Dot grid
Answer
>
Is AI spend worth it

Is our AI spend worth it? How finance shows the board (2026)

Scale the proof to the spend: a sanity check for a small AI budget, and an outcome-by-outcome case for a material one. Last updated: November 2026.

Team Aleph
Shaping the future of AI-native FP&A
Share to
Table of contents
Subscribe to the 10X Finance Blog

Get FP&A best practices, research reports, and more delivered to your inbox.

Thanks — check your inbox.
Something went wrong. Please try again.

To show the board your AI spend is worth it, scale the proof to the size of the spend. For a small AI budget, a sanity check is enough: teams report real value, nothing concerning has surfaced, and more analysis would cost more than it would tell you.

A material AI line needs more. For each investment the board could reasonably ask about, show the outcome it was bought to move, the baseline before, what moved after, and the funding decision that follows. A board answer can be as short as this: "We put $250,000 into AI for support to cut service costs. Resolution time is down 18% and cost per ticket is down 12%, so we're keeping it and expanding it to two more workflows."

Bottom line: Don't build one company-wide AI ROI percentage. Match the depth of proof to the size of the spend, and for every material AI investment show four things: the outcome, the baseline, what moved and what you'll fund next.

Why is it so hard to prove AI spend is worth it?

Because most of the value shows up as time saved and work done differently, and almost nobody measured the "before". In a Forbes Research survey of 1,075 C-suite executives, cited in our budgeting for AI spend guide, 85% said AI is improving their decision-making, but fewer than 1% could show a significant return.

Three things make the gap wide:

  • The benefits are spread thin. A chat assistant saves each person an hour here and there. That's real, but it doesn't land on one line of the P&L.
  • The spend is scattered. Seats sit in one budget, API usage in another and AI inside existing software in a third, so the cost side is hard to total, let alone the benefit side.
  • Nobody set a baseline. Teams adopted AI first and asked about value later, so there's no clean before-and-after.

None of that means the spend isn't working. It means finance has to decide how much proof the board needs, then collect that much and no more.

How much proof does your AI spend need?

It depends on how big and how central AI is to the business. The guide sorts companies into three levels:

AI's role in the businessWhat the board needsWhat finance shows
Small, straightforward line itemA directional answer you can defendA sanity check: value reported, no red flags, spend in a sensible range
Material part of the budgetEvidence for each major investmentOutcome, baseline, what moved and the funding decision, investment by investment
Core to the product or operating modelActive management of value and costThe ROI cases plus unit costs, model right-sizing and a protected experimentation budget

Pick the highest level your spend, complexity or business model requires. Most companies sit at the first or second level today, and the board's expectations rise as the line grows.

Turn AI spend into a number you can defend. Download the guide: Budgeting for AI spend

How do you sanity-check a small AI budget?

Ask three questions, and if the answers are good, stop there:

  1. Do people feel they're getting value? Teams can point to real time savings, or to things they can do now that they couldn't before.
  2. Has anything concerning surfaced? No security, privacy, compliance or control issues that would make you question whether the tools are safe to use.
  3. Is the spend big enough to justify more analysis? As the guide puts it, it isn't worth spending $100,000 of effort proving the return on a $50,000 line item.

Then check that the total is in a sensible range for a company like yours. In BCG's AI Radar 2026, companies said they plan to invest about 1.7% of revenue in AI in 2026, roughly double the year before. By industry, planned investment runs from about 2.1% of revenue in technology and 2.0% in financial institutions to 0.8% in industrials and real estate. These are planned investment figures, not realized spend, and they include talent, training and data as well as tools, so treat them as an outer bound rather than a target.

A small-budget board answer sounds like this: AI is a small part of the budget, the teams using it report real value, no material security or control concerns have come up, and deeper analysis wouldn't change the decision today.

How do you build an AI ROI case for a material AI budget?

