ESG Exclusion Screening

ESG exclusion screening — also called negative screening or exclusionary screening — keeps the companies a fund has promised to exclude out of its portfolio: controversial weapons, thermal coal or tobacco above a revenue threshold, UN Global Compact violators. ESG Exclusion Screening tests every holding against each fund's ESG exclusion policy with seven AI agents that read each issuer's own annual report in its language, compare it with two ESG data feeds and test every policy at once. An analyst proposes each change to the exclusion list, compliance approves it, and only then does it reach the order system.

ESG screening tools that rely on vendor data alone see estimates or revenue bands, often a year or more old, and the vendors disagree. Here the revenue share comes from the issuer's own segment note, reconciled to the income statement and cited to the page, and both feeds are kept beside it. Any gap over 2 points is escalated to a person with the three readings and the reason for each, never averaged.

Yes. Each policy is a set of versioned rules with thresholds, so one manager can run firmwide baselines, the EU Paris-aligned benchmark exclusions and Climate Transition exclusions that ESMA's fund naming guidelines tie to fund names, US sustainable funds, index funds, faith-based screens and clients' own mandate lists side by side. In the working solution, the PAB exclusions include hard coal and lignite at 1% of revenue, oil fuels at 10%, gaseous fuels at 50% and high-carbon power at 50%.

For controversial weapons screening, the Weapons and Norms agent follows ownership chains in the issuer's own filings. In the working solution it found cluster munitions at a subsidiary 54.0% owned by a Korean parent, cited in the parent's Korean business report; the ownership rule counts subsidiaries 50% or more owned. It never excludes on news alone. The quarterly board report then shows controversial weapons exposure as the SFDR PAI 14 indicator: 0.001% of AUM in the working solution, nil once that position's cure ends on December 3.

The Vendor Check agent shows the filing and both feeds with the reason for each gap — an old year, a misclassified activity such as metallurgical coal counted as thermal, a subsidiary mapped to the wrong parent. It drafts a challenge to the vendor with the page attached, and an analyst decides whether to send it. Disagreements that change no outcome can be noted in bulk after checking a sample.

After a compliance approval, the List Publisher agent publishes the new list version to the order management system as its pre-trade restricted list and reads the entry count back. The Pre-trade agent warns traders on orders in names proposed for exclusion, blocks new buys once approved, and always allows sells that reduce a breach. It never cancels or amends an order.

Yes. The agents read, compare, test and draft; they never change a policy, propose or approve a list change, or cancel or amend an order. An analyst proposes, a compliance officer who did not propose it approves, policy changes need two signatures, and every action is on a hash-chained audit trail.

The Agentic Solution Engine builds and deploys it from your requirements and documents — your prospectuses, client agreements and exclusion policies, your data feeds and a sample of past decisions — and it goes live once every quality gate has passed. A guided setup connects the order system, portfolio accounting, security master, feeds and document management, loads the universe and codes each policy from its document with two signatures; the agents are tested on a golden set and one quarterly re-test runs beside the old process before go-live.