Roiva
Product
Customer Support Deflection and the capacity it frees
Sales Tie AI tools to closed-won revenue
Finance Audit-grade ROI for the CFO
Operations Run AI initiatives like a portfolio
Engineering What AI coding assistants give back
Integrations Pricing Start free
AI ROI for Engineering

Price what your AI coding assistants actually give back.

Roiva counts the suggestions and edits your developers accept, prices them at a figure you set, and sets them beside Copilot's seats and Claude Code's spend.

Start free Or get your AI ROI exposure report →
Measure up to 3 initiatives free No card, no expiry

The problem

Seat counts and acceptance rates show your engineers use AI. They don't say what that's worth, and an hours-saved estimate isn't money your CFO can bank.

Worked example

One quarter of Copilot and Claude Code, line by line

Illustrative figures, not a customer's: 120 engineers on GitHub Copilot, 40 of them also using Claude Code.

  1. 1

    Count what developers accepted

    Copilot's acceptances come from GitHub's Copilot report, Claude Code's accepted edits from its analytics. They're different units: one Copilot acceptance can bundle several completions, so each is counted and priced on its own.

    Copilot acceptances, from GitHub 54,000
    Claude Code edits accepted, from Anthropic 18,000
    Engineers using either 120
  2. 2

    Price each at a figure you set

    Nothing published measures the minutes one acceptance saves, so Roiva's convention is 1 minute, marked low confidence. Your engineering lead set their own figures with the reasoning on record, and each entry says whose figure it used.

    Minutes per Copilot acceptance (Roiva's convention: 1) 1.5
    Minutes per Claude Code edit 3
    Developer cost per hour $95
    Freed capacity $213,750
  3. 3

    Check it against Jira

    Throughput is the same engineers' time measured another way, so it sits beside the acceptances rather than being added to them. If acceptances climb and throughput doesn't, that's the conversation the next review should have.

    Issues resolved, the quarter before and this quarter 2,600 → 2,900
    Average cycle time, days 6.1 → 5.2
    Added to value No
  4. 4

    Count as cash only what a document shows

    Your engineers are paid the same, so the hours aren't cash. Contract developers you stopped paying for are, recorded for each month with the contract cited, and never more than the capacity approved for that month.

    Freed capacity, step 2 $213,750
    2 contract developers not renewed, contract cited $96,000
    Capacity not yet cash $117,750
  5. 5

    Set it against everything the AI costs

    GitHub's plan seats and the rest of its invoice aren't counted: only Copilot is the initiative's cost.

    Copilot Business, 120 seats, from GitHub's bill $6,840
    Claude spend for Claude Code, from Anthropic's cost report $21,000
    Enablement and rollout time: 80 hours × $95 $7,600
    Total cost $35,440
Cash value
$96,000
Cost
$35,440
Cash ROI
171%
Capacity not yet cash
$117,750
Beside the ROI, not in it

Priced as savings, the same quarter reads 503%. The difference is engineering time no budget has shown yet.

None of it counts until a person approves it, and each figure says whether it rests on your minutes or on Roiva's convention. How Roiva calculates value →

What a reviewer sees in Roiva

September, one month of the quarter above, as Roiva calculates it: the formula, Copilot's acceptances and Claude Code's edits with what's behind them, and each minutes figure and the rate with who set them and when.

The formula inputs of a Roiva entry for Developer Capacity Freed by GitHub Copilot and Claude Code: Copilot acceptances and Claude Code accepted edits with their sources, the developer hourly rate, and the minutes saved per acceptance and per edit, each with who set it and when.
One month of the quarter above as a calculated entry in Roiva's demo account, a fictional company.

Example initiatives

What teams are tracking with Roiva

The AI Coding Assistant Rollout template ships with a formula for each way an assistant reports its work. Use the one your tools support.

Copilot and Claude Code together

Copilot acceptances × minutes saved per Copilot acceptance, plus Claude Code's accepted edits × minutes saved per edit, at your developer rate. Each is priced in its own unit, because a Copilot acceptance can bundle several completions.

Claude Code on Claude Enterprise

The same, with Claude Code's accepted edits read across your Claude Enterprise organization, where the seats are, rather than from API workspaces.

OpenAI Codex, by lines kept

Where an assistant reports no acceptance count, the lines of code it wrote that developers kept × minutes saved per line. The weaker measure, and the formula says so.

Issue throughput from Jira

Issues resolved, cycle time and story points against the months before rollout. Keep it as a cross-check beside the assistant's figures rather than pricing it too, or the same developer time is counted twice.

Connects to

Works with the systems you already run

The integrations most relevant to Engineering initiatives, and what Roiva does with each one.

GitHub

Roiva prices the time GitHub Copilot gives your developers back and books Copilot's seats from GitHub's bill, so an AI coding rollout shows both sides.

Anthropic, Claude Enterprise

Claude Code's accepted edits, from Anthropic's Admin API or across your Claude Enterprise organization, and Anthropic's Claude spend by workspace.

OpenAI Codex

Roiva prices the developer time OpenAI Codex saves from the lines of code it wrote that your developers kept, read from ChatGPT Enterprise.

Jira

Roiva tracks the Jira throughput behind an AI coding rollout, issues resolved, cycle time and story points, against the months before it, beside what it costs.

See all 30+ AI ROI integrations

Common questions

AI ROI for Engineering — FAQ

How do I measure ROI from GitHub Copilot or Claude Code?
Roiva counts what developers accept, from GitHub's Copilot report and Claude Code's analytics, and prices it: acceptances × the minutes one saves × your developer hourly rate. That is freed capacity: your engineers are paid the same, so it is reported beside cash ROI and never in it. Copilot's seats come from GitHub's bill and Claude's spend from Anthropic's cost report, as costs on the initiative. The time becomes cash only when someone records what changed for a month, such as a contractor not renewed or a hire not made, and cites the document that shows it.
Where does the minutes-saved-per-acceptance figure come from?
From you, ideally. Roiva ships a convention of one minute per accepted suggestion or edit, marked low confidence, because nothing published measures it: GitHub and others report task-level speedups, not minutes per acceptance. Set your own figure under Method › Assumptions with the reasoning behind it. Until you do, every entry that uses Roiva's figure says so, and ROI › Provisional lists what rests on it. A time study times tasks rather than acceptances, so it doesn't set this figure.
We use both Copilot and Claude Code. Will Roiva count the same work twice?
Not within one formula. Pick the formula for both, which prices Copilot acceptances and Claude Code's edits each in its own unit and adds them. A formula for one assistant reads acceptances from the connection you make its source, so two connections never feed it the same count. Keep Jira's throughput as a cross-check: it measures the same engineers' time, so pricing it with its own formula as well would count that time twice.
Why not just count lines of code?
Because a hard line and an easy one count the same, and a long generated file isn't a long saving. Roiva uses lines kept only where an assistant reports nothing better, such as OpenAI Codex from ChatGPT Enterprise, at a convention of half a minute a line that you can replace.
Does Roiva prove the assistant made us faster?
No, and it doesn't claim to. It books the before-and-after arithmetic for a person to approve: acceptances priced at a figure you set, and Jira's issues resolved and cycle time against the months before rollout. It runs no experiments and makes no causal claim. What it gives you is the figures and their sources, so a reader can judge the claim.

Measure your first AI initiatives in Roiva, free.

Measure up to 3 initiatives at a time on your own data, list every idea you're weighing, and keep it all when you measure more. No card, no expiry.

Start free → Get your exposure report