ESG risk monitoring that finds a controversy at its first source, in any language, and shows what each fund's screen requires
Controversies in holdings scored within hours, and no fund's screen breached unseen.
See one case, screen by screen ↓ESG risk monitoring is the daily watch an asset manager keeps on its holdings for ESG controversies — environmental harm, labor and human rights abuses, corruption — and the action each fund's policy requires. ESG Risk Monitoring puts seven agents on every holding: they read news, NGO, court and regulator sources in 31 languages, merge each event into one incident, score it twice, map it to the UN Global Compact principles and each fund's own screen, and propose the step for an analyst to decide and the ESG Committee to approve.
AI in asset management: a buyer's guide to agentic AI for investment managers →
Updated
ESG risk monitoring is how an asset manager keeps the promises in its prospectuses, SFDR annexes and client agreements: no holdings with UNGC violations or breaches of the OECD Guidelines, and engagement, then divestment, when a company is in a severe controversy. Under the ESMA fund names guidelines, an EU fund with an ESG or sustainability term in its name must exclude those companies too. Today the Sustainable Investing team leans on vendor controversy ratings that disagree with each other and alert on a cycle, sets English-language news alerts by hand, and rarely sees the local court ruling, regulator order or NGO lab report that comes first, in the issuer's own language.
Then the slow part starts. Someone works out which funds hold the issuer from a portfolio accounting extract, reads each fund's wording to see what it requires and by when, and drafts a letter, all in spreadsheets and email. That is where screens fail: a holding stays in a screened fund because the vendor data lagged, duplicate reports hide how severe an event is, and the decision is never recorded with a reason. In 2024 the SEC settled with an adviser for $4 million because its ESG funds held companies their stated screens excluded, with no policies and procedures over the screening process, as SEC Rule 206(4)-7 requires.
Estimated split for one severity 4 incident held in three screened funds, from first publication to an approved engagement letter, by hand and with the solution.
Seven agents read, merge, score twice and map every incident; an analyst decides, the committee approves.
On Monday, October 19, 2026 at 4:52 p.m., the Kolaka District Court publishes its judgment against PT Seruni Nikel Tbk, a nickel miner held in three Quillfield Capital funds. Here is how the ESG risk monitoring agents and the people at Quillfield handle it, screen by screen, in the working solution.
Mei covers 140 Asian issuers. Her home screen says the agents read 52,318 articles, filings, court and NGO documents in 31 languages in the last 24 hours; 1,206 mentioned a holding, 214 incidents were opened or updated, and 6 need a person. The watchlist holds 41 issuers worth $1.92B, and 0 of 64 screened funds are breached. First on her list: “PT Seruni Nikel: court finds it liable for poisoning a river.”
“Severity 4 · UNGC principle 7 · 3 funds hold $86.0M · the policy needs a watchlist decision by Mon Oct 26.”
Each holding with an incident in the last 90 days has a trace; each spike is one incident, merged from every report in every language, and its height is the severity. Seruni's trace shows two earlier spikes of severity 2, the NGO water tests on Aug 28 and a village protest on Sep 30, then Monday's 4. The selected spike shows both scores, the principles and the three funds that hold it.
Filters for severity, Global Compact principle, fund, region, language and status, across 1,186 issuers with an incident.
The Watcher agent read judgment 41/Pdt.G/LH/2026/PN Kka on the Kolaka District Court registry, in Indonesian, and matched the defendant to SRNI IJ and to its bond issuer, Seruni Nikel Finance B.V., through the security master. By 6:40 p.m. the Merge agent had folded 37 reports into one incident: 22 Indonesian, 9 English, 4 Chinese, 2 Japanese. The reader shows the original beside the English, with every figure the agents used highlighted and cited to its page: nickel at 3.8 mg/L against a standard of 0.5 on page 9.
“Translation checked against the original by the Watcher agent (97%).”
The Severity agent scored 4 of 5 at 91% confidence on Quillfield's scale, citing each factor: toxic metals in the drinking water of 1,240 households, the company's own settling ponds, a weak response and three incidents in 90 days. The Second Opinion agent, on a different model and without seeing the first score, also scored 4 (88%). Within one level, so no escalation. Data vendor A still shows “Category 2”, last updated Oct 7.
“Not 5: a first-instance judgment that can still be appealed by Nov 2.”
The Mapping agent tied the ruling to UN Global Compact principle 7, the precautionary approach, with principle 8 secondary and OECD Guidelines Chapter VI. Under Quillfield's rules Seruni is a potential violator, not a violator, because the judgment is not final. Asia Pacific Sustainable Equity holds $38.4M (0.92% of the fund): watchlist within 5 business days, committee decision within 90 days. Energy Transition Materials holds $31.2M and EM Corporate Bond $16.4M of the 7.25% senior notes due 2029, each with its own clause.
