Let your team use AI on real customer data, safely
Rig gives every person and every AI agent its own permissions on your warehouse, hides the fields they should not see, and logs every question asked. Fraud and disputes get their case context in minutes instead of an hour, and nobody has to hand out database credentials to make it happen.
// In plain terms
Rig is AI-native data infrastructure. Think Stripe or Vercel, but for data.
We connect all your sources, build a layer on top of them that knows what every table means, and let your team and your AI tools reach that data safely. For a fintech that means fraud, disputes, ops and finance can each ask the warehouse a question directly, see only the fields their role allows, and leave an audit trail behind every one.
// What you'd build
What a fintech builds first
Fintechs were already careful about who could reach customer data. What changed is that the people who cannot get access now have Claude and ChatGPT open in another tab, and pasting data into one of them is quicker than asking.
Dispute and chargeback evidence packs
Every case assembled in one pass: the transaction and its history, the customer's prior disputes, the support contacts around it, and the device and session context. This is the job that took Cleo's analysts 30 to 60 minutes per case before anyone could decide anything.
Your warehouse · Stripe · Zendesk · Intercom
KYC and onboarding review queue
Verification outcomes ordered by risk, so analysts spend the day on the cases that need judgment.
Sumsub · Your warehouse
Approval-rate diagnostics
Why approvals moved last week, cut by segment, cohort and channel, against the rule and model changes that shipped in the same window. Every payments and lending team asks this, and few can answer it the same day.
Your warehouse
Repayment risk and limit alerts
Cohorts drifting off plan before they reach collections, and limit changes surfaced as alerts the day they happen. Cleo's finance and disputes teams built both.
Your warehouse · Stripe
Fraud rule performance and false positives
What each rule catches, what it costs in good customers declined, and which rules no longer catch anything.
Your warehouse
Support quality and escalation patterns
Every conversation reviewed rather than a 2% sample, with escalation patterns and weekly summaries produced automatically. Cleo runs conversation-quality reviews and escalation tracking this way.
Zendesk · Intercom · Front
Channel and campaign performance
Spend against the accounts it acquired and what those accounts went on to be worth, rather than against platform-reported conversions. Cleo built top-performing ad detection and channel reviews on this.
Meta Ads · Google Ads · Your warehouse
Per-agent access and audit
Each person and each AI agent scoped to its own role, queries running in a sandbox, PII masked by policy, and a full log of who asked what. Without this, nothing above it is allowed near customer data.
Rig MCP · Claude · ChatGPT · Cursor
// In production
"Every team has a Rig agent now. Ops, support, fraud, marketing. We stopped being the bottleneck."
Cleo, consumer fintech, Series C
What is not in place yet
There is no fintech connector suite. We connect the commercial stack you run around the product, plus Sumsub for KYC, but there is no core banking, ledger, card issuing or lending connector in the catalogue. Rig's place in a fintech is on top of the warehouse your product data already lands in, which is how Cleo runs it.
On compliance: Rig holds SOC 2 Type II, and we do not hold ISO 27001. PII masking is applied by policy, not on by default, so classifying columns and setting those policies is part of onboarding.
// Proof
Cleo runs on Rig, and so do these
Cleo is the fintech, running Rig across five functions on its own warehouse. It is the only one so far, so the rest of the wall is labelled with what each company does.
Consumer fintech
Sustainable oral care
Homecare software
Online education
Workplace management
Hospitality techCommon questions
No. The catalogue covers the commercial stack a fintech runs (Stripe, CRM, support, ads, finance tools) plus Sumsub for KYC. There is no core banking, card issuing or lending connector. That is not usually the blocker, because your product data already lands in your own warehouse and Rig sits on top of that, which is how Cleo runs.
Rig holds SOC 2 Type II. We do not hold ISO 27001 or a financial services accreditation. Data can stay in your own warehouse so it never moves to us, or you can use EU-based Rig hosting. The security page carries the current status, and we can walk your risk team through it on a call.
Access is role based per person and per agent, queries run in a sandbox rather than against the raw database, PII columns can be classified and masked by policy, and every query is logged for audit. Masking is applied by policy, not on by default, so classifying your PII columns and setting those policies is part of onboarding.
At Cleo, a consumer fintech at Series C, fraud and dispute analysts had been spending 30 to 60 minutes assembling context per case before they could decide anything. Rig built a self-healing context layer over a 4,000-table warehouse, and the rollout went from a fraud-team pilot to org-wide in 16 weeks, with seats going from 37 to 204 and five functions building on it.
No, it removes them from the request queue. Cleo's own framing was going from three analysts automating data documentation to roughly 200 business users automating data-heavy work. The data team still owns the definitions and the access policy, which is the part that needs judgment.
That is close to Cleo's exact number, and Novakid runs 5,000+. The context layer is generated from the warehouse rather than hand-written, and it re-derives as the schema drifts, which is the only reason a warehouse that size stays documented at all.
Put permissions between your AI tools and your warehouse
Bring your warehouse and your access rules. We'll show you what each role sees, and what the audit log records.