Counterparty risk management with every rating driver cited to the bank's own filing, in its own language
Every bank and broker on a current review, each rating driver cited to its own page.
See one case, screen by screen ↓Counterparty risk management is the due diligence, rating, limits and monitoring a firm applies to the banks, broker-dealers, futures brokers and prime brokers it trades with, so that a counterparty's failure does not become its own loss. Counterparty Risk Management gives an asset manager's credit team seven agents that read each counterparty's own filings in any language, propose a cited rating and limit, challenge it against peers and watch between reviews, while the credit officer decides and the CRO co-signs above $50M.
AI in asset management: a buyer's guide to agentic AI for investment managers →
Updated
An asset manager trades with hundreds of banks, broker-dealers, futures commission merchants and prime brokers, and each counterparty credit risk assessment — the review, an internal rating and a credit limit — must be redone at least once a year, with the name watched in between. Today the counterparty credit analyst downloads the annual report — often published only in the bank's own language — retypes a hundred-plus lines into the spreading template, scores it against the firm's scorecard, argues the rating in a memo and walks it to a credit officer.
The risk sits where nobody can trace or nobody is reading. A notch change argued in a memo, with the figures in a spreadsheet that never links to the page, is a rating internal audit cannot re-perform. Between reviews, news, regulator notices, rating actions and bond spread moves arrive every day, and a twelve-month cycle misses them unless someone watches every name. When one family office collapsed in 2021, a single prime broker lost about $5.5 billion, and the report found warning signs that were seen but not escalated. Supervisors expect more: the Basel Committee's 2024 guidelines for counterparty credit risk management cover due diligence, metrics and limits, and governance, and say financial statements alone may be insufficient.
Estimated split of analyst hours for one annual bank review, by hand and with the solution.
Seven agents read, rate, challenge and watch every counterparty; the credit officer decides and the CRO co-signs above $50M.
Quillfield Capital reviews 412 banks and brokers a year. Selvira Yatırım Bankası, an Istanbul investment bank that four of its funds face in FX forwards and reverse repo, publishes its annual report only in Turkish — and its annual review is due Oct 30. Here is what happens, in the working solution.
Farah opens to seven decisions awaiting her, 38 annual reviews due in 30 days, 14 counterparties on the watchlist and two limit excesses, both with a cure plan. Overnight the agents prepared 11 reviews and read 64 new filings. The agents' first line tells her the one decision that matters most today: Selvira's capital fell to 11.4% and the agents agree on one notch down.
“Ask me anything about a counterparty, a limit or a filing — I answer with sources, and I act when you tell me to.”
The Filing Intake agent collected the 2025 Faaliyet Raporu — 248 pages, Turkish only — from the bank's investor relations site when the review window opened. The Spreading agent read it with the Pillar 3 disclosure and the BRSA audited statements into Quillfield's 184-line template, FY2024 beside FY2025, the Turkish label beside each line, at USD 1 = TRY 41.20. Three lines are flagged for the analyst, among them a restated FY2024 NPL ratio.
The Spreading agent never estimates a missing figure; restatements and reclassifications go to the analyst.
Any figure opens its source. Page 142 of the annual report shows the Sermaye Yeterliliği table in Turkish, with the CET1 ratio row highlighted — 13,1 to 11,4 — and the English reading beside it. The note says where the document came from and when it was received.
“The highlighted figure is the one the agents used.”
Scorecard v4.2 gives 54.6, down from 61.2 — the QR14 band. On the rating notch ladder Selvira sits on QR13 with the proposed rung QR14 beside it, and each driver is an arrow sized by its push and weight: capital −0.55, liquidity −0.25, asset quality −0.30, profitability +0.10, market signals −0.15, a net −1.15 notches at confidence 0.91. The Limit agent reads the grid — $25M for a bank at QR14, below the $32.2M capital test of 2% of Tier 1 — and proposes $40M → $25M against $18.7M of exposure across four funds, with UCITS headroom checked fund by fund.
“No trade needs to be unwound. Below $50M, so the credit officer approves alone.”
The Peer agent re-rates every proposal from banks of similar size and business model — here three Turkish and three Central European banks. Its peer-implied rating is QR14, so the difference is zero notches and nothing escalates. Selvira's CET1 of 11.4% is the lowest of the seven; the peer median is 14.3%. Asked, it re-runs without the development bank in the set and still finds QR14.
