With AI tools becoming more powerful and popular, it is important that purpose-driven businesses carefully consider how they use AI. The UK Government recently released a framework for the use of AI within government, much of which is also relevant to businesses. Our top ten tips are:

1) Purpose and ethics

Think about whether the AI tool you are planning to use fits in with your broader aims and purposes. How does the use of AI align with your company values and ethics? Consider whether the use of AI could impact your reputation or detract from your goals. If you are a B Corp then consider whether your use of AI is transparent and whether it will contribute to an improvement in your social and environmental impact.

2) Understand your goals

Before you start to use an AI product in your business make sure you understand what you want to achieve and how the tool you have selected will help you do this. Make sure you understand what the tool's limitations are, as no tool or software is going to be a silver bullet. Understanding the tool you are using will reduce the risk of "black box problem" – where an AI tool makes decisions and it isn't clear how that decision was made. This makes it difficult to diagnose and fix any problems which occur.

3) Quality of input data

Think about, and make sure you understand, where the input data has come from. Any AI tool you use will reflect the data set it has been trained with and there is a risk that biases in the training data will cause biases in the outputs produced. Consider when the input data was collected, as there will have been a cut-off date for the inputs, which means that the outputs produced by the AI tool may not reflect the latest data available.

4) Team expertise

Consider whether your team have the expertise to select, implement and use the AI product you have selected, and if necessary look to fill any skills gaps. You could also look to put in place training for team members to upskill them to deal with AI.

5) Environmental impact

Consider the environmental impact of AI. Like everything we use, AI has an environmental impact. Training a new model uses a huge amount of energy and water, and produces "e-waste" which can contain difficult to dispose of chemicals. Using a pre-trained model will help reduce this, but as with everything your business uses you will need to consider the environmental impact.

6) Impact on individuals

AI should be deployed in a way that ensures fairness, transparency and accountability. You have an obligation under the Equality Act not to discriminate, so it is important that you consider how the use of AI impacts individuals and certain groups, and whether the use of AI could have more of an impact on one group than another. For example, if you switch to an "AI chat bot" for your customer service could this disadvantage users who are less tech-savy? Think about what you could do to mitigate this risk.

7) Confidentiality, data and IP

What data and confidential information are you going to be sharing with the AI tool? Do you have a right to share this information, and is it ethical to do so? Do you understand whether that data or information will be kept confidential, or will it be used to train the AI model? If you are going to share any personal data with the AI tool then you need to consider how you will comply with your obligations under data protection legislation when doing so. You may need to ask for additional information or reassurances from the supplier. If you'd like to learn more about data privacy considerations when deploying AI, you can find our insights here.

8) Include a trial and testing period

Don't rely on AI from the get go – most AI tools don't provide accurate information all the time, and there are known issues with AI "hallucinations". You will need to think about whether this is likely to be an issue and put in place safeguards to prevent this.

9) Oversight and review

Think about where human oversight is needed. There are some circumstances in which individuals have a legal right to human review (for instance where personal data is being processed) but it is helpful to have human reviewers regardless to ensure that any issues are picked up promptly.

10) Think critically

Think critically at each stage – do not allow hype or excitement to detract from your decision-making processes, or the normal steps you would take to check a software product before you implement it.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.