“No restriction is coded without a committee decision.”
The Action agent proposed the watchlist in all three funds and an engagement letter to the Board of Commissioners, with no divestment while the judgment can be appealed. Mei moves Seruni to the watchlist with reason code WL-02, potential violator, and the portfolio managers are told no trade is required. The letter is drafted in English and Bahasa Indonesia with four asks, each cited to the ruling: a remediation plan within 60 days, independent quarterly water tests, the grounds of any appeal, and a meeting with representatives of the 1,240 affected households. She submits it for approval.
The ESG watchlist goes from 41 issuers and $1.92B to 42 and $2.01B.
Rafael reads the letter in both languages and approves it with reason code AP-02, approve with edits, and the stewardship team sends it through the engagement log. Mei cannot approve her own letter. The agents keep four watch conditions running: whether an appeal is filed by Nov 2, a remediation plan by Jan 18, 2027, severity rising to 5, and an appeal court upholding the ruling.
“With ‘approve with edits’, your edits are kept beside the analyst’s version.”
A former Nordhavn Shipping vessel was beached for scrapping at Chattogram and two workers were injured. The Severity agent scored 3, treating the sale to a cash buyer as indirect involvement; the Second Opinion agent scored 5, citing Nordhavn's own policy covering sales through cash buyers and the Hong Kong Convention, in force since June 26, 2025. A gap of two levels stops the flow and goes to Mei with both rationales; the Mapping agent waits for her score before applying each fund's policy.
“Agents never settle a gap over one level: read both rationales and set it.”
Taraval Agro Plantations has been severity 5 since March 2025, with forced-labor indicators at its Sabah estates and 3 of 4 engagement milestones missed in 18 months. Mei proposed divestment on Oct 12 and, as the proposer, cannot vote; Rafael voted on Oct 16. Sofia's vote, due Mon Oct 26, is the second. After it, compliance codes restriction R-ESG-0412 in the order management system and the portfolio managers sell $54.7M in 4 funds to zero by Dec 1; the Action agent estimates 4.2 trading days at 20% of volume.
“Rafael Ortiz approved on Oct 16. A second vote approves the divestment. You may vote against with a reason.”
Every action by a person or an agent is recorded with who, what, when, the reason code, the evidence and the agent's version: the Watcher agent's read at 5:03 p.m., both scores, the mapping, Mei's WL-02 decision and Rafael's approval. Entries are read-only and the chain is verified. Internal audit asks the Evidence agent to prove every severity 4 had a decision in its window, and exports an evidence pack for an issuer, a fund or a period.
“In Q3, 9 of 9 severity 4–5 incidents in screened funds had a signed decision inside the policy window.”
The same controversy, seen by the people who carry it from the first source to a decision — what their week looked like, and what it looks like now.
Seven agents read, merge, score twice and map every incident; an analyst decides, the committee approves.
Reads news, NGO, court and regulator sources and company filings in 31 languages, about 50,000 documents a day, and matches each one to the issuer, its subsidiaries and its local names through ISIN, CUSIP and LEI in the security master.
Folds duplicate and related reports of one event into a single incident across languages, by place, parties and period, and explains every merge.
Scores each incident 1 to 5 on the firm's severity scale from impact, involvement, response and recurrence, each factor cited.
Re-scores every incident of severity 3 or more on a different model, without seeing the first score, and escalates a gap of more than one level.
Maps each incident to the UN Global Compact principles and OECD Guidelines chapters, then to every fund that holds the issuer and that fund's coded policy, with the clause and the deadline in business days.
Proposes watchlist, engagement, no action or divestment as each policy requires, drafts engagement letters in English and the issuer's language and divestment files, and sets watch conditions.
Keeps the evidence file for every incident, builds committee packs, PAI data and reports with every number cited, and verifies the audit chain.
Decide, approve, vote, restrict. The agents propose; a named person decides.
On every screen, ask the agents in plain words or tell them what to do. They answer with sources and act only with your approval.
Why principle 7?
Principle 7 asks companies to take a precautionary approach to environmental challenges. The court found the company let tailings run off into a river for two years, with nickel and chromium far above the standard: a failure to prevent harm it could foresee. Principle 8 is secondary because the remedy is about environmental responsibility.
Which funds must act, and by when?
Three funds, $86.0M: Asia Pacific Sustainable Equity $38.4M, watchlist by Mon Oct 26 and a committee decision by Jan 18, 2027; Energy Transition Materials $31.2M, watchlist and engagement; EM Corporate Bond $16.4M, watchlist and bondholder engagement. No fund has to sell today.
What changed since last week?
The water tests (Aug 28) and the protest (Sep 30) were already on the trace; on Monday the court ruled, 37 reports followed in 4 languages, and this morning the company said it “will consider legal remedies”. Our data vendor has not updated since Oct 7.
