Skip to main content

The Globe and Mail

Could a robot catch the next Madoff? Ottawa startup bets AI can root out fraud

MindBridge CEO Eli Fathi is pictured during a meeting with staff in Ottawa, on May 10, 2017.

Dave Chan/The Globe and Mail

Eli Fathi was travelling in September, 2015, when the veteran tech executive – one of the most sought-after mentors in Ottawa's tech scene – got a call from a stressed-out entrepreneur named Solon Angel.

Mr. Angel had set out to build artificial intelligence (AI) software to help auditors detect fraud and other accounting anomalies. Incidents of fraud had surged in the past decade, but while many were exposed by whistle-blowers, external auditors uncovered just 4 per cent of known offences and software tools caught even less, according to the Association of Certified Fraud Examiners. The Brazilian-born Mr. Angel, who had moved to Canada in 2007, was convinced that regulators armed with the self-teaching tool he was developing would have caught notorious fraudster Bernie Madoff much earlier.

Suddenly the calls were starting to come in to Mr. Angel's six-month-old Ottawa startup. A top Canadian insurer wanted to see the software in action. A global retailer and a Bay Street financial institution were interested.

Story continues below advertisement

The problem: Mr. Angel's product was little more than a concept, backed by rudimentary technology he described as "garbage."

"We were selling it with a deck as if the product was completely built," said the 34-year-old, who is more product marketer than developer. "There was nothing. …We had not even a prototype. But people started saying, 'OK, bring the technology in.' "

He needed help – fast.

To Mr. Fathi – who had founded or co-founded six companies since moving to Ottawa from his native Israel in the early 1970s and sold his latest to SurveyMonkey the previous year – this wasn't just another distress call from an entrepreneur in over his head. It was the opportunity he'd been looking for.

"I knew [for my next company] I was going to do fintech. I knew it was going to be data science and big data," Mr. Fathi said. "[Mr. Angel] showed me a problem that was huge, and the timing was right. I knew that once you find the problem, to build the company around it, I could do that."

So he joined as CEO in early 2016.

Now, 18 months later, MindBridge Analytics Inc. under the 65-year-old Mr. Fathi's watch has transformed into what could be Ottawa's first serious contender in the red-hot AI sector – a sector that has been dominated by developments in Montreal and Toronto.

Story continues below advertisement

Mr. Fathi has recruited one of the most seasoned early-stage management teams in Canada. It includes chief technology officer Robin Grosset, former chief architect with IBM's Watson Analytics group, and local executives from IBM and Canadian Bank Note Company Ltd. to lead sales and corporate development. (Mr. Angel, formerly a product manager with an audit analytics firm, is now chief strategy officer).

MindBridge has raised $4.3-million from investors such as Montreal's Real Ventures and Silicon Valley venture capital firm 8VC. Mr. Fathi has an aggressive set of goals that sees MindBridge in use by 1,000 organizations in 2019, ready to go public by 2020 and generating $100-million in revenue in four years.

The company – whose software has been built out by Mr. Grosset and a team of developers – is winning praise from its 20 audit customers and partners in Canada, the United States and the U.K. (up from seven in April), including BDO and Grant Thornton.

"Their machine learning capabilities appear to be unique in the audit field and go well beyond the capabilities of any data analytics [provider] we've seen," said Scott Spradling, vice-president, audit and accounting with Thomson Reuters, which is adding MindBridge to its tax and audit accounting offerings this year. "It will allow us to deliver data analytics capabilities to our audit customers in a way the market has never seen."

Kirstie Tiernan, a director with BDO in Chicago, said: "Some of what we've been able to do with MindBridge wouldn't have been possible [with prior software] – period," including a recent review of millions of journal entries for clients in a three-day period, during which it uncovered two instances of fraud. She said BDO plans to use MindBridge for forensic audit investigations.

The software isn't meant to replace auditors, but rather help them up their game while saving time. Auditors typically apply standard rules – for example, flagging unusually sized transactions – to a sampling of entries from a firm's books. Reviewing everything would take too long using traditional methods. MindBridge applies the same standard rules automatically, but the software reviews all entries in a fraction of the time auditors would take.

Story continues below advertisement

"Fraud is a one-in-10,000 scenario, so the idea you'll find that from a statistical sample is ridiculous," Mr. Grosset said.

The software can also spot irregularities auditors might miss, such as sequence gaps in transactions, and apply statistical methods such as Benford's Law – which holds that individual figures in a typical set of numbers most commonly start with one and most rarely start with nine – to root out phony entries.

But the algorithms also learn from how auditors interact with the data. "Every action you take to follow up on something you find tells us you've found something valuable," Mr. Grosset said. "The converse is if you don't follow up, this is normal."

After reviewing the data, the software produces a neat, user-friendly table summarizing how many entries it flagged as suspicious. It is then up to auditors to review those entries and decide what to do next.

"MindBridge helped us realize that machine learning and AI will be far more efficient and powerful for auditors to do their jobs [than the] antiquated and very cumbersome processes [they now use]," said Real principal Sam Haffar. "This is a case where there's a massive arbitrage opportunity that will benefit clients and auditors with new standards."

The company plans to initially target chartered professional accountants – there are 38,000 such firms in the U.S. alone – before aiming for the larger and more lucrative internal corporate audit market, as well as regulatory and investigative agencies. Each audit will typically generate thousands of dollars in revenue for the startup. "This is a very significant and profitable opportunity" for MindBridge, Mr. Spradling said.

Eric Au, leader of the analytics innovations group with Grant Thornton, said MindBridge's biggest challenge isn't proving the technology works, but convincing audit professionals to trust it. "I think that's one of their biggest hurdles," he said. "Even if it's demonstrably shown a machine is better than humans, you still have to get people comfortable with it."

But BDO's Ms. Tiernan said auditors will come around once they see AI auditing in action: "They'll say, 'How can we possibly go back to doing it the way we've done it before?'"

Report an error Licensing Options
About the Author

Sean Silcoff joined The Globe and Mail in January, 2012, following an 18-year-career in journalism and communications. He previously worked as a columnist and Montreal correspondent for the National Post and as a staff writer at Canadian Business Magazine, where he was project co-ordinator of the magazine's inaugural Rich 100 list. More

Comments

The Globe invites you to share your views. Please stay on topic and be respectful to everyone. For more information on our commenting policies and how our community-based moderation works, please read our Community Guidelines and our Terms and Conditions.

We’ve made some technical updates to our commenting software. If you are experiencing any issues posting comments, simply log out and log back in.

Discussion loading… ✨

Combined Shape Created with Sketch.

Combined Shape Created with Sketch.

Thank you!

You are now subscribed to the newsletter at

You can unsubscribe from this newsletter or Globe promotions at any time by clicking the link at the bottom of the newsletter, or by emailing us at privacy@globeandmail.com.