Which is harder, not easier.
Then there is the correction. Forrester’s 2026 Technology and Security Predictions puts it bluntly: enterprises will defer a quarter of their planned AI spend into 2027, with “fewer than one-third of decision-makers able to tie the value of AI to their organization’s financial growth.” Forrester’s read on what happens next is the part you should tape to your monitor. “CEOs will lean more on their CFOs to approve AI investments based on their ROI in 2026.” Sharyn Leaver, Forrester’s Chief Research Officer, framed the year this way: “In 2026, the AI hype period ends as the pressure to deliver real, measurable results from secure AI initiatives intensifies.”
So the CFO has more authority over technology decisions than she did two years ago, and she just watched a category of spend fail to produce a number. That is who is reading your budget. Plan accordingly.
The Four Lines You Actually Control
Most engineering budgets I review are organized by team or by project, which is also how most IT budget planning templates are laid out. Both are wrong for this purpose, because neither maps to how the money gets scrutinized. Finance sees four things. Build the document their way and half the argument disappears before it starts.
| Line | What it covers | The question finance will ask | Evidence that survives |
|---|---|---|---|
| People | Salaries, benefits, contractors, recruiting fees, severance | What happens to output if we hold headcount flat? | Named roles tied to named commitments, with a start date and a ramp assumption |
| Run | Cloud, hosting, observability, on-call tooling, support contracts | Why does this grow faster than revenue? | Unit cost per transaction, tenant, or user, trended across four quarters |
| Change | New capability, migrations, integrations, modernization | What breaks or stalls if we defer this two quarters? | An external clock: a contract date, an end-of-life notice, a regulatory deadline |
| Tools | Licenses, seats, AI assistants, CI minutes, security scanning | Who is using this, and what did the last renewal buy us? | Seat utilization pulled from the vendor console, not from the vendor’s slide |
Four lines. That is the entire structure. No more. Anything you cannot file under one of them is either somebody else’s budget or a project you have not thought through yet.
The reason this works is boring and mechanical. Finance can hold four categories in their head across eight departments. They cannot hold your twelve squads. Give them a shape they already use and your document stops being homework. Boring wins here.
Show Me the Data, and Then Show Me Who Paid for It
I ask for evidence on everything. My teams know it. What newer leaders miss is the second half of the question, and it is the half that determines whether a number helps you or quietly hurts you in the room.
Take cloud waste, since it is the single most common line item people try to reclaim. Flexera’s 2026 State of the Cloud Report, built on responses from more than 750 cloud decision-makers, put wasted cloud spend at 29%, the first increase in five years, and found 85% of respondents naming cloud cost management as their top challenge. Useful figures. Flexera also sells cloud cost management software.
That does not make the number false. It makes it a number your CFO is entitled to discount, and if you present it as neutral and she recognizes the logo, you have spent credibility you needed later. So say it first. “That figure comes from a vendor who sells the fix. Here is ours.” Then show yours. Every time I have run that comparison the internal number came back well under the published industry figure and still carried the room, because nobody in it could argue with where the number came from. Lower and better. That surprises people.
The same trap runs the other direction with macro forecasts. Gartner’s July 2026 update has worldwide IT spending reaching $6.37 trillion, up 14.2%, with data center systems alone jumping 62.5% to $822 billion. Every vendor in your inbox will quote that at you. It is a real number and it is nearly useless as justification, because “the industry is spending more” is not an argument. It is a weather report. Your CFO will say some version of “other companies are not my problem.” She is right.
Use macro data to set context in one sentence. Use your own data to ask for money. That is the whole rule.
Under the Hood of the Headcount Line
People are usually somewhere between sixty and eighty percent of an engineering budget, which means every other line is a rounding error by comparison and yet gets three times the discussion. Fix the headcount line and you have fixed most of the document.
Three things I would put in front of finance, in this order.