Answer four questions for every AI investment the board could reasonably ask about. You don't need one blanket enterprise-wide return; you need a short, specific case per investment:

  1. Pick the outcome. What did we buy this to improve? Be specific: "improve productivity" is too vague. Reduce close time, increase rep capacity, lower support cost per ticket, improve conversion or reach a product milestone sooner.
  2. Set the baseline. Where were we before AI? Close took eight days. Reps handled 40 accounts each. Support cost $18 per ticket.
  3. Show what moved. What changed after the investment? Close fell to six days. Rep capacity rose 20%. Cost per ticket fell 12%. Clear quantitative change beats perfect attribution.
  4. Make the call. What do we do next? Expand what's clearly working, adjust what's mixed and stop funding what isn't delivering.

The fourth step is the one boards care about most. A case that ends in "so we're expanding it" or "so we're stopping it" turns AI spend from a cost the board has to trust into a decision it can see.

Worked example: the support AI case

The guide's template answer is about a $250,000 investment in AI for customer support, bought to lower service costs and add capacity. Since rollout, resolution time is down 18% and cost per ticket is down 12%. The decision: keep funding it and expand it to two more workflows.

To show the board the money as well as the percentages, put the baseline into dollars. Say the team handles 100,000 tickets a year at $18 each, or $1.8 million. A 12% fall in cost per ticket is worth about $216,000 a year. Against a $250,000 investment, the cost saving alone pays it back in a little over a year, before counting the faster resolution or the capacity freed up for growth.

The ticket volume here is illustrative; use your own. The structure is what matters: outcome, baseline, result, decision, with one number the board can check.

How do you measure AI value when there's no baseline?

Rebuild the "before" from systems that already kept records, and compare against teams that didn't adopt. Most companies rolled out AI before anyone measured anything, so this is the normal case, not the exception. Four ways to get a usable comparison:

  • Use system history. Ticketing tools, the close calendar, the CRM and the general ledger all hold data from before rollout. Resolution times, close dates and accounts per rep can be pulled for the prior year.
  • Compare against a team that hasn't adopted yet. If one region or team rolled out later, its numbers over the same period are a rough control group.
  • Start measuring now with a holdout. For the next investment, keep a small group on the old process for a few weeks and compare.
  • Discount self-reported time savings. Surveys are useful for direction, but self-reported hours saved are hard to check. Use them to rank where value shows up, not as the dollar figure.

Tell the board which method you used. A rebuilt baseline is fine if you say it's rebuilt.

When should you stop funding an AI investment?

When the outcome it was bought for hasn't moved and there's no clear reason to expect it will. A few signals to watch at each review:

  • The primary metric is flat after two review cycles, with steady usage. The tool is being used but isn't changing the result.
  • Usage has faded. Seats or runs are falling and the team has gone back to the old way of working.
  • The unit cost keeps rising. Cost per ticket, per run or per user is climbing faster than the value it produces.
  • A cheaper option does the same job. Another tool you already pay for now covers the use case.

Stopping is part of the answer the board wants. A finance team that can show what it stopped funding is more credible when it asks for more.

What should you measure for each kind of AI spend?

Measure the outcome the spend was meant to move, using a metric the business already tracks. A few common pairings:

  • Finance AI: days to close, hours spent on variance commentary, or forecast cycle time. Our guide to automating month-end close commentary shows where AI shortens the close.
  • Sales AI: accounts or opportunities per rep, or time from lead to first meeting.
  • Support AI: cost per ticket, resolution time, and the share of tickets resolved without a person.
  • AI agents: cost per run and the work each run replaces. Our guide to AI agents for finance and FP&A covers how agents fit into the finance workflow.
  • Product AI: gross margin per customer, conversion or retention for the features that use it.
  • Chat assistants for everyone: adoption and a short quarterly survey of time saved. For broad seat spend, that's usually as precise as it's worth getting.

Pick one primary metric per investment. A case with six metrics invites the board to argue about the weakest one.

How do you present AI value to the board?

In three parts, on one page. Start with the total and where it went, then show the cases, then the decisions:

  1. The spend. One reconciled AI total for the year, split by type: employee tools, product development, customer-facing cost of revenue and any capitalized work. Our guide to whether AI spend is COGS or OPEX covers the split. Our guide to finding hidden AI spend covers the costs a first total usually misses.
  2. The evidence. A short table with one row per material investment: cost, outcome, baseline, result.
  3. The decisions. What you'll expand, adjust and stop, and how that shapes next year's AI budget. Our guide to forecasting AI spend shows how to turn those decisions into the FY2027 number.