Any disagreement of more than one notch goes to the credit officer, with both arguments.
The Watch agent has added that Selvira postponed a $500M Eurobond issue, from a Turkish news item, and that its USD 2029 bond spread widened 86 bps to 498. Daniel Osei, the analyst, accepted the restated NPL ratio and submitted on Friday. Farah asks why liquidity counts against a bank whose LCR is well above 100%. The answer: LCR fell from 168% to 142% and NSFR from 121% to 114%, the steepest fall among the seven peers; the scorecard scores the one-year change, and it costs 0.25 notch.
The approval states what she signs — QR13 → QR14, limit $40M → $25M, prepared by Daniel Osei, with Farah as the checker. She picks reason code APR-01 “Proposal supported by the evidence” and confirms she has reviewed the cited evidence. Only then are QR14 and the $25M limit written to the counterparty master and the limit system, Kenji Watanabe on the FX desk is told, with $6.3M of headroom, and the next review is set for Apr 19, 2027 — six months, as Selvira is on the watchlist.
“Limit at or below $50M — your signature completes it.”
For Banco Ribeira do Sul in São Paulo, the Rating agent proposes one notch up, QR12 → QR11, from a Portuguese annual report. The Peer agent disagrees: against six Brazilian mid-size banks it implies QR13, because the top 20 depositors hold 38% of deposits against a peer median of 21%. Two notches apart, it escalated to Farah with both arguments, and the analyst Sofia Reyes recommends keeping QR12. Farah chooses the rating and gives her reason.
Asked why the two agents disagree: “Scorecard v4.2 has no funding-concentration driver; it is on the list for v4.3.”
The same counterparty review, seen by five of the people who carry it — what their year of reviews looked like, and what it looks like now.
Seven agents read, rate, challenge and watch every counterparty; the credit officer decides and the CRO co-signs above $50M.
Collects annual reports, Pillar 3 disclosures, FOCUS and 1-FR-FCM reports and audited statements from investor relations sites, regulator sites, document management and the data vendor, and detects each file's language, period and accounting basis.
Reads each filing in its original language into the firm's 184-line spreading template, two years side by side, and cites every figure to its page with the original passage beside the English.
Applies the firm's scorecard to propose a rating notch with five weighted drivers — capital, liquidity, asset quality, profitability and market signals — a score, a confidence and the country ceiling.
Builds a peer set of similar size and business model and re-rates the counterparty from it, as a challenge to every Rating agent proposal.
Turns the notch into a limit from the grid and the capital test, checks live exposure by product and fund and UCITS headroom, and drafts a cure plan for every limit excess.
Reads news, regulator notices, filings, rating actions and bond and equity moves for every counterparty between reviews, in the language they are published in, and proposes what to do.
Drafts the credit memo from the review and the monthly counterparty risk report for the Risk Committee, every sentence cited.
Approves; the CRO co-signs above $50M. The agents propose; a named person decides.
Ask about any counterparty, limit or filing — or tell the agents what to do. They answer with sources, and act only when you say so.
Why does liquidity count against them?
LCR is still above the 100% minimum, but it fell from 168% to 142% and NSFR from 121% to 114% — the steepest fall among the seven peers. Scorecard v4.2 scores the level and the one-year change; the change costs 0.25 notch, the level costs nothing.
What if we keep the limit at $40M?
At QR14 the policy maximum for a bank is $25M. Keeping $40M is a policy exception: it needs reason code EXC-02 and the CRO's signature. UCITS headroom would still be fine (0.43% of assets), but I would not recommend it: capital is the weakest of the peer set.
Which counterparties got worse this month?
Four, by the evidence: Selvira Yatırım Bankası — CET1 13.1% → 11.4%, downgrade to QR14 proposed; Tianrui Securities — a regulator warning letter on margin-lending controls; Castellane Futures — adjusted net capital down 22% in September; Hoshida Trust Bank — bond losses doubled to ¥184.2B.
Who can approve a $90M limit?
Two people must sign: a credit officer (Farah Malik) and the CRO (Martin Ostrowski), because it is above $50M. The analyst who prepared it cannot sign, and no desk can set a limit. Albury Northern Bank's $75M → $90M is waiting for the CRO now.
13 screens from the working solution, on its sample data. Pick one to see it large.
