Failed trade management where agents find the cause, price the miss and check the fix, and two people decide before each market's cut-off
Target 20–30% fewer fails, each fix cited and checked before the market’s cut-off.
See one case, screen by screen ↓Failed trade management — also called settlement fails management, or trade fails management — is the work of finding, fixing and accounting for trades that are unaffirmed, unmatched or failing to settle: reading the status, finding the cause, chasing the counterparty, re-instructing before the market's cut-off, and reconciling the penalties afterwards. Failed Trade Management gives settlement and trade support teams seven agents that link every message to its trade, name the cause with evidence, price a miss, and draft and check the fix, while an analyst approves and a second person releases.
AI in asset management: a buyer's guide to agentic AI for investment managers →
Updated
Failed trade management is a race against cut-offs. Since the US moved to T+1 under SEC Rule 15c6-1 on May 28, 2024, a settlement team has one evening to affirm a trade and one morning to fix it, and Europe, the UK and Switzerland move to T+1 on October 11, 2027. Every market and custodian has its own amendment deadline in its own time zone; once it passes, the trade fails for the day. Yet the work is still done by reading MT548 reason codes in custodian portals, opening the allegement, searching a shared mailbox for the counterparty's new instruction — often written in German, French or Japanese — checking the SSI record, re-keying the instruction and tracking who is chasing what in a spreadsheet.
The cost lands on the funds. Under CSDR settlement discipline every late business day draws a cash penalty — 1 basis point of the trade a day for liquid shares — charged through the custodian, and a sale that fails can leave a fund short of the cash for its next purchase. Wrong standing settlement instructions cause about one fail in five, the most preventable cause there is. And at month end someone still has to reconcile the penalty reports, dispute the errors and chase the credits brokers did not pass on.
Estimated split of one day's exception work for a desk of six settlement analysts, by hand and with the solution.
Seven agents link, explain, price, draft and check every fail; an analyst approves and a second person releases.
Friday, October 16, 2026, at Quillfield Capital, a global asset manager with settlement desks in New York, London and Singapore: 40,212 trades settle today in 31 markets, 183 exceptions are open and 37 need a person. Here is one of them — a EUR 11.8 million sale failing at Euroclear Bank — screen by screen, as the analyst works it with the agents.
Anika works the Euroclear, CREST and Clearstream lanes from London. Today shows 8 items waiting for her of the 14 exceptions assigned to her, 1 h 25 min to Euroclear Bank's 15:30 Brussels cut-off, 96 fails so far against 118 at this time yesterday, and EUR 186,420 of penalties avoided in October. Since 7 a.m. the agents have read 18,406 status messages, 2,214 allegements and 412 counterparty emails in 11 languages, and linked 99.98% of them to their trades.
Checker agent: “Re-checked 386 proposed fixes before they reached a person; stopped 9, including a quantity question on QF-88177.”
One lane per market — DTC, Euroclear Bank, CREST and Hong Kong — on one hours-from-now axis, each with a red line at its local cut-off and a clock in its own time zone. Every open exception sits where the agents expect to fix it, colored by cause and sized by the money at stake a day. 19 trades are projected to miss, EUR 1,418 a day at stake if nothing changes. Hong Kong has closed, and its lane is marked stale because the sub-custodian's file is 24 minutes late.
“Closed for today. 13 trades rolled to Monday, October 19. Chasers sent in Chinese and English.”
The sale of 410,000 Hallberg Industrie AG to Brennmoor Securities AG for the Quillfield European Equity Fund — EUR 11,840,800, about $13.7 million — is unmatched at Euroclear Bank, reason REAG. The Intake agent linked the MT548 status and Brennmoor's MT578 allegement. The Cause agent, at 0.97 confidence, names the reason: our instruction still carries Brennmoor's old account 23715, because the trade was enriched on October 14 at 10:12 and the SSI team verified the new account 41908 on October 15 at 16:20.
Cause agent: “Same cause on 3 more Brennmoor trades settling Monday.”
