Factsheet automation can be introduced without disruption if the switch is not planned as a cut-off date but as a controlled transition: the new pipeline first runs in parallel with the existing process, the old route stays available as a fallback, the scope grows step by step, and the line between automation and professional sign-off is fixed in advance. For investment foundations and pension funds as much as for fund management companies and asset managers, FINMA-regulated or not, that is what ensures the recipients of the factsheets notice nothing of the switch, except that the documents arrive on time more reliably.
The parallel run is the most important element. For one or more reporting cycles, the factsheets are produced twice: once along the existing route, once through the automated pipeline. The existing version is still the one that gets published, and every difference between the two is investigated: a rounding difference, a manual correction nobody had documented, or a genuine error in the new logic. Only once reconciliation holds, run after run, does publication move to the pipeline. The parallel run costs effort, but it replaces trust with verifiable evidence.
The second element is a fallback path that remains in place after the switch. For the first cycles after the changeover, it should be clear who can still run the previous process in an emergency, where the templates are kept, and at what point before the publication deadline the decision is taken to return to the old route. A fallback that nobody can operate any more is not a fallback. It is retired only once the pipeline has delivered several cycles without intervention and the people responsible decide so deliberately.
The third element is a gradual rollout. Instead of switching every product, language and share class on the same date, the pipeline starts with a narrow slice and then grows group by group. Each extension brings its own exceptions, and they can be resolved at small scale rather than surfacing all at once during the month-end close. The rollout at IST Investmentstiftung follows this pattern: Amula AI's reporting pipeline has been live there since October 2025, with 39 investment groups in scope and being rolled out group by group, rollout ongoing. The factsheets are produced in German and French in IST's own design, and monthly effort is down from 20+ to 3–4 hours (an internal estimate based on first measurements).
The fourth element is a clear line between what the pipeline does and what a person is accountable for. Data intake, calculations, layout and language variants run deterministically and traceably, with the same result on every run. Professional sign-off stays with the people who already hold it today, in portfolio management or compliance, for example. This line is written down before the parallel run begins, so that nobody on the team has to wonder whether a figure still needs checking or whether the system now does that.
The human side matters at least as much as the technology. The team that has been producing the factsheets by hand knows the exceptions that appear in no documentation, so it belongs in the parallel run from the start, not at the end. One accountable person on your side who collects differences, takes decisions and keeps the pace prevents the switch from stalling between the business and the implementation partner. A short, regular check-in during the transition cycles replaces large training sessions: the team learns the new pipeline on its own factsheets. Who keeps the pipeline current after handover, your team, your partner or both, is also settled before the switch.
A switch without disruption is therefore less a question of technology than of sequence: prove first, then switch, then extend. That is how Amula AI's reporting automation is built. You own the pipeline, templates and logic; it runs in your own environment, on the systems you already use. If you want to judge the output first, you can start with a proof of concept: up to 10 key pages of your report rebuilt in your design with dummy data, before committing to a full build. Know your bottleneck? Start with reporting. Not sure where to start? Start with the AI Assessment.