A fund management company can automate above all the processes that recur on a fixed cycle, rest on structured data, occur in large numbers and leave a person accountable for the decision: factsheets and share-class reporting, reconciliation with the custodian bank, rule-based checks against investment restrictions, and the recurring tables of the annual and semi-annual reports. Investment decisions, valuations that involve judgement and communication with the supervisor are not on that list. Four questions place any process of a fund management company in a few minutes, before a project begins.
The four questions are these. Does the process repeat on a fixed cycle, daily, monthly or quarterly? Is the input data structured, as a data delivery, a spreadsheet or an interface, rather than only free text? Does the decision stay with a person, so that the automation prepares and checks but does not sign off? And is the volume large enough to justify the build, for example across many funds, share classes or languages? Four yeses mark a strong candidate. A no is not a disqualification, but it shows where data or responsibilities need clarifying first.
Factsheets and share-class reporting answer all four questions with yes. Produced by hand, every additional share class and every language multiplies the effort; in an automated pipeline it becomes a parameter. Data intake, calculations, layout and language variants run deterministically and traceably, with the same result on every run, while professional sign-off stays in portfolio management or compliance. That is why reporting is the obvious first step for many fund management companies.
Reconciling positions, transactions and cash with the custodian bank meets the four criteria as well: a daily or monthly cycle, structured data on both sides, high volume. What gets automated is the matching, not the resolution: items that agree pass through, and differences reach the responsible person together with their context. Investment restrictions work in a similar way. Limits and concentrations can be expressed as a rule set that returns the same result on every run and logs it. How a flagged breach is to be assessed, and what follows from it, is still judged by a person.
For the annual and semi-annual reports, the answer is split. The statement of net assets, the income statement and the inventory tables rest on structured data and can be produced from the same data basis as the monthly reporting. The cycle is less frequent, however, and review and sign-off are more extensive, so extending the pipeline usually pays off only once it is established for regular reporting. For documents such as prospectuses or key information documents, AI can help flag changes between versions or check details against master data; the fund management company remains accountable for the text.
Clearly outside the scope are investment decisions, valuations that involve judgement, for example for real estate or illiquid assets, and communication with FINMA or the audit firm. Here at least one of the four criteria is missing, usually the third: the decision itself is the work. Automation can support these processes by compiling documents and preparing data, but it does not replace judgement, and responsibility for automated steps stays with the fund management company. Writing this line down in advance prevents a project from failing on compliance rather than on technology.
The result of the four questions is a short list: which processes can be automated now, which first need clean data or clear responsibilities, and which deliberately stay manual. That is exactly the mapping the AI Assessment from Amula AI works out: typically in one to two weeks, ending in a report and a live walkthrough with three tiered implementation options. What gets built runs in your own environment, on the systems you already use, and you own the pipeline, templates and logic. Know your bottleneck? Start with reporting. Not sure where to start? Start with the AI Assessment.