Every statement cites its source. Jana Kessler of Brennmoor wrote on October 14 at 08:31 Frankfurt: “Neues Konto: 41908 (bisher 23715)”. The Intake agent's English sits beside the original, with the same lines highlighted in both. The other tabs hold the status message, the allegement, the SSI record, our instruction, the penalty and the history.
The Intake agent translates every message and keeps the original beside the English.
The Penalty agent has priced the miss: EUR 1,184.08 a day — 1.0 basis point of EUR 11,840,800, liquid shares — or EUR 2,368 by Monday's close, charged to the fund through the custodian. The sale also funds Monday's Vantaa Grid purchase, so a miss would overdraw the fund by EUR 9,630,100. Anika types “What if Brennmoor doesn't match by 15:00?” and the agents answer with their sources: call Brennmoor's settlement desk on the number on file and escalate to Rafael Mendes at 15:00, half an hour before the cut-off.
Penalty agent on buy-ins: “Not applicable: mandatory buy-ins are not in force under CSDR.”
The Chaser agent drafted the cancel-and-replace of our MT543 — receiving-agent account 23715 to 41908, the only field that changes — and a short note to Brennmoor in German. The Checker agent passed it 6 of 6: the verified SSI, a match with Brennmoor's own instruction, a valid format, no sanctions hit, and quantity, amount and date unchanged. Above $5M the release is the supervisor's. Anika approves with the reason “Verified SSI — re-instruct”.
“You are the maker. You cannot also release this instruction.”
Partials, shapes, buy-in responses and anything above $5M wait in Rafael's queue, each with the rule that sent it there and the Checker agent's result. He releases QF-88213 with the reason “Checked against the evidence”; the SWIFT gateway sends the cancel and the new MT543 at 14:07:40, and the custodian acknowledges at 14:07:52. Nobody releases what they approved: Anika's own partial on QF-87952 reads “You approved this: someone else releases it”.
Rafael Mendes: “Nothing above five million leaves without my eyes on it.”
The custodian's status comes back matched at Euroclear Bank. The Penalty agent records EUR 1,184 a day avoided and no overdraft on Monday, and October's penalties avoided rise to EUR 187,604. The maker, the checker, the gateway and every agent step are on the trade's record, with the time and the reason.
Cause agent: “Three more Brennmoor trades settling Monday still carry account 23715. Shall I prepare the re-instructions?”
The Forecast agent scored Monday's 31,980 settlements on trade date: 214 score 0.60 or more. The top three are the Brennmoor trades still on account 23715 — QF-88240 at 0.96, QF-88231 and QF-88214 at 0.95. One approval prepares all three re-instructions: two go to a second analyst, Diego Ramos, and QF-88240, at EUR 5,917,200, to Rafael. Over the last 20 days, 38% of the trades it flagged failed, against 0.86% of all trades.
Forecast agent: “I propose pre-chases; I never send one on my own.”
Kaito Shoken's letter, in Japanese, offers 300,000 Kinsei Robotics shares on Monday and the remainder on Tuesday, and the Chaser agent drafted an acceptance of the partial. The Checker agent disagrees: our trade is 320,000 and Kaito's instruction is for 300,000, so the gap is a quantity mismatch (DQUA), not only short securities. Both arguments are on the screen, and the analyst chooses: accept the partial of 300,000, or query the quantity with Kaito.
Checker agent: “Accepting the partial would leave 20,000 unexplained … A person decides.”
The same afternoon, seen by five of the people who carry it — what their day looked like, and what it looks like now.
Seven agents link, explain, price, draft and check every fail; an analyst approves and a second person releases.
Reads settlement status (MT548 / sese.024), allegements (MT578 / sese.028), pending statements and penalty reports (MT537), affirmation reports, custodian fails files and counterparty emails and letters in any language, and links each one to its trade, keeping the original beside the English.
Scores tomorrow's settlements for fail risk on trade date, from positions and cash in portfolio accounting, each counterparty's settlement history, central matching status and the SSI change log, and says why in plain words.
