AI Governance & Ethics: The Non-Negotiables for Leaders

An effort to use AI to approve loans at car dealerships in Canada proved successful, boosting annual profits by 8%. That's a great story and a good example of how AI can help. Unfortunately, in this case, the same AI also discriminated against women, incorrectly turning down their applications. The good news, in that story, is that this was recognised and the algorithm was corrected.

There are many, many stories like that and not all of them have a happy ending. Not because AI is inherently unethical or misogynistic but simply because AI is trained on real data, which tends to have human biases baked in.

We are all rushing to implement AI and position our organisations for the future. There is no two ways around this and it's something we all need to do. AI adoption is racing ahead, but ethics and governance often lag behind with 49% of business leaders admitting they are not prepared to use AI responsibly.

How do you make sure your business will recognise when your AI solutions cause unethical outcomes? And how do you correct it?

One approach is to do everything yourself. Another is to work with a partner who can help you, not only with the AI implementation itself, but also with the ethical and compliance implications. A competent AI partner will help you explore and implement AI use-cases for your organisation, and they will also help you set up policies and implement governance frameworks. They may even be able to manage some of those things for you. In either case, keep on reading for some pointers.

You are a good person. You want to run your business fairly and ethically. You also want to use AI in your organisation. That means you will need to spend some time thinking about the ethical risk and implications of what you are doing and how AI may handle the problems you give to it.

What is required to do this well?

However you go about it, the minimum you need are;

1. Ethical guidelines or policies – a set of principles (e.g. fairness, transparency, accountability) that your AI projects must adhere to.
2. Governance structure – appoint AI ethics officers or a committee to review projects. You will probably want to involve both the HR department and your legal advisor (if you have one). This can be incorporated into your existing software project review processes that you may have in place.
3. Risk assessment & compliance – integrate AI into existing risk management processes and ensure compliance with regulations.
4. Continuous monitoring – regularly audit AI solutions for bias or drift. This is similar to how you would monitor any software solution for fault rates: Bias in an AI model is essentially a bug - it's just that you need to monitor for it differently to other types of bug.

Some of this is similar to things you probably already have in place for other things (data protection, security etc) while other things are completely new.

Working with an AI Partner

No-one can do it all for you. You still need to have the right policies in place and you need to spend at least a bit of time thinking about the ethical implications and potential pitfalls in what you are doing. This is no different to anything else you do in business. But a partner can help with a lot of the practical implementation.

Here is what you should ask of any partner that you are working with on AI:

1. What are your Ethical AI guidelines and policies? Can you help us write ours in such a way that they fit with how we work with you and take account of what we outsource to you?
2. Once the AI solution is up and running, are you able to monitor it and flag up potential bias or drift in the solution?
3. When you design AI solutions, how do you go about assessing the ethical dimensions? How do you do risk assessment and management? What are your policies and processes?
4. How do you go about testing, in the real world, that the AI solution does not have unintended consequences?

At NewOrbit...

We have always considered ourselves "Product Partners" rather than just a development partner. That is why we bring ISO accreditations, risk management, project governance, product management, security expertise, data privacy governance and much more to our customers in our development projects. This doesn't free our clients from all responsibilities, but it does reduce the overhead of managing a software product immensely.

We are currently establishing our own Ethical AI framework so that clients can rely on us for governance, compliance and ongoing monitoring – and we will help them define what stays in-house vs what they can confidently outsource to us.

AI is already proving transformative in how we all work. But it comes with significant dangers, some business-ending, that must be managed. Striking the right balance between rushing to reap the benefits of AI and ensuring you continue to treat your stakeholders right is important.

If you want to explore how to put these principles into practice without overwhelming your team, let’s talk.