It is not unusual for institutional asset managers to work with multiple administrators, often eight to ten at a time. Each sends data in its own format with its own definitions and sometimes two teams inside the same administrator send it differently again. A lot of it still arrives as PDFs that someone extracts into Excel before it reaches the systems that need it.
APIs are often presented as the answer to that kind of fragmentation and they do help. Data still reaches investment accounting in many shapes and through many routes, though, from providers that each define things their own way. APIs will keep mattering as one route among several. Connecting investment operations depends on what happens once data arrives: how it is mapped, validated and traced, and how quickly a new source can be added.
A connected operation needs each of those routes to arrive at the same place, mapping that can change without a software release and governed access for the AI tools and agents firms are now adding. Here’s how FundGuard approaches each part of that problem.
Why Flat Files and Point-to-Point APIs Still Run So Much of Investment Operations
Much of the industry’s data still moves in files. Sending data to a vendor can mean generating multiple flat files, zipping and encrypting them and dropping them on an SFTP server for the vendor to retrieve. This introduces delays, increases operational complexity and creates multiple copies of data that can quickly become outdated. It also adds reconciliation burdens and limits the ability of downstream systems to access trusted, up-to-date information when they need it.
Point-to-Point APIs
Point-to-point APIs improve on that, but each provider connection can still require its own setup and ongoing maintenance. Bringing on a new custodian or data vendor has typically meant two weeks of building an interface, with calls and meetings before any data flows. The same work starts again for the next provider. Each interface then needs maintenance when either side changes a field or a format, so the integration estate grows with each relationship a firm adds.
Data Adds a Second Problem
Administrators often use different data formats, definitions and communication methods, sometimes even across teams within the same organization. Greater standardization, whether through a shared industry utility or an agreed minimum data package, could significantly reduce this complexity. While larger asset managers may have more influence over their service providers, scale alone cannot resolve the lack of common standards. As we hear consistently across our industry roundtables, meaningful progress requires industry-wide collaboration rather than individual firms pursuing their own solutions.
The Cost Shows Up in People’s Time
Oversight teams often spend more time chasing data, converting formats and reconciling inconsistencies than performing the oversight activities they're responsible for. Data preparation and validation have become intertwined, with the preparation consuming much of the team's capacity.
Until common standards exist, the practical answer sits inside your platform. If data can come in through whichever route suits the provider and be mapped through configuration, each new source becomes a configuration task rather than a software development project.
Four Ways In and One Core
FundGuard’s investment accounting platform takes data in through four routes, each suited to a different kind of work. Recurring, high-volume data such as trades, corporate actions, prices and custodian files arrives through streaming feeds from IBM MQ, Kafka or from SFTP without anyone starting a load. One-off and remediation loads come in as CSV, Excel or delimited files through the web application. Partner integrations use REST and GraphQL APIs secured with OAuth 2.0, for single records or in bulk.
Manual entry through the application’s screens is kept for exceptions and corrections. The rule of thumb is that recurring data generally belongs on an automated feed, which keeps keyed-in changes to the cases that need a person’s judgment. Uploads have a clear job too. When a provider sends a corrected file or a batch needs reloading, it comes in through the same mapping and validation as the original, so the fix leaves the same trail as the data it replaces.
All four routes feed a common validation, persistence and accounting core, with mapping applied wherever source data needs to be transformed. A trade that arrives on a Kafka stream and a correction keyed in by an accountant pass the same checks before they reach the books, so data quality doesn’t depend on which route a record took. Because FundGuard is event-driven, a change that comes in through any of these routes updates the books as it arrives, without waiting for an overnight batch.
Each row carries its provenance through to the record it becomes, including the file, the row number and the source system. Any field loaded from a file can be traced back to the exact row that supplied it, so when a value looks wrong at month-end, the investigation starts at the source. Once data is in the books, FundGuard records each change as a new version with its source attached, whether that’s a file, a feed or the person who keyed it in, so the history of a position shows where each update came from.
The core also handles the awkward cases that tend to break traditional loads. A file arriving in a layout the platform hasn’t seen before is identified on its first row, before a failed load. Rows that refer to a record that doesn’t exist yet are parked and retried once it does, so the rest of the load isn’t held up.
Mapping as Configuration, Inbound and Outbound
Between a raw file and a record in the books sits a mapping rule set. Each one is a versioned definition, specific to a client, of how one shape of file becomes one type of record, such as a trade, a position or a price. It sets which columns to read, how to convert dates, decimals and codes and which rows to process.
Mapping Rule Sets Are Configuration
An implementation consultant writes them in the web application, so onboarding a new feed doesn’t wait for a software release. A library of operators covers lookups, text, numbers, dates and conditional logic, which takes care of the quirks that vary from one provider to the next, such as date formats, sign conventions and code translations. That keeps provider-specific logic in one visible place, where the operations team can see how each field is derived. Because rule sets are versioned, a change to how a provider’s file is read is recorded alongside the rules themselves.