Be clear about what's proven and what's promising. A board will accept "we're still measuring this one" for a new investment; it won't accept a confident return figure that falls apart under one question.

What about AI spend that's hard to measure?

Ring-fence it and judge it differently. Some AI spend is exploration: new tools, new models and new use cases that haven't reached a workflow yet. Give it a fixed experimentation budget and judge it by what you learn and what graduates into real work, not by whether each experiment pays back immediately.

The same goes for power users. AI usage is concentrated in a small group of people, and heavy usage can look like waste on a cost report while being the most productive work in the company. Before you cut a heavy user's spend, look at what they produce. Our guide to reducing AI costs without losing the gains covers which costs to protect.

Tell the board you're doing this on purpose. A named experimentation budget with a clear rule for what graduates is easier to defend than spend nobody can explain.

How is this different from FP&A software ROI?

It answers a different question. Proving AI spend is worth it is about the return on everything the company spends on AI, across every team. Estimating the return on a specific software purchase, such as an FP&A platform, is a single business case with its own costs and savings. Our guide to estimating ROI from an FP&A platform covers that case, including why vendor ROI calculators overstate.

The methods share one rule, though: measure what's measurable, leave out what isn't, and say so.

Common mistakes when proving AI spend is worth it

  • One AI ROI percentage for the whole company. It's either unprovable or meaningless. Show cases per investment.
  • No baseline. If you didn't measure the "before", say so and start measuring now, rather than estimating it after the fact.
  • Giving AI credit for everything that improved. If close time fell because you also hired a controller, the AI case is weaker than it looks.
  • Proving the small stuff in detail. Spending weeks on the return of a modest seat budget costs more than it's worth.
  • Ending without a decision. Evidence that doesn't change what you fund isn't an answer to the board's question.

Show what your AI spend delivers with Aleph

Aleph connects your general ledger, AI provider usage and HRIS to Excel, so AI spend refreshes on its own, sits next to the rest of the P&L and carries variance against plan, which gives every ROI case a reliable cost side. Get the full method in our budgeting for AI spend guide, watch the webinar from AI spend to AI budget, or book a demo to see it on your own data.

See your FP&A upside in 15 minutes   Discover how top finance teams eliminate manual work and unlock value with Aleph.  
Subscribe to the 10X Finance Blog

Get FP&A best practices, research reports, and more delivered to your inbox.

Thanks — check your inbox.
Something went wrong. Please try again.

Frequently asked questions

Measure it per investment, not for the company as a whole. For each AI investment, name the outcome it was bought to improve, record the baseline before rollout, measure what changed after, and decide whether to expand, adjust or stop funding it. Put the change into dollars where you can, such as cost per ticket times ticket volume.

There's no single right figure. In BCG's AI Radar 2026, companies planned to invest about 1.7% of revenue in AI in 2026, from about 2.1% in technology to 0.8% in industrials and real estate. Those are planned investment figures that include talent and data, so build your own number from your seats, usage and AI projects.

Say so, and show how you'll find out. For a small budget, a sanity check (teams report value, no red flags) is a defensible answer. For a material one, pick one outcome per investment, start measuring the baseline now, and set a date when you'll make the funding call.

Finance should own the total and the reporting format, and each business owner should own the case for their investment: support for support AI, sales for sales AI, engineering for coding tools. Finance checks that every case has a baseline and ends in a funding decision.

At least once a year with the budget, plus a short update when a major AI investment is approved or reviewed. Where AI is a material or core cost, add a quarterly line on AI spend against plan and the status of each major case.

Discover Aleph today

A 30-minute demo on your numbers, not ours.
Screenshot of an income statement spreadsheet comparing revenue, cost of revenue, and operating expenses for Jan 25 and Feb 25, alongside a sidebar menu with options including 'Income Statement,' 'Analyze with AI,' and other budget categories.
Dotted grid