Every trade carries context the order itself doesn't contain. Mandates, investment guidelines, restricted lists, current positions, reference data and operational state can all affect whether an order can proceed, or whether an exception needs review. We put that context in one place before the order goes out, and route anything genuinely ambiguous to the person who owns the decision.
Work a new earnings transcript against the analyst notes already on file, and where this quarter's read diverges from the last one, flag the divergence rather than smoothing it over. Every material claim stays attached to its source so the investment team can verify it without hunting for where it came from.
Screen a proposed order against the mandate rules, investment guidelines, restrictions and account-specific terms that apply to it, alongside current holdings and the reference data those checks depend on. Conflicts and genuinely ambiguous clauses go to compliance for a decision instead of being resolved by inference.
Investigate a break against the trade, the account and the settlement state behind it, and identify the system the discrepancy originated in rather than the one that reported it. Open breaks are prioritized against settlement deadlines, ownership and downstream impact.
Verify the security master and vendor reference data against the custodian's records position by position. Where an identifier resolves to two different instruments, or a price is stale enough to matter, quarantine the position for a human rather than picking a winner. In a separately managed account that error lands on one client's own statement, where nothing dilutes it.
Write attribution commentary from the same holdings and performance data the investment team has already reviewed, call out any position that drifted from what the guidelines describe, and hold one voice across every account on the cycle. Commentary should read as though one desk wrote it, not whoever was on rotation.
Track a side letter alongside every exception it has created against standard process, and bring the non-standard term forward the next time that account comes up, with the reasoning attached. An account's history shouldn't reset when the person who negotiated it changes roles.
The investment book and the accounting book disagree about a position for reasons that are entirely explainable and still cost an operations analyst an afternoon. A security master keyed to the wrong identifier does not fail loudly; it prices a position wrong until someone reconciles by hand. A restricted-list check that runs only at order entry will not catch a name added later the same day. Reference data goes stale quietly, because nothing about a stale price looks wrong on screen. And a side letter negotiated three years ago by someone who has since left is the class of exception nobody recalls until it is breached.
The order management system, the custodian's records and the internal accounting book, which are meant to agree and often don't. Around them sit the mandate documents and investment guidelines behind each account; restricted lists; the security master and its identifiers; vendor reference and pricing data; settlement instructions and trade breaks; side letters; and the performance and attribution data behind the reporting cycle.
On the investment side, it accumulates in re-derivation: the same synthesis or commentary rebuilt each cycle from inputs the systems already hold. In operations and compliance, it accumulates around exceptions, where one break or identifier mismatch is traced by hand through three systems before anyone can say which of them is wrong, and around guideline checks that get re-run manually because nobody is confident the last run saw current data.
Early conversations are free and low-pressure. If AI isn't the right tool for the job, we'll say so.