AI spend decisionsbefore the bill arrives.
Every AI dollar gets an owner. Spending stops at the limit you set. Decisions take minutes — and you get the receipt.
AI spend record · today
livegpt-4o → gpt-4o-mini on support traffic
policy: downgrade-rule · non-critical
contract-analysis hit its limit — $4,120 prevented this month
policy: budget-limit · legal-tech monthly
+$2,500 increase · owner notified where she works
Sarah Chen · one tap · on the record
Every request evaluated against your policies — and recorded
One control surface across every provider
Dashboards show the past. These are decisions.
Every AI dollar has an owner and an explanation.
By feature, by team, by customer — with a named owner, a threshold, and an approval path on every budget. When the bill doubles, you answer in one click, not one week.
- Cost attribution down to the feature that caused it
- Named owners — nobody's spend is anonymous
- Chargeback and showback your finance stack can ingest
SELECT * FROM caphound_spend WHERE charge_period = yesterday()your warehouse| feature | team | customer | BilledCost | enforced |
|---|---|---|---|---|
| contract-analysis | legal-tech | clearpath-legal | $1,284.42 | 247 |
| customer-support | product-eng | nexus-tech | $612.18 | 0 |
| document-intelligence | legal-tech | clearpath-legal | $498.07 | 61 |
| claude-assistant | platform | — | $396.55 | 12 |
| semantic-search | growth | meridian-co | $203.91 | 0 |
FOCUS 1.0 conformant · x_caphound_* extension columns · pushed nightly
Spending stops at the limit you set — before the bill.
Not an alert about an overrun. The overrun doesn't happen. Five policy types — budget limits, model downgrades, routing preferences, model restrictions, rate limits — evaluated on every single request, each decision recorded with the policy that made it.
- The runaway loop gets stopped, not reported
- Non-critical traffic moves to cheaper models automatically
- Routing keeps requests compliant — region, provider, approved models
- Every record names its policy — what, why, and how much it saved
decision-streampolicy:any · env:productiongpt-4o · contract-analysis
budget-limit · legal-tech monthly
gpt-4o → gpt-4o-mini · customer-support
downgrade-rule · non-critical traffic
claude-sonnet · doc-extract
routing-preference · EU data residency
o1-preview · engineering-sandbox
model-restriction · approved models only
batch-embed · growth-experiments
rate-limit · 20 req/min non-prod
Every request, evaluated against your policies
$2,144 never spent this month
Decisions take minutes, not meetings.
A budget needs an increase at 6pm. The owner approves it with one tap in Slack — or one click in email. Rejections ask for a reason. Temporary exceptions revert themselves. Nothing waits for Monday.
- One-tap approve / reject, right where the owner already is
- First valid decision wins — everyone else is notified
- Exceptions expire on schedule; no forgotten overrides
CapHoundapp6:04 PM
Budget increase requested — contract-analysis
dev@acme.com requests +$2,500 on the contract-analysis budget.
“Quarter-end document volume — need extra capacity through Friday.”
Approved by Sarah Chen · one tap · 6:06 PM · on the record
A receipt, not a report.
Once a month: what was prevented, what was decided, who decided it, and what reverted on schedule. Every figure is a recorded event — never an estimate. Forward it upward without a meeting.
- Spend prevented is the real value of stopped requests
- Every decision carries its decider, channel, and timestamp
- Backed by an immutable audit trail
From: CapHound
Controls Receipt — May 2026: $12,840 prevented, 14 decisions
Prevented
$12,840
Decisions
14
Median
38 min
Controls applied
Budget limit raised to $17,700
maya.patel · $2,500 governed
Exception applied · auto-reverted on schedule
cfo@ · $500 capped
247 requests blocked at limit — contract-analysis
policy · $4,120 prevented
Every figure is a recorded event — backed by the immutable audit trail.
Works where your teams already work
Warehouse export is FOCUS 1.0 conformant — loads into any FinOps pipeline unmodified.
“I watched AI spend explode across the industry with none of the rigor we'd apply to any other production system. AI spend should be owned, approved, and controlled — not discovered on an invoice.”
Amarinder Singh · Founder, CapHound · 15 years in engineering