Artificial Intelligence (AI) is one of the cutting edges of technology along with Quantum Computing and Augmented Reality. As part of a business’ technology toolkit, these technologies present amazing possibilities and opportunities. Quantum Computing and Augmented Reality enable faster processing and enhance our ability to operate in our world. Importantly, these capabilities extend what we already do yet don’t attempt to replace what we do.

On the other hand, AI is different as its name states. AI is essentially a non-human performing tasks traditionally restricted to intelligent forms – humans. For businesses, this is extraordinarily powerful. Creating possibilities for enhancing processes, enabling previously dangerous tasks to be completed faster than human decision-making – the list is endless. However, this power comes with its own risks. Like a Ying must have a Yang; up a down; left a right, so too does AI opportunities with its dangers. While businesses shouldn’t stop exploring AI capabilities, they should go into these with eyes wide open. For starters, doing analysis to fully understand the potentials and risks in their own context and not that of the vendor. After all, no one understands themselves better than themselves, and the same applies to businesses and understanding where they accept risk.

At the core of managing both personnel risk and AI risk is the specific understanding of what is to be achieved. In other words, where is the line between acceptable and unacceptable behaviours, and what indicates the line is crossed? If we consider AI to be an artificial human, and therefore put it through the same process as recruiting a human worker, wouldn’t you want to understand the full capabilities both positive and negative? An active process of vetting a person or background checks on their trustworthiness and capabilities which is different to the individual’s self-attestation in a resume or interview.

So how do we perform vetting or background checks on AI? And what are the risks of not doing this?

Just like you need to have a good understanding of a role before it is advertised as vacant, a business must have a good understanding of what it wants to achieve with the AI product. Just like a role where you may define clear boundaries where things become unacceptable, such as too high an hourly rate or too low in experience, there must also be clear boundaries for what you want the AI to do. Considerations must be taken into account, such as the level of decision-making autonomy the system might have and what actions/decisions you want it to do. Does the system directly impact people or will the decisions still be checked by a human before being actioned? All these are essential in understanding the AI requirements.

Even before AI is looked at, failing to consider these during the initial phases means a business runs the risk that an inappropriate AI product is selected. It could be that the product does not have all the capabilities required, or worse, the product has more capabilities and is left unchecked to act outside what is required.

Having thought about what the AI needs to do, then comes the consideration of what the AI must not be doing. Again, as humans are bound by levels of acceptable behaviour such as ethics and morals, so too does AI need to be bound. And this is where some of the challenges with AI business risk arise. For humans, it is possible to create legal requirements that each of us is responsible for our own actions. We have a social understanding of accepted norms, behaviours and limits. For AI, it is a human that needs to keep oversight and hold the AI in check. That’s right – AI doesn’t alleviate the need for humans, in fact, it requires a human with the capability and capacity to oversight its actions.

Not understanding what the acceptable bounds of actions are and how to monitor the AI increases the risk that the AI will move past acceptable operating boundaries without anyone knowing. Of all the risks to businesses that AI brings, this is by far the one with the most potential for impact. Impacts such as:

• Inability to complete designated tasks
• Moving past the scope of the task to directly or indirectly impact aspects outside the task
• To change in ways that are unforeseen and therefore cease operating as previously understood

These can result in detrimental consequences including:

• Damage to reputation
• Damage to operating efficiencies
• Increased costs of managing the AI
• Inability to operate due to the AI impacting systems or processes in ways unknown
• Inability to recover from an outage or incident as the AI processing is no longer understood
• Unethical or immoral operations due to AI decisions
• Potential for litigation or loss of licences/certifications or ability to operate

In fact, depending on what the AI has done, the consequences are as endless as the possibilities AI presents.

Need help?

Anchoram Consulting is finely placed to help organisations understand their AI risk, manage the risks and create more trustworthy AI systems. Our skilled professionals have decades of experience in mapping specific technology, people, processes and physical analysis into enterprise-understandable and relevant risk with actionable risk management. We are unique in mixing our people to form highly experienced, multidisciplinary teams that can analyse business needs. Pair this with deep technology and AI understanding and in-depth experience in the pragmatic application of internationally recognised standards and frameworks to produce business-relevant and actionable advice.

To ensure that suitable governance documents the requirements of the organisation, we focus on:

• mapping the risks from implementing AI to determining their relevance to the business
• creating measures that enable monitoring of achievement of business requirements
• flagging if the AI is straying and planning for how to manage identified AI risks

We aim to align and incorporate AI risk into existing business risk frameworks as after all, AI risk rolls up to be a business risk.

If you are interested in finding out more, feel free to reach out to myself or any of the Anchoram team for a no-obligation chat.