The scorecard, thresholds, limit rules and roles every counterparty risk assessment runs on are settings, versioned, and every change needs a second approver.
| Setting | Default | Choose from |
|---|---|---|
| Scorecard driver weights | Capital 30% · Liquidity 20% · Asset quality 20% · Profitability 15% · Market signals 15% | Your weights, totaling 100% · saved as a new scorecard draft for approval |
| Agent confidence below which no rating is proposed | 0.85 | Your threshold · below it the analyst rates without a proposal |
| Peer disagreement that escalates to the credit officer | More than 1 notch | Your escalation rule |
| Watch trigger: capital fall in one month | 15% | Your trigger · watch signals rated high, medium or low |
| Capital test | Banks: limit ≤ 2% of Tier 1 capital · broker-dealers: ≤ 10% of excess net capital | Alongside the limit grid by notch and counterparty type |
| Approvals | Credit officer up to $50M · CRO co-signs above | Policy exception above the grid: reason EXC-02 and the CRO |
| UCITS checks and temporary excess | 10% of assets per credit institution · 5% other · 20% combined · excess up to 10 business days | Checked fund by fund against fund administrator NAVs |
| Data residency and retention | EU (Frankfurt) · US (Virginia) · Singapore · 7 years after the relationship ends | Legal hold on or off · desks see only the counterparties they trade with |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample data · “target” = the design goal, measured in the live solution · “estimated” = our estimate · “typical” = published figures (the 2021 Paul Weiss report on prime-broker losses and its press coverage). Regulations named for context: Basel Committee Guidelines for counterparty credit risk management (Dec 2024), Federal Reserve SR 21-19, SEC Rule 18f-4, UCITS Article 52, SEC Rule 204-2. People, firms, funds and counterparties named on this page are fictional — characters and sample data in the working solution.
Counterparty risk management is the due diligence, rating, limits and monitoring a firm applies to the banks, broker-dealers, futures brokers and prime brokers it trades with, so that a counterparty's failure does not become its own loss. Counterparty Risk Management gives an asset manager's credit team seven agents that read each counterparty's own filings in any language, propose a cited rating and limit, challenge it against peers and watch between reviews, while the credit officer decides and the CRO co-signs above $50M.
The Filing Intake agent collects the annual report, Pillar 3 disclosure and audited statements in their original language, and the Spreading agent reads them into one template with every figure pinned to its page. Any figure opens the original page beside the English reading, so counterparty due diligence no longer depends on a translated or standardized dataset.
The Limit agent turns the proposed notch into a limit from your grid and the capital test — 2% of Tier 1 for banks, 10% of excess net capital for broker-dealers — and checks it against live exposure by product and fund and against UCITS Article 52 headroom. A credit officer approves up to $50M, the CRO co-signs above, and the limit system is updated only after the signature.
The Peer agent re-rates every proposal from similar banks. If it disagrees with the Rating agent by more than one notch, it escalates to the credit officer with both arguments side by side, and she chooses the rating with a reason code.
The Watch agent reads news, regulator notices, filings, rating actions and bond and equity moves for every counterparty in the language they are published in. Each signal comes with an English reading and a proposed action — a watch, an off-cycle review or a limit hold — that a credit officer approves or dismisses with a reason.
Yes. The audit trail records every agent and human step with who, what, when, why and the evidence, cannot be edited and is kept for seven years. One export builds an evidence pack — decisions with reason codes and signatures, spreads with source pages, original filings, agent steps and versions, peer sets and the policy in force on each date.
Yes. The agents propose; they never set a final rating, never sign and never write to the limit system before a signed approval. The analyst prepares, a credit officer approves, the CRO co-signs above $50M, nobody approves their own work and trading desks can only see limits.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your scorecard, limit grid, approval rules and a sample of past reviews — and it goes live once every quality gate has passed. We will walk you through it on your own counterparties first.
Every holding bucketed monthly, every override explained, no filing trigger missed.
1-min demoLiquidity Stress TestingEach fund's breach day found and evidenced, and the quarterly pack signed in a day.
1-min demoDerivatives Risk Management ProgramEvery daily VaR test explained in plain words, every escalation signed, no late N-RN.
1-min demoWe'll run Counterparty Risk Management on a sample of your own banks and brokers and the filings they publish.