# AI ROI Calculator: Will AI Pay for Itself in Year One? | SpliceOps URL: https://spliceops.lrgfrmt.com/tools/ai-roi-calculator/ > Work out the first-year return on rolling AI out to your team: cost, value of time saved, ROI and payback. Free calculator built on DORA's open method. # Will AI pay for itself in year one? Rolling AI out costs money before it saves any: tools, training, and a dip in output while people learn. Put those costs next to the time AI frees and you get a first-year return you can defend. ## How the number is built - **Spend** = (AI tools + training) × people + platform and security - **Output lost while learning** = people × cost per person × % of output lost × (months it lasts ÷ 12) - **Year-one cost** = spend + output lost while learning - **Value of time saved** = people × cost per person × % of work time saved, counted only for the months after people have learned - **Year one, after costs** = value of time saved − year-one cost, and **ROI** = that ÷ year-one cost - **Payback** = the months until time saved covers the year-one cost. Nothing is saved while people learn, and from year two the AI tools and platform costs come back. If year two loses money, it doesn't pay back. - **Year two** = a full year of time saved − AI tools − platform and security. Training and the learning dip don't repeat. - **Break-even** = the % of work time AI must save for year one to come out even The costs and the starting values follow [DORA's AI ROI calculator](https://dora.dev/ai/roi/calculator/), which publishes its formulas in the open. The starting values describe an example software team, not a benchmark. We start from a team of 50 rather than DORA's 500. The biggest difference from DORA: DORA counts a full year of time saved on top of the dip, while we count time saved only once the dip is over, because a team can't be slower and faster in the same months. That makes year one lower and payback later than DORA's, so we add year two to show where the return comes from. Payback also takes off the costs that come back in year two. Time saved can't go below zero, and where a figure can't be worked out, we say why. The math runs in your browser. Your numbers are saved in the page's web address so you can share the result, which means anyone who opens that link, including the site's host, can see them. Updated 1 Nov 2026 · SpliceOps ## A worked example Take a 50-person team where each person costs $176,000 a year, counting pay, taxes, benefits and overhead. AI tools run $330 a person a year, training $9,600 a person once, and platform and security $100,000 a year. People lose 15% of their output for the first 3 months while they learn. Year one costs $926,500: $596,500 of spend and $330,000 of lost output. If AI saves 12.5% of each person's work time, the nine months after that save time worth $825,000, so year one comes out $101,500 behind. It pays back in about 13 months, and year two comes out about $983,500 ahead, because the training and the lost output don't repeat. Year one would break even at about 14% time saved. ## What the number leaves out It's an estimate, not financial advice. It counts the value of time saved only, and it holds every rate flat into year two. It doesn't include: - revenue from shipping more, sooner - fewer outages and faster recovery - the cost of managing the change, governing AI use and reviewing more code - tool prices or time saved that change after year one DORA's full calculator models the first two. Saved time is also only worth money if you put it to work, so decide where it goes. Then test your numbers on one of your real workflows, starting with time saved and how long the dip lasts. With the starting values, each extra month of dip costs about $202,000, and each point of time saved is worth about $66,000. ## How we keep the dip small The calculator puts your whole team through the dip at once, for three months. That's a big-bang rollout: everyone gets new tools on the same day and works out alone what to do with them. With the starting values, that dip costs $330,000, more than three times the $101,500 that year one comes out behind. We don't roll AI out that way. - **Start in one place.** We pick one workflow and the few people who own it. Only they go through the dip. Everyone else keeps working as they do today. - **See it working on day one.** In one focused day we build a working prototype of that workflow with you, so the learning starts right away and the dip stays short. - **Learn while it's cheap.** The first workflow is an experiment. We measure the time it really saves and fix what doesn't work while the stakes are small. Your measured number replaces the calculator's default. - **Then speed up.** Each next workflow lands faster, because the setup is in place and your people have seen one work. Confidence comes from a result, not a training course. There's a trade-off, and a CFO will spot it: starting small means the rest of the team gets its gains later. What you get in return is a dip that costs a fraction of the starting case, and a real time-saved number before you commit the full budget. The starting case commits $926,500, $596,500 of it in spend, before anyone knows whether 12.5% is true for your work. We'd rather you find out on one workflow first. We don't model this in the calculator. We'll publish the numbers once we can show them from client work. [How it works](https://spliceops.lrgfrmt.com/how-it-works) ## Questions about AI ROI ### How do you calculate the ROI of AI? Add up what the rollout costs in year one: tools, training, platform and security, and the output lost while people learn. Then work out the value of the time AI saves: people × cost per person × the share of work time AI saves, counted only for the months after people have learned. Net benefit is that value minus the cost, and ROI is net benefit divided by the cost. [Run the numbers](https://spliceops.lrgfrmt.com/tools/ai-roi-calculator). ### Is time saved by AI real money? Only if you use it. Freed hours become money when they go into work you'd otherwise hire for, ship later or not do at all. If nobody decides where the time goes, it disappears into the week. Name the work it goes to before you count it. ### How much time does AI really save? It depends on the work, so measure one workflow before you commit. Our calculator starts at 12.5% of each person's work time, the default in DORA's AI ROI calculator. Treat that as a starting point, not a promise, and use your own number once you have it. ### Why does AI make teams slower before they get faster? People need time to learn new tools and change how they work, so output dips first. That dip is a real cost, and it belongs in the business case. Keep it short by starting with one workflow and the people who own it. ### How do you keep the productivity dip small? Don't put everyone through it at once. Start with one workflow and the few people who own it, build a working prototype of it in one focused day, and measure what it really saves. Then roll out to the next team, faster each time. The rest of the team gets its gains later, but you find out what AI saves before you commit the full budget. ### How long does AI take to pay back? Count the months until the time AI saves has covered the year-one cost. Nothing is saved while people are still learning, and from year two the tools and platform costs come back, so take those off. With our calculator's starting values, a 50-person team pays back in about 13 months. Your own numbers will move that a lot, which is why we test them on a real workflow. More answers in the [FAQ](https://spliceops.lrgfrmt.com/faq). ## Bring your business case to a working session We'll test your numbers on one of your real workflows and tell you what it would take to hit them. **Book a working session** → /book A paid day, credited in full if you go on to work with us.