How ADB is using AI to refine audits

By Yun Xuan Poon

Interview with Hock-Chye Ong, Auditor General at the Asian Development Bank.

The Asian Development Bank (ADB)’s internal audit team is to its projects what a good baker is to a recipe. Just as a baker meticulously checks that all the ingredients have made it into the mixing bowl, internal auditors make sure that all the conditions set before developing a new project are met.

This is crucial in fulfilling ADB’s mission to alleviate poverty in Asia and the Pacific by supporting economic and social growth. “Our role is [to give] an independent assurance that things that should be happening are happening as intended,” says Hock-Chye Ong, Auditor General at ADB.

What can tech do to help auditors with this process? “For us, it’s the only way forward”, says Ong. He shares how internal auditors are using tech to conduct better checks, and how the team is building its auditors’ digital skills.

ADB’s tech ventures

Image from ‘ADB Internal Audit Project Sites in Kalimantan (Indonesia)’ video

There are four ways tech is helping the internal audit department to conduct better checks. Ong shares more on his team’s ventures with drones and AI.

First, auditors use drones to track on-the-ground progress of building projects. This saves auditors from having to travel to remote sites in person. It’s also helpful for monitoring the progress of a larger area, as drones can provide a bird’s eye view.

For instance, ADB collected drone footage of several projects in Kalimantan, Indonesia, including the building of new roads and transmission towers. The internal audit team is now exploring ways to harvest more information from the images they collect, “to make it something useful other than just taking a picture of the location”, says Ong.

Second, ADB uses natural language processing, a way of teaching computers to make sense of and analyse a written document, to check that key covenants in a contract are reflected in their system. This tool reads a document, compares it against what’s in the system and points out any differences. In the past, this process took around two hours for a single loan, but natural language processing has shortened it to seconds.

This “completeness test” ensures that auditors are checking against the right terms when they audit projects. “When our colleagues monitor, they use the system, they don’t go back to the agreements (anymore),” explains Ong.

Third, the internal audit team uses an AI technique to check that all payments are done according to ADB’s rules. For example, the software checks that the person initiating a payment cannot then give approval for it, and that the transaction has proper confirmation before it goes through.

Fourth, ADB uses machine learning to identify risk areas more accurately. “When you set something up based on a rule, and if it's not meeting the rule, [the system] will throw out as an exception, but some of the exceptions are really okay,” explains Ong. This narrows down the areas that auditors need to conduct more in-depth checks for.

The internal audit department is tapping ADB’s digital sandbox to expand its tech tools, Ong shares. It provides a low-risk, regulated space for testing out new tech.

Building auditors’ tech skills

ADB as a whole has embarked on a journey to digitally transform its operations and services by 2030. The internal audit department is taking “baby steps” towards that goal, Ong says.

Building a learning culture is key in this stage of the department’s tech transformation, says Ong. After all, it’s impossible for tech to become a core part of operations if nobody knows how to use it.

The internal audit department brushes up on their digital skills with online courses on platforms like Coursera. It also learns about the latest tech research from academia. The team works very closely with Columbia University in New York, which has a very strong AI programme, says Ong.

It’s also important to learn from other institutions who are further along in tech innovations. The department looks outward to see what tech exists, what other companies are doing, and how it can adapt them. “The average age in my department is 34 years old,” notes Ong. “They're very fast, they're very techy and you don't need to manage them, you have to inspire them.”

As an organisation, ADB has also taken steps to help employees understand data better. It has pulled together a ‘data dictionary’ to standardise definitions for around 300 terminologies. For more than 50 years, there had been no consistency in the way ADB talks about data, says Ong. This dictionary can help employees understand what sets of data to collect, and how they can be used.

Connecting public sector auditors

Ong’s efforts to help auditors understand tech better extends beyond his team. In 2017, ADB’s internal audit department launched AuditWithoutWalls!, an initiative that equips auditors with digital know-how and connects them through social media. Public sector internal auditors from more than 100 countries can take short e-learning courses together, ask questions and share learnings.

Ong also established the ASEAN Central Banks Head of Internal Audit Network to help internal auditors from the 10 countries learn from one another. “They can help each other out so the weaker countries can benefit from stronger countries,” notes Ong.

For ADB’s internal audit department, tech uptake is all about slow and steady. Tech will help ADB ensure its funds meet their full potential for improving lives, and it’ll also bring fellow auditors along in the journey towards digital transformation.