Names the real reason for each fail — not just the status code — with cited evidence from the message history, the allegement, the email thread and the SSI master, and finds other trades with the same cause.
Prices the CSDR penalty, funding and overdraft cost and buy-in risk before the cut-off, reconciles daily and monthly penalty reports against its own calculation, and drafts appeals and credit claims.
Drafts the chaser in the counterparty's own language, from templates in 11 languages, and the amended instruction: a cancel-and-replace, a partial or a shape request.
Re-validates every fix before a person sees it: the verified SSI, the counterparty's allegement, format, sanctions, unchanged economics and the release rule. It runs on a separate model family from the agent that drafted the fix.
Plans the day backward from each market's cut-off, in its own time zone and holiday calendar, re-plans when a clock or a status moves, and escalates what will not make it.
Approve, release, verify SSIs. The agents propose; a named person decides.
Every working screen has the agents beside it. Ask about a trade, a market, a counterparty or a penalty, or tell them what to do: each answer cites its source, and anything that moves a position is only prepared for a person to confirm.
Why is QF-88213 failing?
QF-88213 is unmatched at Euroclear Bank because our instruction carries Brennmoor's old account 23715. Brennmoor moved to 41908 in a German email on October 14, and its own instruction already uses 41908. Our SSI master has had the verified account since Oct 15 at 16:20; this trade was enriched the day before.
Which trades will miss today's cut-offs?
19 trades are projected to miss today's cut-offs, EUR 1,418 a day of exposure: 6 at Euroclear Bank (cut-off 15:30 Brussels), 5 at CREST (15:00 London) and 8 at DTC (14:00 New York). I have highlighted them on the lanes.
Re-instruct every Brennmoor trade still on the old account
Three more Brennmoor trades settling Monday still carry account 23715: QF-88214 (EUR 2,633,600), QF-88231 (EUR 3,658,200) and QF-88240 (EUR 5,917,200). I have prepared all three re-instructions for your approval: two will go to a second analyst to release, and QF-88240, above $5M, to Rafael Mendes.
What did CSDR penalties cost our funds in October?
In October to date, CSDR penalties charged to Quillfield funds are EUR 41,880; credits received are EUR 29,310 of EUR 33,940 expected, so EUR 4,630 has not been passed on by two brokers. The Quillfield European Equity Fund bore the most: EUR 9,860 charged, EUR 2,980 after credits.
13 screens from the working solution, on its sample data. Pick one to see it large.
Thresholds, cut-offs, escalations and penalty rates are settings. The platform administrator proposes a change with a reason, and it takes effect only after a second administrator approves it.
| Setting | Default | Choose from |
|---|---|---|
| Second-analyst release | Amendments below $5M | Set by the firm in Thresholds and rules; held until a second administrator approves |
| Supervisor release | $5M and above; every partial settlement, shape and buy-in response | The amount is the firm's to set (the screen's example change: $5M to $4M) |
| Cause agent confidence floor | 0.85 — below it, the agent asks a person | Set per agent, with its version and last evaluation |
| Intake linking threshold | 0.95 — below it, the message waits for a person | Set per agent, with its version and last evaluation |
| Follow-up chaser | 20 minutes without a reply | Service levels and escalation |
| Escalation | To the supervisor 60 minutes before cut-off; to the Head of Trade Support 30 minutes before, above $5M | Service levels and escalation, with follow-the-sun ownership |
| Amendment cut-offs | DTC 14:00 New York · Euroclear Bank 15:30 Brussels · CREST 15:00 London · Hong Kong 14:30 | Per market, in local time, with its holiday calendar and penalty regime — 31 markets |
| CSDR penalty rates | Commission Delegated Regulation (EU) 2017/389: 1.0 bp a day for liquid shares | A rate change is a reviewed setting with a new version; each day is priced with the version in force |
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 a pilot · “estimated” = our estimate · “typical” = published figures and rates (DTCC equity trade volume insights — DTC fails 2.83% before T+1 and 3.01% in April 2026; Commission Delegated Regulation (EU) 2017/389 — CSDR cash penalty rates; TABB Group for DTCC — about 20% of fails from incorrect SSIs; SEC Release 34-96930 — T+1 under Rule 15c6-1, and Rule 15c6-2; ESMA final report on settlement discipline for T+1 — October 11, 2027). EUR figures are sample data. People, firms, funds, counterparties, issuers and identifiers named on this page are fictional — characters and sample data in the working solution.
