KYC Screener
Parses onboarding docs, runs the rules engine, packages escalations for compliance.
Financial services
Anthropic shipped 10 financial-services agents on May 5. We deploy them in your production environment in 6 weeks.
Polystat is an applied AI engineering firm. We productionize Anthropic's open-source agent templates at banks, asset managers, insurers, and fund administrators. Six weeks. Fixed price from $60,000. Founder-led. One specific workflow at a time.
Book a call New York / Chicago / Miami
01 / The news
May 5, 2026 — Anthropic announcement
Anthropic released ten production-grade Claude agents for financial services. Microsoft 365 integration across Excel, PowerPoint, Word, and Outlook. Moody's, Dun & Bradstreet, Verisk, Guidepoint, and four other data partners launched MCP connectors at the same time. Citadel, BNY, Carlyle, FIS, Mizuho, Travelers, Walleye, and Hg are quoted as customers. The templates are Apache-licensed and open. Read the announcement →
If you read the Anthropic announcement and thought "great, but who actually deploys this for us at our scale?" — this page is about that. We wrote a detailed thesis on the six gaps between an Anthropic template and a production tool at your firm: read the essay
The ten agents we deploy.
02 / What shipsAnthropic ships them as starting-point templates. We productionize them against YOUR data, YOUR workflows, YOUR compliance regime, in six weeks fixed price.
GL Reconciler
Finds breaks, traces root cause, routes for sign-off. Reconciles GL accounts and NAV.
Month-End Closer
Runs the close checklist, prepares journal entries, produces variance commentary.
Statement Auditor
Reviews financial statements and LP reporting for consistency and audit-readiness.
Valuation Reviewer
Ingests GP packages, runs valuation templates, stages LP reporting for review.
Earnings Reviewer
Reads transcripts and filings, updates models, flags thesis-relevant changes.
Pitch Builder
Comparable companies, precedent transactions, LBO scenarios, branded pitchbook end to end.
Meeting Preparer
Assembles client and counterparty briefing packs before every meeting.
Market Researcher
Sector landscapes, peer comps, idea shortlists. Synthesizes news and research.
Model Builder
DCF, LBO, 3-statement, comps — live inside Excel via the Microsoft 365 add-in.
How a six-week deployment runs.
03 / How it runsDiscovery.
Our engineer sits with your operators doing the workflow today. We map every step, time the friction points, and pick the agent template that fits. Compliance lead in the room.
Build + customize.
Production-grade code inside your environment. We customize Anthropic's template against your data sources, your chart of accounts, your escalation rules, your audit log format. Weekly demos.
Eval + adoption.
Build the eval set from real production data. Pair with operators during real workflow time. Tool ships when your team is using it without us in the room.
What an in-house team would otherwise hit.
04 / The six gapsFrom our essay on the gap between template and production. Each of these is solved engineering by week six.
01
Data connector paywalls.
Moody's, FactSet, S&P Capital IQ, PitchBook are gated. We map your existing subscriptions to required connectors before the build starts.
02
Firm-specific workflows.
Anthropic's GL Reconciler runs a generic close checklist. Yours has a specific chart of accounts, your reinsurance accruals, your fund-admin conventions. We encode them.
03
Adoption pairing.
Most internal AI tools die at deployment. We sit with your operators in weeks six through eight. The deliverable is your team using the tool, not the code.
04
Eval set construction.
We assemble 30-50 representative cases from your real production data, label them, run the agent against them, iterate. You can safely change prompts and upgrade models.
05
Firm-specific compliance.
The agent's escalation logic, audit-trail format, and retention requirements need your compliance officer's input. We bring them into scope in week one.
06
System integration.
The unsexy 70%. Authentication, parsing, rate limits, retries, audit logging. We scope to one or two integration points so the pilot ships.
Firms we work with.
05 / CoverageRegional banks.
$5-50B asset banks. KYC, credit memos, regulatory filings, reconciliation.
Asset managers.
$1-25B AUM RIAs, family offices, wealth managers. Research, client briefs, portfolio analysis.
PE & private credit.
Lower-middle-market funds. Sourcing, screening, valuation review, IC memos.
Fund admin & insurance.
Fund administrators, specialty insurers. GL reconciliation, NAV tie-out, statement audit, claims triage.
Why Polystat.
06 / AboutFounder-led delivery.
The person you talk to is the person who ships. Tomasz Bogus, founder, spent ten years running a software engineering consultancy in deep tech (Kafka, Scala, big data) for enterprises. Now applied to Claude.
Senior teams, built per engagement.
A typical pilot: one senior applied AI engineer (8+ years engineering, 2+ years shipping production LLM systems), Tomasz as founder/PM. Specialists added per scope. No junior bench, no sales floor.
Built for one tool at a time.
We ship one production AI tool per engagement. Six weeks. Fixed price from $60,000. Two weeks of adoption support after launch. Your team using it is the deliverable, not the code.
Pick one workflow. We ship the tool that absorbs it.
Tell us which Anthropic agent you'd want first, or describe the workflow that bleeds the most analyst hours. 30 minutes is enough to know if we're a fit.
Book a 30-minute conversation