The Same Engine Works in Reverse
Outbound rule sets turn FundGuard records into whatever extract layout a client specifies, using the same operators, so a client-specific extract for a vendor or downstream system is a configuration task too. Other variants of the engine handle market data feeds, with each vendor’s conventions for prices, FX and security master data, as well as user-defined objects for data structures clients design themselves. Those outbound requirements become configurable rule sets that can be updated when the vendor changes its specifications.
AI Accelerates Mapping
Upload a single sample file and the mapping wizard drafts most of the rule set, covering obvious name matches, standard date formats, sign conventions and code translations. The consultant then reviews the draft, starting with the rules the model flagged as low confidence. Ambiguous fields and client-specific business logic stay with the consultant. The draft is validated against FundGuard’s data model before it can be saved.
An Open Ecosystem Built for Implementation and Interoperability
Modernizing investment accounting involves more than replacing a legacy platform. Firms need to connect existing systems, onboard new data sources, preserve established workflows and ensure trusted accounting data is available wherever it's needed across the enterprise. The ability to integrate and adapt without lengthy development projects is critical.
FundGuard addresses this through an open, API-driven architecture, configurable data integration and a partner ecosystem that supports implementation and ongoing interoperability. Rather than requiring firms to rebuild interfaces for every new provider or downstream application, the platform supports multiple integration methods, allowing data to be ingested, mapped, validated and distributed according to each firm's requirements.
Technology partnerships extend these capabilities. For example, FundGuard's relationship with Gresham provides for existing custodian connectivity while Snowflake and Databricks serve for downstream access needs.
Equally important is the expertise required to implement and maintain a modern investment accounting environment. FundGuard's Implementation Partner Program provides a structured training and certification pathway for technology consultants and systems integrators.
Together, these capabilities give firms greater flexibility in how they modernize. Existing technology investments can remain in place where they continue to add value, new data sources and applications can be introduced as requirements evolve and implementation expertise can extend beyond the platform provider itself.
The result is an investment accounting foundation designed to work within the broader enterprise technology ecosystem, rather than becoming another isolated system that firms must continually integrate around.
MCP: Governed Access for AI Tools and Agents
AI introduces a different access requirement. Model Context Protocol (MCP) is an open standard that gives AI tools a consistent way to discover and use approved data and functions.
FundGuard’s MCP layer sits above its existing services and APIs, translating requests from compatible AI tools such as Claude into governed calls to FundGuard. Each request is subject to the entitlements of the person using the tool, so it can access only the data and capabilities that person is permitted to use.
Using a standard MCP interface reduces the need to build a separate connection for each compatible AI tool. Through that interface, tools can access approved holdings, NAVs, prices and other FundGuard records.
Because MCP uses FundGuard’s existing services, AI tools work with the same validated and traceable accounting data as other users and applications. Where source lineage is available, the tool can retrieve it alongside the underlying record. FundGuard’s own AI runs within its controlled environment, and client data is not used to train its models.
For operations teams assessing how connected their platform is, a few questions say more than whether it has APIs. Do all the routes run into the same validation? Can a new feed or a client-specific extract be set up through configuration? Can each value be traced back to the row, feed or person that produced it? Can AI tools reach the data under the same entitlements as your people?
See how FundGuard brings in data from your custodians, administrators and vendors and gives your AI tools governed access to it. Request a demo.
Frequently Asked Questions
What does connecting investment operations involve beyond APIs?
APIs are one route for data to reach investment accounting, alongside streaming feeds, file loads and manual corrections. Connecting operations also depends on mapping each provider’s formats and definitions into one data model, applying consistent validation regardless of the route it took and tracing each value back to its source.
How does data get into FundGuard?
Through four routes that feed a common validation, persistence and accounting core, with mapping applied wherever source data needs to be transformed. Recurring, high-volume data arrives through messaging and streaming interfaces such as IBM MQ and Kafka, or through automated SFTP file delivery. File uploads handle one-off and remediation loads, REST and GraphQL APIs secured with OAuth 2.0 serve partner integrations and manual entry is kept for exceptions and corrections.
What is a mapping rule set?
A versioned definition, specific to a client, of how one shape of file becomes one type of record in FundGuard, such as a trade, position or price. Rule sets are configured in the web application by an implementation consultant, so a new feed or a client-specific extract doesn’t need a software release.
How does AI help with data mapping in FundGuard?
The mapping wizard drafts most of a rule set from a single sample file, covering name matches, date formats, sign conventions and code translations. A consultant reviews the draft, starting with low-confidence rules. The draft is then validated against FundGuard’s data model before it can be saved.
How does FundGuard support implementation and integration with existing systems?
FundGuard combines an open, API-driven architecture, configurable data integration and a partner ecosystem to support implementation and interoperability. Technology partnerships, including Gresham, extend connectivity to custodians and data providers, while Snowflake and Databricks support downstream access to accounting data. FundGuard's Implementation Partner Program equips technology consultants and systems integrators with the training and certification needed to support implementations and integrations.
What is MCP in investment accounting?
Model Context Protocol is an open standard that gives AI tools a consistent way to discover and use approved data and functions. FundGuard’s MCP layer sits above its existing services and APIs, translating requests from compatible AI tools such as Claude into governed calls to FundGuard. Each request is subject to the entitlements of the person using the tool, so it can access only the data and capabilities that person is permitted to use.