Failed trade management — also called settlement fails management, or trade fails management — is the work of finding, fixing and accounting for trades that are unaffirmed, unmatched or failing to settle: reading the status, finding the cause, chasing the counterparty, re-instructing before the market's cut-off, and reconciling the penalties afterwards. Failed Trade Management gives settlement and trade support teams seven agents that link every message to its trade, name the cause with evidence, price a miss, and draft and check the fix, while an analyst approves and a second person releases.
The usual causes are an SSI mismatch, an unaffirmed or unmatched trade, short securities, short cash and other data errors such as a wrong amount or settlement date; incorrect standing settlement instructions alone cause about one fail in five. The Cause agent reads the status code, the counterparty's allegement and its emails or letters in any language, names the real reason in one sentence with its sources, and finds other open trades with the same cause.
Under Commission Delegated Regulation (EU) 2017/389 the daily cash penalty is a rate on the trade's value — 1.0 basis point a day for liquid shares, 0.5 for illiquid shares, 0.20 for corporate bonds, 0.10 for sovereign debt, and the central bank rate for lack of cash. For CSDR penalty management, the Penalty agent prices each fail before the cut-off, reconciles every daily and monthly penalty report against its own calculation, drafts appeals and chases credits brokers did not pass on; the supervisor prepares the month end and the Head of Trade Support signs it off. Where there is no cash penalty regime, as in the UK and US, it prices funding and overdraft cost instead.
Yes. SEC Rule 15c6-2 asks brokers to complete allocations, confirmations and affirmations by the end of trade date, and unaffirmed or unmatched trades have their own view. For each unaffirmed US trade the Chaser agent drafts a chaser asking the broker to affirm in central matching; the design goal is every one chased before the 9 p.m. ET affirmation cut-off on trade date. In the sample data, 96.6% of Thursday's US trades (6,002 of 6,214) were affirmed by 9 p.m. ET. The Forecast agent proposes pre-chases before 9 p.m. ET and 23:00 CET, ahead of Europe's move to T+1 on October 11, 2027.
Yes. The Forecast agent scores the next day's settlements on trade date from positions and cash, counterparty history, central matching status and SSI changes, and says why each one is at risk. In the working solution it scored 31,980 Monday settlements and flagged 214; over the last 20 days, 38% of flagged trades failed against 0.86% of all trades. It proposes pre-chases and never sends one on its own.
No. Settlement staff can only request an SSI check. Changes go to the SSI team, which records them, calls the counterparty back on the number on file and has a second person approve; the maker never checks their own change. Emails from look-alike domains are flagged as possible fraud before anyone acts on them.
Yes. The agents draft and check; they never send an instruction, accept a partial, shape or buy-in, or book a penalty. The analyst approves every fix with a reason code, a second analyst or the supervisor releases it, nobody can release what they approved, and every step is on a chained audit trail.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your release rules, cut-off calendar, systems of record and a sample of past fails — and it goes live once every quality gate has passed. First-time setup ends with a test on yesterday's fails, so you see what the agents would have caught before you rely on them.
Every SSI change proven at the source and called back before a dollar moves.
1-min demoPosition and Cash ReconciliationEvery break explained from its source before the NAV is struck, every fix proved first.
1-min demoCorporate Actions ProcessingEvery event one cited golden record, every election made before the custodian’s cut-off.
1-min demoWe'll run Failed Trade Management on a sample of your own past settlement fails, status messages and counterparty emails.