The hard work of making things simple: Putting AI to work at Fund Recs.

“I would have written a shorter blog on AI, but I didn’t have the time.” ChatGPT

AI makes things easier in a multitude of ways, including the world of Fund Data Automation. At Fund Recs, we’re on a mission to simplify data for our clients, helping them focus on solving core business challenges instead of battling complex processes. Artificial Intelligence (AI) has become an important tool for us in that mission, and we’d like to share how we’re incorporating it into our platform to drive innovation and efficiency.

This is Part 1 of a 2-part blog series that will take you through our approach to AI. In this post we’ll look at why AI matters to us, how we approach AI development, and how we’re keeping Security and Data Privacy at the forefront of our development process. Part 2 will discuss how we’re going to keep our clients up to date and up to speed on our AI features, some use cases we’ve already implemented and what the future holds for AI at Fund Recs.

Why AI Matters to Us

AI isn’t just a buzzword for us—it’s a tool to create tangible value for our clients. We want to be laser focussed on integrating AI where it can make a transformational difference to how people use our products. For us, that started with taking a step back and defining the areas where we feel AI could have the greatest impact. Our AI initiatives focus on three key areas:

1. Simplifying Our Products: We believe software should be intuitive. We pride ourselves on having a no-code platform that doesn’t require any technical expertise to learn. Natural Language Processing (NLP) will allow our clients to interact with our solutions using prompts and commands, reducing the need to learn our terminology and configurations. We see this as a huge unlock for how software is used across all industries.
2. Scaling Smartly: By automating repetitive tasks using AI, we’re freeing up Operations teams to focus on strategic initiatives, ultimately enabling our clients to support more business without stretching resources. AI is an enabler for our users, not a replacement.
3. Expanding Product Capabilities: With AI, we’re solving problems we couldn’t before. For example, in the Financial Services industry, a huge amount of time and effort is still expended extracting data from large unstructured documents. With previous methods of template creation using defined rulesets, the cost/benefit of creating automations for unstructured datasets often didn’t make sense. Using AI, the process can be much more automated and scalable.

Security First: Protecting Client Data

We operate in an industry where security and data privacy should be the first thought when developing new tools or features, not an afterthought. We have highly restrictive clauses built into our contracts around our clients’ data and how it’s used and stored. None of those requirements change with the use of AI. Some of the ways we’re ensuring that’s the case are:

• No Use of Client Data for AI Training: Currently, we don’t use client data to train our AI models, and we don’t anticipate this changing in the short term. Much of our focus is on using NLP to simplify how our product is used, and to intelligently create rules to extract and transform data for use downstream. Think “how do we turn the users’ natural language prompts into a set of Fund Recs rules that will give them the outcome they need”. This means that the transactional data itself is not required, it’s the users’ interaction with our ruleset that matters here.

There are certainly AI use cases we’ve considered where using anonymised client data would be a prerequisite to making it work. Automatching of data in a reconciliation is probably the most relevant use case here, where analysing previously manually matched transactions could be used to train an AI model on what future matches should/could look like. If and when we do decide to look at use cases where client data would be required, we commit to being fully transparent, obtaining explicit client consent with opt-out options.

• Data Containment: We are using open-source models as the foundation for our AI features, fine-tuning them within our infrastructure to maintain control and security. This means that all AI processes run within our infrastructure, ensuring no client data is exposed to external vendors.
• Robust Data Protection Policies: As mentioned, we adhere to the highest standards in data privacy and security, and none of that changes with the use of AI. We are ISO 27001 certified, SOC 2 compliant, and comply with all data protection regulations where we operate, including GDPR in the EU.

How We Innovate: Iterative AI Development

At Fund Recs, we believe that innovation thrives on iteration. Rather than rushing to implement sweeping changes, our AI development process is deliberate, lean, and focused on delivering maximum value with minimum disruption. Here’s a closer look at how our iterative approach works:

1. Proof of Concept (POC): Starting Small, Thinking Big

Every new AI feature begins its journey as a Proof of Concept (POC). This phase allows us to test the feasibility of an idea in a controlled environment before committing significant resources to development. Here’s how we handle POCs:

• Controlled Testing Environment: We use User Acceptance Testing (UAT) environments to simulate real-world usage scenarios without impacting live operations. This ensures that even experimental features remain risk-free for our clients.
• Clear Success Metrics: Every POC has defined goals, such as improved processing speed, enhanced accuracy, or user satisfaction. These metrics guide our decision on whether the feature should move forward.
• Low-Cost Experimentation: By starting small, we minimise costs and resource allocation, allowing us to test multiple ideas without committing to one prematurely.
• Rigorous Security Reviews: Every feature undergoes a comprehensive security audit to ensure it adheres to our strict data protection policies.

2. Feedback-Driven Refinement: Listening and Learning

Once a POC demonstrates potential, the next step is refining the feature based on feedback from key stakeholders, including clients and internal teams. Feedback plays a pivotal role in ensuring our AI features meet real-world needs. Here’s what this process involves:

• Internal Expertise: Our Client Solutions and Product teams provide initial insights, ensuring the feature aligns with both client needs and our overall product strategy.
• Client Insights: During testing, we often ask clients to use the feature in controlled scenarios. Their feedback helps us identify pain points, usability issues, and opportunities for improvement.
• Iterative Adjustments: We make multiple rounds of updates, addressing feedback and continuously fine-tuning the feature. This ensures the final product is as robust and user-friendly as possible.
• Performance Testing: We stress-test features under various scenarios to confirm they can handle real-world usage at scale without compromising speed or reliability.

3. Planned Releases: Delivering Confidence

Once a feature has passed through POC, refinement, and controlled rollout, it’s ready for a full production release. This final stage is all about ensuring the feature meets our high standards for performance and usability:

• Feature Flags: All new AI features are gated behind feature flags. These flags allow us to activate or deactivate a feature for specific clients or environments, ensuring a phased and safe introduction.
• Documentation and Training: To support clients, we provide detailed documentation, training sessions, and tutorials for each new feature, ensuring they can take full advantage of the capabilities AI offers.
• Real-Time Monitoring: During rollouts, we closely track usage data, system performance, and error rates to identify potential problems early.

Our iterative development process ensures that every AI feature we release is well-tested, thoughtfully designed, and aligned with client needs. By starting small, listening to feedback, and rolling out features gradually, we reduce risks and maximise the impact of our innovations. This method also allows us to adapt quickly to changing circumstances and emerging opportunities, keeping Fund Recs at the forefront of AI-driven solutions.

This approach isn’t just about building better products—it’s about building trust with our clients by showing that we prioritise their experience and security at every step.

That’s it for Part 1 of the series. Join us in Part 2 for a look at some real use cases, areas we’ve pinpointed for future development, and how we aim to ensure that our clients are aware of, and getting the most value from, our AI features.