The cost of the seat staying empty. Most leaders budget the salary and skip the vacancy. If a senior platform engineer is the only person who can safely ship a service, and the search runs five months, you did not save five months of salary. You bought five months of a bottleneck, and Forrester expects developer hiring timelines to roughly double in 2026, which turns a nuisance into a planning assumption. Put a monthly number on the empty seat. Even a rough one. Any number defensible in a hallway beats no number in a spreadsheet.
The mix, not just the count. Four permanent hires and a contractor bench are not the same request even at identical cost, and finance knows it, because one commits the company for years and one does not. A contract engagement that covers a migration with a hard end date is often the easier approval, and it should be, since the risk profile is genuinely lower. I have traded a headcount request I probably would have lost for a six-month contract I won in a day. One day.
Comp bands you did not make up. If your salary assumptions came from what you paid the last person you hired in 2023, you are going to miss, and you will miss in the direction that blows up in Q3 when the offer gets declined. Pull current market data. KORE1’s salary benchmark assistant is one input. Two or three independent aggregators plus recent live offers is better, and the variance between them is worth putting in the document, because a range with a stated method reads as rigor while a single number reads as a guess. Method beats precision.
There is a version of this where the gap is not a budget problem at all. You need a leader, not four more engineers, and no amount of headcount fixes an org that has nobody empowered to decide. That is when VP of engineering staffing becomes the actual line item, or a fractional or interim VP of engineering if the need is real but the permanent seat is not funded yet. Interim leadership is one of the few requests that gets easier to approve the tighter the budget is, because it is an expense with a stop date rather than a permanent addition to the run rate. Stop dates sell.
The Build vs Buy Line Nobody Prices Honestly
Every budget cycle produces at least one line where somebody has quietly decided to build something a vendor already sells, and the estimate is wrong. Not maliciously. Engineers price the first version, which is the cheap part, and forget that they just signed up to own it in perpetuity. Forever.
I wrote a full framework for build vs buy decisions elsewhere and will not repeat it here. The budget-specific version is one line: whatever you estimate to build, add the annual carrying cost for three years and put that in the document too, because that is the number finance will discover in year two and remember in year three.
AI tooling is where this is going wrong most often right now. The seats look cheap. The productivity claim is enormous. Stack Overflow’s 2025 Developer Survey found 84% of developers using or planning to use AI tools, and, in the same population, 46% actively distrusting the accuracy of what those tools produce against 33% who trust it. The top frustration, cited by 45%, was AI output that is almost right but not quite, which is the expensive kind of wrong.
None of that is an argument against buying the seats. I buy them. It is an argument against writing “30% productivity gain” in a budget document you will be held to, because your CFO will remember that sentence in twelve months and ask what happened to the headcount you said you would not need.
Claim the tool. Do not claim the miracle.
Decide Your Cuts Before Somebody Else Does
Budgets get trimmed. Not because your plan was bad. Because revenue came in soft, or an acquisition landed, or a competitor did something and the board wants a response funded by Friday.
The leaders who come through that well have one thing in common, and it is the least glamorous habit in this entire article. They arrive with the cut list already written.
Three tiers, drafted in October, before anyone asks:
- What comes out at a 10% cut, and what specifically gets slower as a result. Usually a tooling consolidation and one deferred hire.
- What comes out at 20%, and which commitment you are formally withdrawing. This tier should hurt enough that you name a project that dies.
- What you will not cut at any number, and the sentence you will say when asked. Mine is almost always the thing with an external clock on it. A compliance date. A contract renewal. Something nobody in the room can negotiate with.
Handing that over feels like negotiating against yourself. It is the opposite. The leader who cannot answer “what would you cut” gets cut arbitrarily by somebody with no context, and I have watched that go badly enough times to consider the exercise mandatory. You are choosing to be the person who decides.
One caution. Do not put anything in tier one that you actually need, hoping it survives. Finance remembers. Do that twice and every future budget you submit gets treated as padded, which it now is.
