A series on AI and fund administration, part six: what changes in the skills an administrator’s team actually needs, and what a client can reasonably expect of them.
As AI tools take on more of the routine matching, extraction and drafting described elsewhere in this series, the skills that make a good fund administrator shift with them — not away from technical competence, but towards a different, and in some ways more demanding, application of it.
The administrator whose value lay mainly in careful, accurate transcription — reading a document and typing its figures into a system correctly — finds that value eroded as tools do that transcription reliably on their own. What remains valuable, and arguably becomes more valuable, is the judgement applied to the exceptions a tool cannot resolve: recognising when a flagged anomaly is genuinely significant rather than a quirk of formatting, and knowing which figures to question even when a tool has not flagged them at all.
This is a skill built from experience with the funds themselves, not from any tool, and it is the reason a smaller, senior-led team — described earlier in this series — has an advantage: that judgement sits with people who already know each fund's particular structure and history, rather than with a generic support team working from a script.
A related skill, newer to the profession, is knowing how to work with an AI tool productively: framing a request clearly, recognising when an output looks plausible but is wrong, and verifying a result rather than accepting it because it arrived quickly. Industry surveys of administrative and operations professionals report a steady increase in people actively seeking out these technical skills alongside their existing expertise, rather than treating them as someone else's responsibility — a sign that the profession already recognises this as part of the job, not an optional extra.
This is not the same skill as being a technologist. An administrator does not need to understand how a model works to use it well, in the same way an auditor does not need to build accounting software to use it competently. What is needed is the discipline to treat a tool's output as a draft requiring verification, every time, regardless of how confident that output appears.
Because more of the first-pass work is automated, the knowledge that distinguishes a good administrator moves further into the detail of each fund's governing documents, its regulatory obligations in each jurisdiction it touches, and the particular expectations of its investors and auditors. A team that understands why a clause in a limited partnership agreement matters is in a far better position to catch an AI tool's misreading of it than a team relying on the tool's own summary of what the clause says.
This argues for administrators investing in technical training and in deepening the fund-specific and jurisdictional knowledge their people already hold — the two reinforce each other, rather than one replacing the need for the other.
Berkley Fund Services is building both of these skills deliberately across its team: practical fluency with the AI-assisted tools now part of day-to-day administration, alongside the senior-level fund and jurisdictional knowledge that lets that fluency be applied with proper judgement rather than blind trust in a tool's output. Clients dealing with Berkley should expect both from the people handling their fund, not one at the expense of the other.
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.