Can AI Detect Fraud? Case 2 - Bridging Finance

6/13/24 10:50 AM

PwC, the liquidator of Bridging Finance, a now defunct private lending fund in Canada, estimates that investor losses exceed $1bn (they are, amongst other things, suing Bridging's auditor, KPMG, for $1.4bn Canadian Dollars.) This is likely the largest alternative asset fund failure of the past 5 years. DiligenceAI gives Bridging an Operational Risk Score of just 32% - despite the appointment of KPMG as well as a top-tier administrator. So can AI pre-emptively identify funds at higher risk of fraud?

To "test" DiligenceAI against this actual case of fraud, we gave the following data to the AI engine:

  • Castle Hall's DiligenceExchange verified data pack on a Bridging fund, with comprehensive data points across the management company, fund structure and control environment.
  • A list of all DiligenceExchange risk flags triggered by the manager and fund based on that information.
  • Castle Hall's DiligenceExchange Benchmark data, to enable the AI to interpret how frequently risk factors occur in the industry across our data set of 2,000 managers / GPs and 5,000+ funds.

With this three dimensional map of data -> risk flags -> benchmark data, the AI can interpret and analyse the overall risk profile of the fund.

Clearly, a review of desk based data does not replace a traditional, full scope ODD review (and in the case of Bridging our full scope work at the time uncovered other issues which further - negatively - impacted the risk assessment). And, it must be noted that Bridging was primarily an investment fraud (bad lending policies, kickbacks etc - this article provides a chronology and plenty of colour as to the facts and circumstances). But, stronger operational controls could have protected investors around existence, valuation and governance.


  • Strong controls and an honest asset manager - good performance
  • Weak controls and an honest asset manager - perhaps even better performance
  • Strong controls and a dishonest asset manager - we all get out alive
  • Weak controls and a dishonest asset manager - we all go down with the ship

Perhaps most importantly, AI can provide an "independent" and unbiased review. In this case, the AI doesn't care if an allocator personally knew Natasha and David Sharpe, the CEO / CIO duo who led Bridging, had had drinks with them / golfed with them / partied with them / spent time at their cottage.

And AI is also immune to FOMO (Fear of Missing Out). Just as with many other investments and strategies over the past 15 years of Goldilocks low interest rates, FOMO appears to have been one of the main drivers for individuals and wealth platforms in Canada to allocate to this manager. Bridging had great returns, everyone else was in, and clients were clamoring for the funds to be on your own distribution shelf.

Here is the DiligenceAI summary on Bridging, automatically created by the AI in under one minute:


Bridging Finance Inc. ("BFI") is (was!) a Toronto-based private lending firm with $x billion in AUM as of March 31, 2019. Founded in 2012, BFI has 34 employees across 3 Canadian offices. The firm offers various credit strategies to both retail and institutional investors.

A key risk is that BFI's CEO David Sharpe and CIO Natasha Sharpe are married. This level of nepotism is extremely rare, existing at only x% of peer firms. It creates significant conflicts of interest and key person risk, especially given BFI's small size. The lack of an independent COO or CFO (which 93% of peers have) further weakens oversight.

BFI also has a very high level of external ownership (x%) by the Coco Group with a seed capital relationship. Only 11% of peers have this structure, which can misalign incentives. Employee insider investment is also low at <5% of AUM, below the 79% of peers above this threshold.

BFI's AUM has grown by >25% in the last year, faster than 87% of peers. While positive, this rapid growth creates operational risk if not managed prudently, especially as x% of AUM is in SMAs that can add complexity.

Regulation and Compliance

On the compliance front, BFI does not use an external compliance consultant, unlike 71% of peers. Given the regulatory demands of the private credit space, this is a meaningful gap.

BFI has also faced civil litigation within the past few years from a portfolio dispute and an xxxx. While not highly unusual (24% of peers report some litigation), it still raises concerns about internal practices.

Information Technology

From an IT perspective, BFI appears to use capable but generic systems (x for deal tracking, x for portfolio management, x for accounting). However, the lack of an integrated, controlled environment specific to BFI's needs is a potential weakness versus peers.

Asset Controls

While reputable service providers are used (xxxx for administration, KPMG for audit, xxxx for custody), there are still material gaps. The administrator does not independently verify assets for the NAV (a best practice for 100% Level 3 holdings). The custodians also weren't fully transparent.

Trading Controls

BFI appears to rely on the same generic systems noted above for trading and operations. For an illiquid credit book, more specialized tools to track deal flow, manage collateral and test compliance would be expected. The lack of an external ops/tech consultant like most peers also limits best practice insights.

Valuation Controls

The fund is 100% invested in Level 3 assets that are highly subjective to value. No third-party valuation agent is used. Nor is there an independent valuation committee to approve marks. This leaves valuations largely at BFI's discretion, a major conflict given performance fees are paid on these estimates. This is a red flag, especially with the CEO/CIO's incentives.

Reporting Controls

Positively, NAVs have been produced on time with no restatements to date. The concentrated/illiquid portfolio, small asset base and relationship conflicts suggest elevated risks that reporting may obscure

Operational Risk Score and Narrative

Operational Risk Score: 32% (bottom 10% of peer group)

This places BFI in the lowest tier of managers with identified operational deficiencies that introduce heightened fraud risk.

Key Negative Factors:

- Extreme conflicts of interest with husband/wife in CEO/CIO roles at small firm

- Very high key person risk with concentrated authority in two principals

- 100% Level 3 assets with valuations controlled by conflicted managers

- Inadequate governance and controls, especially asset verification and pricing

- Misaligned incentives from lopsided external ownership and low employee investment

- Lack of institutional infrastructure for a first-time fund in a demanding strategy

- Elevated business risk due to small fund AUM, high client concentration and short track record

Key Positive Factors:

- Use of reputable service providers for core fund administration and audit

- Timely reporting of NAVs with no issues in short history to date

- Portfolio management experience of leadership team prior to launching BFI

However, the positives are outweighed by negatives in this case. Having multiple red flags in a small, new, niche fund is a risky profile. The nepotism, valuation, and governance issues are especially serious and point to a high risk of fraud/abuse. It's rare to see this combination of problems in a fund being marketed to institutions.


AI is clearly not omnipotent, and with the benefit of hindsight, Bridging had multiple operational weaknesses for the Monday morning quarterback. Equally, there are plenty of other funds which have husband / wife (or brothers, sisters, fathers, daughters etc) which operate successfully. There are also plenty of private credit funds which don't have an external valuation agent, for example, (and have lovely, smooth performance gains, high Sharpes and virtually no volatility!!!)

To revert back to the earlier comments, AI is both extremely quick to produce an analysis and is not influenced by relationship investing or commercial pressures from the bus dev team to get the latest hot fund on the distribution list. And yes, the AI can hallucinate - but human analysts are hardly infallible either.

AI opens a new door to screen, score and consistently monitor thousands of funds across the industry: it is hard, then, not to see AI as a transformational tool in the operational risk management toolkit.  

And for Bridging and their principals? A final judgment from the Ontario Capital Markets Tribunal is expected shortly.

Subscribe by Email