5 Areas Where AI Could Disrupt Fund Operations
AI has arrived, but do we really know where it will have the biggest impact on back-office fund operations? Speak to tech vendors and you may be led to believe that everything has already been figured out. Speak to the people doing the work, and you will hear a different story.
It is worth acknowledging the challenge of implementing AI and deciding where fund administrators and fund managers should focus first. This is not a typical technology enhancement. Because AI has such broad capabilities, it can be difficult to design a proof of concept, mainly because teams do not always agree on what concept they are trying to prove.
For fund managers who do speak to tech vendors, it can be easy to believe that significant efficiencies have already been found, only to face a surprise when quarterly reporting comes around. At this stage of technological change, I have outlined five areas where AI could have the biggest, measurable impact.
1. Checking Quarterly Reports
First and foremost, the key investor document shared on a quarterly basis. For many Fund Admins and Fund Managers alike, even though this is a repeatable process, it's the checking of the data behind these reports that takes time each quarter — with any mistakes proving costly, especially if it is a large institutional investor that spots it.
The old historic process would involve a preparer and most likely two checkers at the Fund Administrator level (one technical, one more administrative and grammatical). The reports would then be shared with the Fund Manager/Controller for their review, which would most likely involve cross referencing against shadow books and records. This is where large bottlenecks in the process can occur.
The AI impact: The human led review process can be replaced by a click of a button type review as agentic AI tools learn from past errors and cross reference key reports. Of course, there is the question of the data being fed into AI tool in the first place, with Admins needed to ensure cash reconciliations and books and records are consistently maintained and are error free. But if a cadence of preparation, review and AI oversight can be achieved, report preparation times can be drastically reduced.
2. Data Retrieval
One of the largest bottlenecks in Fund Administration is the retrieval of data which feeds into key reports. Whether it's data from a deal team following an investment completion, or perhaps a syndicated loan portfolio or fund of funds investment. Either way, the connection of databases and the manual retrieval of information on a regular basis can lead to a lot of errors and time-consuming exercises by the Fund Admin.
Historically, Fund Admins would map out data points and must physically build in retrieval of data into their quarterly reporting process. This creates more chances of error and complicates the reporting process, as reconciliations dominate the reporting cycle.
The AI impact: There are many tools being built which use OCR technology to map data points, however a slight change in infrastructure leads to a heavy lift in terms of recalibrating the data retrieval tool. By coaching AI tools on what is needed and the common areas of retrieval, fast reconciliations can be performed that shine a light on missing information. A future state may include AI agents which log into investor portals and retrieve documents, or automated data flows connecting various data sources within an operating model. The increased access to data is the single most important impact of AI on the back office. Not just for investor reporting requirements, but also for regulatory reporting which requires timely and accurate sign off.
3. Capital Call Accuracy
Capital calls and distributions are a significant source of error in a fund administrator's operating model, especially as investors negotiate bespoke capital call templates and require extensive legal language to be included because of complex LPAs or side letter arrangements.
Although most fund administrators have capital activity files built from complex and sophisticated accounting software, allowing them to build capital call and distribution templates rapidly, the tailoring of individual investor notices still leads to errors and extensive checking processes.
The AI impact: AI tools have been built to help with key processes to issue capital calls. AI checking models to check notices and learn from previous errors and changes, LP Commitment and expense allocation models and capital scenario modelling are being used in the market today through agentic AI. What was once a lengthy, time-consuming exercise prone to error, is now becoming much easier thanks to AI involvement. This allows managers to negotiate more favorable terms with investors and deliver bespoke templates without taking on the same level of downstream operational risk.
4. LPA De-Codification
LPAs are being drafted in more creative ways than ever before as GPs seek to attract institutional capital and negotiate favorable fund terms. This creates a downstream effect, with accountants and administrators spending time trying to understand the LPA and build financial models that reflect the legal terms — especially when thinking about expense allocations or waterfall distributions.
Historically, fund administrators would appoint a subject matter expert to get into the detail of the LPA and work closely with lawyers, building complex Excel models to reflect the terms. This can create key person risk and potential errors, sometimes not unearthed until the last moment.
The AI impact: AI tools are already being used to extract key accounting information from LPA’s relating to expense allocations and waterfall calculations. The key developments going forward are the connectivity from legal drafting through to a model build out within a key accounting system. The preparation of complex calculations and the number of errors is already reducing, but system connectivity is the next hurdle which looks in reach. Checking these files is also time-consuming for the fund manager and often involves a form of shadow accounting. AI takes on this process, creating significant time savings could be found within accounting teams at both the fund administrator and the fund manager.
5. Investor Closing, AML and Ongoing Relations
As funds try to attract capital from institutional investors around the world, the closing process becomes more complex and the relationship stakes are raised, as funds rely heavily on investor capital. Funds may establish feeder vehicles in different countries or attract investors from multiple jurisdictions, complicating AML processes and slowing closing.
In the past, fund managers relied on fund administrators with a global footprint and global AML expertise to review the documentation. But this remains a heavy lift: manual checking is required, AML laws are constantly being updated, and the risk of missing a requirement is high. The number of documents changing hands at this time can create bottlenecks in AML processing teams, and strains relationships with Investors at a key time of fundraising – particularly if additional documents are needed.
The AI impact: In the future, AI can take the brunt of the work by reviewing AML documents received and assessing their completeness, before final review and sign-off by the MLRO. Getting rapid responses to investors is the most important task in managing the investor relationship at this stage of the Fund. Completing the initial AML review and highlighting red flags before a key AML/Compliance employee must review is extremely valuable and drastically speeds up the closing process and improves relationships with Investors.
As identified above, there are many areas where AI can impact the operating model. The challenge is not necessarily finding the right tool or developing that tool, as many already exist; it is identifying the impact, the cost saving, and the specific processes to target. In my experience, human judgement is still needed throughout fund processes, however the human input is free to add more value as opposed to being bogged down in repetitive checks looking for transposition errors. AI frees humans up from this mundane challenge. What is exciting and challenging in equal measure, is what we use them for next!
The key? A clear target operating model, with AI implementation and impact mapped out.
If you need help mapping these processes, identifying where AI can have the greatest impact, and quantifying the effect on the target operating model, reach out and we can set up an initial consultation.