A series on AI and fund administration, part three: the specific, operational places the work genuinely gets faster, and what stays a human task regardless.
The first article in this series set out, in general terms, where artificial intelligence is being applied in fund administration. This one goes further into the operational detail: what changes, task by task, when these tools are used well, and — just as importantly — what does not change at all.
Bringing a new investor onto a fund's register has always meant reading a stack of documents closely: the subscription agreement, the trust deed where one exists, bank statements and identity documents, each checked against the others for consistency. Done by hand, that review is thorough but slow, and the slowest part is usually the first pass — simply finding where in a long document a particular fact sits.
AI tools shorten that first pass considerably. A document is read once by the tool, the relevant fields are extracted and flagged with a confidence score, and a person then reviews the extraction rather than the raw document — correcting what the tool got wrong or flagged as uncertain, rather than re-deriving every figure from scratch. The review step does not disappear; it becomes materially shorter.
Matching bank deposits to capital calls and redemption instructions is one of fund administration's more tedious tasks, made worse by incomplete payment references, split transfers and currency conversions that obscure what a payment actually relates to. A reconciliation that once needed a full day of careful, manual matching can, with good exception-handling, be reduced to an hour or so of review once the routine matches have already been made automatically.
The gain here is specifically in volume, not in judgement. The tool clears the straightforward matches so that a person's attention goes to the handful of payments that genuinely need investigating — which is, in fact, a better use of an experienced administrator's time than working through hundreds of routine matches to find the few that do not reconcile.
Investor reports have always drawn on data that already exists in the fund's records — NAV figures, performance, transaction activity — assembled and written up by an analyst each reporting period. AI tools can now produce a first draft of that commentary directly from the underlying data, which a person then edits for accuracy, tone and anything the data alone would miss. The analyst's time shifts from assembly to review, and report turnaround shortens accordingly.
The same applies to the routine investor queries that make up much of an administrator's day-to-day client service — a NAV figure, the status of a redemption, a copy of a statement. These can increasingly be answered without waiting for a person to become available, freeing that person's time for the queries that actually need judgement: a dispute, an unusual instruction, a question about the fund's structure rather than its numbers.
None of this removes the administrator from the process. Every figure that reaches an investor or a regulator still needs a person accountable for it, every exception still needs a person to resolve it, and every tool's output is only as good as the review applied to it. What changes is the balance of the working day: less time spent on the transcription and matching that AI now does well, more time available for the exceptions, the judgement calls and the client conversations that remain, as they always have been, a person's job.
That shift in balance is the real efficiency gain — not that administration becomes faster in some generic sense, but that the people doing it spend their time on the parts of the job that actually need them.
This article is provided for general information purposes only and does not constitute legal, regulatory, tax or financial advice. Readers should seek advice specific to their circumstances before acting on any matter referred to above.