Where were we ahead of our data vendor this month?
This month we flagged 9 severity 3+ incidents before our data vendors, on average 11.4 days earlier, and missed none they flagged. PT Seruni Nikel is the latest: the vendor still shows “category 2”, last updated Oct 7, while the court ruled on Oct 19.
13 screens from the working solution, on its sample data. Pick one to see it large.
The thresholds every incident is checked against are settings. Core rules are locked, and any change is saved as a new version that compliance approves.
| Setting | Default | Choose from |
|---|---|---|
| Severity 4–5 in a screened fund reaches an analyst within | 4 business hours | Locked core rule |
| Divestment needs committee votes from different members, never the proposer | 2 votes | Locked core rule |
| Restriction coded by compliance after approval within | 1 business day | Locked core rule |
| Second opinion runs on incidents of severity | 3 or more | Editable, as a new version for compliance approval |
| Escalate when the two scores differ by more than | 1 level | Editable, as a new version for compliance approval |
| Log severity 1–2 automatically at confidence of at least | 85% | Editable, as a new version for compliance approval |
| Hand to a person below confidence of | 75% | Editable, as a new version for compliance approval |
| Entity match by name alone needs similarity of at least | 0.90 | Editable, as a new version for compliance approval |
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 (SEC settled order, October 2024, $4 million, ESG screens not applied; Berg, Kölbel and Rigobon, Aggregate Confusion, ESG rating correlations of 0.38–0.71; a controversy data vendor's published research FAQ, 2020, on bi-weekly client alerts). Regulations named as context: UN Global Compact principles, OECD Guidelines for Multinational Enterprises (2023), SFDR principal adverse impact indicators 10 and 11, ESMA guidelines on fund names (2024), SEC Rules 206(4)-7 and 204-2. People, funds, companies, courts and amounts named on this page are fictional — characters and sample data in the working solution.
ESG risk monitoring is the daily watch an asset manager keeps on its holdings for ESG controversies — environmental harm, labor and human rights abuses, corruption — and the action each fund's policy requires. ESG Risk Monitoring puts seven agents on every holding: they read news, NGO, court and regulator sources in 31 languages, merge each event into one incident, score it twice, map it to the UN Global Compact principles and each fund's own screen, and propose the step for an analyst to decide and the ESG Committee to approve.
Norms-based screening, often called UN Global Compact screening, checks companies against international standards such as the ten UN Global Compact principles and the OECD Guidelines for Multinational Enterprises, and excludes, watches or engages with those that breach them. In this solution each fund's screen is coded from its prospectus, SFDR annex or client agreement and approved by compliance, so an incident is tested against what that fund actually promised.
The Watcher agent reads sources in the issuer's own language, including 212 court registries in 46 jurisdictions, 138 regulators and 1,480 NGO sites, and matches them to the issuer, its subsidiaries and local names. The original is shown beside the English with every figure cited to its page; in the demo, an Indonesian court ruling was read 11 minutes after it was published.
It depends on how the fund's policy treats UNGC violations, and the Mapping agent cites the clause it applies. In the demo, a first-instance judgment that can still be appealed makes the issuer a potential violator — watchlist within 5 business days, engagement and a committee decision within 90 days — while a final judgment, or severity 5 without credible remediation, makes it a violator to be sold within 30 business days.
No. It works beside the ESG risk monitoring tools you already license: vendor categories and norms status come in as a feed and are shown beside the agents' own score — compared, never copied. In the demo month the agents flagged 9 severity 3+ incidents before the data vendors, on average 11.4 days earlier, and the vendor still showed the Seruni case at “category 2”.
Yes. The Evidence agent drafts the principal adverse impact data for indicators 10 and 11 with every number cited to holdings and policies; an ESG analyst reviews and submits it, and investment compliance signs it. The ESG Committee report, the stewardship report, US fund board updates and client report extracts work the same way: drafted by the agent, reviewed by one person, signed by another.
Yes. The agents propose and draft; they never send a letter, vote, code a restriction or trade. The analyst decides with a reason code, the chair approves letters, two committee members approve any divestment and never the proposer, compliance codes the restriction, and every step is on a hash-chained audit trail.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your fund prospectuses and SFDR annexes, your escalation policy and severity scale — and it goes live once every quality gate has passed. A guided first-time setup codes each fund's screen for compliance to approve.
Issuer questions answered in minutes, every claim pinned to its source page.
1-min demoESG Data ExtractionHoldings move from estimated to reported ESG data, each figure cited to its page.
1-min demoPortfolio Company ESG Data CollectionEach company confirms prefilled values, and no outlier reaches an LP unexplained.
1-min demoWe'll run ESG Risk Monitoring on your own holdings and fund policies, against a recent month of controversies.