Can AI replace the traditional human compliance officer?

For the last few decades, there has been an AI phenomenon. After years of artificial intelligence being thought of as ‘in the future’, cognitive technologies such as visual recognition and natural language understanding have now become more accurate and affordable, and many companies have therefore found a way to use them. This trend has been greatly accelerated due to the pandemic, and will likely continue to accelerate, especially for the younger generations.

What is artificial cognitive intelligence?

AI technology is often a cyber robot who can communicate with humans and make human-like decisions based on how it’s programmed. Even though ‘AI technology’ may appear simple, making a cognitive bot actually takes a combination of different technologies. For example, in order to make a bot that was trained in English but can converse with people in other languages, it would require a combination of cognitive technologies such as:

• natural language understanding
• intent clarification
• audio recognition
• text recognition
• speech-to-text capability
• text-to-speech capability
• translations capability
• memory of past conversations
• proficiency

In addition, the bot would need to be programmed to give the correct responses to each question, including with accents and customised answers. Basically, an AI bot is a collection of technologies analysing and building each area of human cognition to render it intelligent enough to be like a real human. And just like how human experts specialise in one particular area, the same will apply for an efficiently-trained bot. The bot’s specialty is its ‘domain’.

Can AI replace humans, and perform even better?

Many people believe that bots can replace humans, but, at least in the near future and in most areas, they won’t. The goal of AI is to be as close to a human as possible – meaning most AI will not reach, let alone exceed, this goal. A bot’s processing speed and memory capacity can be faster and greater than a human, but this doesn’t mean the programmed machine can make better decisions than an intelligent and experienced human. Also, because the answers will always be programmed by a human, it’s hard to expect a machine would make a better choice than its trainer (even though in areas like gaming this is possible, with AI solution AlphaGo actually defeating the world’s number one Go player Ke Jie).

Is there any limit to the domain to be acquired by AI?

To make a smart AI solution, some elements are necessary, including:

The more accumulated historical data that is available, the better and faster AI can be created, and the domain can be quite diverse, from medical or legal to entertainment. For example, did you know that AI could create an effective and engaging preview for a new movie – something believed to need human creativity? As long as there is enough data to be patterned from tagged historical film previews, the trained AI can analyse the visuals and find the most common patterns to create one of the most compelling previews viewers have seen.

That said, there may not be any limited area that AI can’t conquer, and likely any domains can be captured by AI as long as there is a demand and a supplier. Meaning conquering a domain is more often a business decision based on return on investment rather than feasibility.

Can AI compliance officers replace humans?

Considering the conditions of building smart AI, it is safe to say AI won’t, and can’t, completely replace human compliance officers in the near future. Strategic compliance officers make everyday decisions by keeping an open mind and analysing each individual case, rather than looking at patterns. There are also difficulties in gathering data for training the AI, as such data is not commonly publicly available or is hard to document.

But utilising a well-trained AI bot to make rule-based decisions does free up the human compliance officer so they have the time to properly consider each case-specific decision.

Which compliance areas can AI assist with?

Among other things, well-trained AI can assist with:

• answering employees’ enquiries and advising on policies, codes of conduct and regulations in different countries, as well as having a general understanding of industry risks based on activity and third party
• automating pre-approvals for gifts, travel and entertainment expenses based on local regulations and limits
• conducting initial reviews and triaging cases, such as conflict of interest disclosures and misconduct cases reported from various channels
• assigning training courses based on the trainee’s background and roles, and populating training reports for managers and executives.