The question every engineering leader now gets
An engineer sends a one-line message: “Can you raise my AI budget to $1,500 a month?”
Most engineering leaders answer it blind. Spend sits in three consoles, one each for Cursor, Claude and Copilot. What the money produced sits somewhere else. So you say yes and hope, say no and slow a strong engineer down, or ask finance for a spreadsheet.
Today we’re launching Milestone Budget Alerts, in AI Spend Hub. Milestone posts a Slack alert when a developer or team gets close to their monthly AI budget for Cursor, Claude and Copilot. Ask Milestone in the thread whether to raise it. It recommends a budget and shows the math: spend, merged PRs and cost per merged PR against the team.
AI spend became a meter
AI spend used to be a seat line: a fixed price per developer, set once a year. Now it’s a meter. It runs per developer, per day, in several tools at once.
GitHub moved every Copilot plan to usage-based billing on June 1, 2026 (GitHub changelog, April 27, 2026). Anthropic puts the average enterprise Claude Code cost at about $13 per developer per active day, or $150–250 per developer per month (Claude Code docs, checked October 12, 2026).
So AI coding spend is now variable, per developer, and split across vendors. Three things break.
- You hear late. Each tool shows its own spend in its own console, and alerts an admin by email or in the product. No single budget covers all three, so you often hear when the invoices land.
- The raise request has no evidence. Spend alone can’t tell a developer who ships more with AI from one who runs long sessions that go nowhere.
- Hard caps stop work. A per-user cap protects the bill, but it stops the developer mid‑task. The people most likely to hit one are the heaviest users, and some of them are your most productive.
Spend tells you what something cost. It doesn't tell you what it produced.
Set. Alert. Answer.
Budget Alerts works in three steps, in the order you'll use them.
Set: Set AI budgets once.
One monthly budget per developer covers Cursor, Claude and Copilot together. Set an org default, override it for a team or a person, and choose when the warning fires. Getting close fires at your Warn at percentage, 80% by default. Reached fires at 100%.
Budgets track variable, billable spend. Seat and subscription costs don't count.
Budget Settings at Corvant: a $900 org default per developer, Warn at 80%. Illustrative data.Alert: Alert the people who decide.
When a developer or team crosses Getting close or Reached, Milestone posts one Slack message a day to each recipient. It covers everyone who crossed, with a link to the Budgets screen filtered to exactly those people.
Each crossing is announced once. Nothing is blocked, and nobody loses access to a tool by going over.
The daily alert in #eng-ai-spend: Priya Nair is Getting close, at 84%. Illustrative data.Answer: Get the answer, with the math.
Ask Milestone in the alert's thread what you'd ask finance. It reviews the developer's recent spend and output. The answer starts with a bottom line: the budget the data supports.
Then come the spend trend and projection, merged PRs and cost per merged PR against the team median, and one lever to lower cost. Milestone recommends a budget and shows the math. You decide, then make the change in Milestone.
The AI spend review in the thread. Bottom line: "Approve $1,200/mo, not $1,500." Illustrative data.
One raise request, start to finish
Here’s how it plays out at Corvant, a fictional company with 10 developers on AI budgets.
October 1. Daniel Brooks, an engineering leader at Corvant, sets the org default to $900 per developer per month. He moves Warn at from 70% to 80%. All crossings go to #eng-ai-spend, a private channel, and to Daniel. The Platform team gets its own rule, and every developer hears privately about their own budget.
October 22, 9:02 AM. Milestone posts in #eng-ai-spend: “1 new budget crossing today.” Priya Nair, a developer on the Platform team, is Getting close: 84%, or $756 of $900. She has already asked for $1,500 a month.
9:06 AM. Daniel replies in the thread: “@Milestone Priya asked to raise her budget to $1,500/mo. Should I approve?”
Milestone answers with an AI spend review of Priya’s last 90 days, July 25 to October 22. Its bottom line: “Approve $1,200/mo, not $1,500.” Four numbers back it.
- 84%Close to her budget. $756 of $900 on October 22, projected $1,065 by month end. Her monthly spend went from $590 in July to $861 in September, up 46%.
- 42She ships more. 42 merged PRs in 90 days. Platform team median: 29.
- $58She costs less per merged PR. $58 per merged PR. Team median: $66.
- $1,200Not $1,500. $1,200 covers her projected month with 13% headroom. At $1,500, her 80% warning wouldn't fire until $1,200 ($1,500 × 80%), so Daniel would lose the early signal.
The review names one lever, too. 61% of Priya’s spend is on premium models. Moving long refactor sessions to a standard model would save about $150 a month. The review ends with how it read each number, so Daniel can check the math.
Daniel opens Budget & Usage from the alert, filtered to Watch + Over, and sets Priya’s budget to $1,200 in Milestone. Milestone recommended. Daniel decided, with the data in hand.
$1,200Priya's new monthly budget, set in Milestone
63%her usage after the change: Healthy
$150/moabout what the one lever saves: long refactor sessions on a standard model
Corvant is a fictional company. Every figure in this example is illustrative, not customer data.
Who sees what
You choose who hears. There are three audiences:
- All crossings. A channel, named people, or both. They see every developer and team that crossed, including developers in no team.
- Specific teams or groups. Extra recipients, such as a team lead, told only about the teams or groups they’re mapped to.
- Each developer, privately. A private message about their own budget, and nobody else’s. It’s off by default. Tell developers before you turn it on.
A channel’s audience is its membership, including people who join later. And a question asked in a channel thread is answered in that thread, so the whole channel sees the answer. Start with a private channel with a deliberate membership.
Recipients never sign in. Milestone matches them to Slack by work email, and they don’t need a Milestone login. To ask Milestone in a thread, each person connects to the Milestone assistant once.
What Budget Alerts doesn't do
Know these before you roll it out.
- It doesn’t block anyone. These are alerts, not limits. Nothing is blocked, throttled or capped. If you want a hard stop, keep the vendor’s own cap as a backstop. Budget Alerts tells the budget owner before it hits.
- It covers three tools. Budgets count variable spend on Cursor, Claude and Copilot only. Other tools appear on the Spend screen, but in no budget.
- It’s Slack only. No email, Teams or in-app alerts in this release. One Milestone workspace connects to one Slack workspace, and Enterprise Grid isn’t supported.
- It’s daily. Milestone checks for crossings every hour and sends each recipient at most one message a day. A crossing found after today’s message waits for tomorrow’s.
Get started
Available now. Your Milestone team turns it on for your workspace. Then three steps get you running:
- Ask your Milestone team to enable it. It’s a switch for your workspace.
- Install the Milestone Slack app. In Milestone, go to Settings → Integrations → Slack → Add to Slack. A Slack admin approves the install.
- Choose who hears, then switch it on. In AI Spend → Budgets & Usage → Slack Alerts, pick the recipients, turn on alerts and press Save.
Switching it on doesn’t announce who is already over. Your first alert covers the next new crossing.
Fund the people who ship.
Book a demo. We'll walk one raise request through Milestone, from the alert to the decision.
Or watch the 51-second film →