What Finance Actually Asks in the Room
How far ahead should I start building this?
Ninety days before the submission deadline. The document takes a week. Gathering evidence your CFO cannot dismiss takes the other eleven.
Most IT budget planning calendars start too late. Unit costs need four quarters of history to show a trend, and pulling four quarters of clean cloud data for the first time is never the two-hour job people assume. Seat utilization has to come out of vendor consoles one at a time. Comp bands go stale in about a quarter. Start in the summer for a January cycle and you will be assembling. Start in November and you will be inventing.
My CFO already told me the number. Why bother with any of this?
Because a top-down number is an opening position, and the leaders who reshape it are the ones who show what the alternative allocation buys. The total may not move. What it funds absolutely can.
I have never once talked a CFO into a materially larger envelope by asking harder. I have several times moved money between lines inside a fixed envelope by showing what the current split was costing, which is a smaller win that compounds every year afterward, because next cycle you are the person whose numbers were right.
Our engineering costs are mostly cloud and it keeps climbing. What do I actually say?
Cost per unit, not total spend. Total cloud cost rising alongside a growing business is normal and finance knows it. Cost per transaction rising is the problem, and that is the only version anyone can act on.
Pick the denominator your business already reports. Transactions, active tenants, loans funded, orders shipped, whatever appears in the board deck. Then show the line. If cost per unit is flat while volume grows, you have a growth story and you should say so out loud, because nobody else will say it for you. If it is climbing, you have found your own strongest budget argument and you did not need a vendor to hand it to you.
Do I need finance in the room before the submission, or does that look weak?
Bring them in early. Every budget I have gotten approved cleanly was reviewed by somebody in finance before it was ever submitted.
The FP&A analyst assigned to your org is the most underused relationship in engineering leadership. That person knows the format that gets approved, the three questions your CFO always asks, and which line last year’s submission died on. Twenty minutes with them in September is worth more than a beautifully formatted deck in December. It also converts them from a reviewer into someone with a stake in your ask surviving, which matters more than anyone admits.
How much detail does finance actually want?
One page they can read in ninety seconds, backed by an appendix they may never open. The appendix is not wasted paper. It is what lets you answer a follow-up in the room instead of promising to circle back.
I have gotten this wrong in both directions. A forty-slide deck that got skimmed in four minutes, and a one-liner that fell apart the moment somebody asked how I calculated it. What works is a single page carrying the four lines, the delta against last year, and the source named beside every number. Everything else sits behind it. When she asks where the 29% came from, you do not say you will follow up. You turn to page nine.
Find the Line You Cannot Defend
Open whatever draft you have right now. Go line by line and find the one where, if she asks “where did this come from,” your honest answer is that it felt about right.
There is always one. Sometimes it is the biggest line on the page.
Fix that one before you touch formatting, before you build a single slide, before you rehearse anything. A budget document is not a persuasion exercise. It is a claim about the future that somebody is going to hold you to, and the fastest way to be believed next year is to be checkable this year. Checkable. Nothing fancier than that.
Before you submit, if there is one line you would rather nobody asked about, I will look at it with you. Connect with me on LinkedIn and paste the line.
Sometimes the weak line is not a line at all. It is a role you have been trying to fill since spring, and money was never what was blocking it. KORE1 has placed senior engineering talent for more than twenty years across 30-plus U.S. metros and holds a 92% twelve-month retention rate on those placements, which is the figure that matters when you have funded a seat once and cannot afford to fund it twice. Their engineering staffing practice works on contract, contract-to-hire, and direct hire, depending on what your budget can carry. Talk to a recruiter there if the seat you are trying to fund is one you have already failed to fill.
Related reading: Why “Tech Debt” Is the Wrong Frame, Show Me the Data: An Engineering Leader’s Framework for Build vs Buy Decisions, and Operational Discipline Is Not Bureaucracy.
