This global trading platform sped up a critical piece of customer onboarding & compliance 10x with Reindeer AI

Verifying where a customer’s money came from is a legal requirement for financial platforms and a mechanism for identifying suspicious activity, like money laundering. But when your review process depends on manual document handling, even well-run operations start to break under volume.

This is the story of how one trading platform used Reindeer AI to automate a critical piece of its customer onboarding and compliance process. As part of its know-your-customer (KYC) and anti–money laundering (AML) obligations, the company must sometimes verify the source of a user’s funds.

Reindeer AI took over the slowest, most error-prone part of that workflow by extracting, auditing, and analyzing the documents that support those decisions without the need for new systems or workflows.

COMPANY

A regulated trading platform needed to improve a critical part of the workflow to meet their KYC and AML obligations

A popular investment platform that blends trading with social networking helps users buy and sell assets like stocks, and ETFs, while following and copying the strategies of other investors. It’s designed to be open, easy to use, and globally accessible, but must also be tightly regulated.

Before any customer can begin trading on the platform, they go through a structured onboarding process. That includes collecting personal and financial information, assessing risk, and making sure the customer’s activity aligns with compliance policies. Onboarding is where anti–money laundering (AML) enforcement begins, and where know-your-customer (KYC) protocols are first applied. Afterwards, if the customer makes a large deposit, the company may need to take the process a step further to verify the source of those funds.

This “source of funds” verification typically applies to users making large deposits or triggering risk alerts. They may be asked to submit documents like payslips, bank statements, or proof of asset sales. These documents help the company determine whether the user’s activity is legitimate or potentially fraudulent.

Verifying a user’s source of funds is a time-sensitive, detail-heavy task. Documents come in a wide range of formats and quality; some might be photos, some might be PDFs with missing pages, and some may not match what the customer initially declared. Reviewers are tasked with extracting the correct information, cross-checking it against the customer’s profile, and deciding whether the activity is legitimate. Automation would speed up a slow, inconsistent process while making it more accurate. If they could increase verification time without sacrificing accuracy, they would be able to handle a higher volume of customers while simultaneously onboarding them faster.

PROBLEM

Manual source-of-fund reviews were too slow and easy to get wrong

When a large deposit or suspicious activity triggers a source-of-fund review, the system flags the case and requests documents from the customer. These may include bank statements, sell and buy contracts, payslips, inheritance documents, marriage certificates, or other financial records, which come in various formats.

Each customer also fills out an economic profile when they join the platform, describing their income sources, which might include salary, savings, property, pension, a company sale, and so on.

Agents are responsible for reviewing submissions, classifying them, extracting relevant data, and verifying whether that data supports the customer’s original claims. It’s also responsible for justifying the volume of their deposit, transfer and activity in the platform. Their job is to prepare the case so the compliance team can decide whether to approve the activity, request more information, or escalate it.

The work is repetitive, time-sensitive, and hard to scale. Because submissions vary in format and quality (some are photos, some are scans, and some are incomplete), reviewers must interpret what they’re looking at in real-time, which makes the process slow.

Each delay stalls the compliance process and can frustrate customers who aren’t engaging in suspicious activity when their transfers remain unreviewed for too long.

“We had one case where a public figure submitted a completely legitimate request, but the reviewer didn’t recognize the name and flagged it,” said an operations leader from this trading platform. “That person wasn’t allowed to trade for a certain amount of time for no reason. The initial review should’ve been a two-minute pass, but it ended up in escalation.”

These kinds of errors are avoidable, but common in manual workflows. Human attention drifts, judgment calls vary, and every mistake adds friction to a process that’s already slow.

SOLUTION

Reindeer AI improved the speed and accuracy of source-of-fund verification

Reindeer AI builds custom AI agents that automate tedious, error-prone back-office workflows, such as managing source of funds submissions, extracting data, and determining whether to validate those submissions. Our custom-built AI agents don’t just read documents; they follow logic, flag issues, and advance work without requiring human supervision.

In this case, Reindeer AI was brought in to handle the repetitive parts of the source-of-fund review and recommendation process. Instead of relying on agents to manually open every file, figure out what it is, extract the right numbers, and check them against a user’s profile, the AI agent does all of that automatically.

It classifies documents, extracts relevant data, checks for inconsistencies or missing information, and structures the results for the compliance team to review. If something’s unclear, it flags it for a human. If something’s missing, it can ask the customer directly.

Unlike generic AI tools or rigid automation scripts, Reindeer’s model adapts when policies or document types change and integrates with the team's existing systems.

First, Reindeer AI performs a fast audit of the customer’s documents as soon as they’re uploaded. It understands what specific data should be extracted from each type of document. For example, bank statements and sale of property documents hold different information.

It can identify and highlight the relevant income entries directly from each, so reviewers can see exactly where the data came from. The input is transparent: every extracted field is anchored to a visible highlight in the original document.

With human reviewers, this step is a black box. There’s no way to know which part of the page they looked at or what they might have missed, but Reindeer AI makes the logic visible, consistent, and auditable.

The AI also flags potential discrepancies or missing fields and presents all the structured data side-by-side with the original file. If data is missing, it immediately requests it from the customer.

After extracting the raw data, Reindeer AI automatically groups related transactions to give reviewers context. It clusters entries that belong to the same employer, counterparty, or transaction type, like salary payments from the same source or repeated deposits from the same ATM. These groups appear in the right-hand panel and are labeled based on the system’s understanding of the relationship.

This structure helps compliance team quickly assess whether the declared income aligns with the supporting documentation, without having to manually trace individual entries.

Grouping is especially useful for spotting patterns, like inconsistent income timing, multiple sources that appear related, or deposit behavior that doesn’t match the customer’s profile. Because the groupings are traceable to the original document, reviewers can always see exactly where each number came from.

Once all the documents have been processed, Reindeer AI assembles the results into a structured verification summary. Each check is listed with a pass/fail status and a short explanation. If something is missing or unclear, the system flags it and recommends a next step, like asking the customer for additional documentation.

In this case, the AI flagged a mismatch: the customer’s deposits are higher than what their payslips and declared savings can justify. It also found missing employer details on the payslip. That could be a sign of money laundering, so the system recommends contacting the customer for more information, specifically, additional payslips or proof of savings.

Each verification result includes a link back to the original source document, so compliance can audit or override any step if needed.

RESULTS

Reindeer AI reduced time to verify on source of fund cases by 10x in just two weeks

To measure the impact of Reindeer AI, the GBS team compared its performance directly against their existing human-led process. They focused on reducing verification errors and reducing their time to verify source of fund cases on average and case throughput.

A reduction in verification errors would mean their AI agent pulled the correct data from each submission, while an reduction in time to verification would mean the team could measure more source of fund cases per day. The results were clear:

• 95% extraction accuracy, exceeding human reviewers
• 10x reduction in time to onboard and verify source of fund cases compared to manual prep

These results were achieved quickly. With fewer than 50 sample cases, Reindeer trained and deployed a working model in just two weeks. Because Reindeer AI runs on top of the company’s existing systems, it was able to step into the workflow without disrupting it.

Now that the source-of-fund model is live, they can apply the same approach to adjacent workflows, like KYC document checks, income verification, or transaction monitoring, using custom AI agents that build on what’s already in place.

Ready to process more cases, faster, with full accuracy and auditability, using a model trained on just a few samples? Book a demo to talk to one of our experts.