How to get started in AI
AI is massive right now, and many businesses have had direction from their Board, CEO or other executives demanding that the business makes use of AI. Unfortunately, some are jumping in with two feet and not really planning their AI journey. So, the question remains, how to get started in AI properly?
What do you want to achieve?
The first question to ask when you want to get started in AI, is what do you want AI to achieve? AI is not a magic wand that can be put in place and find your problems – it is another tool that you can leverage to achieve your aims. In the same way as Cloud is not a solution, and the same way that Blockchain is not a solution, AI is a tool and an approach that you can leverage – or focus on, in order to achieve business aims.
Is your goal to reduce staff with AI?
That may be an end result, but simply putting in AI will not do that. You may be able to remove some menial or repetitive tasks with AI, or streamline activities that normally take many human-hours, but it all requires a change to process and activities. In the same way as mechanisation in the Industrial Revolution allowed businesses to have machinery work 24×7 to repetitively produce products, and then reduced the workforce that previously did the task. AI can remove the need for some low-level roles rapidly, but in most cases it needs additional work.
Is your aim to replace people with AI?
Slightly less controversial, but AI still needs oversight and human review – so it will not be able to replace people completely, but it can accelerate and support decision-making and improve efficiency by doing a lot of the work that a person would normally do.
Is your goal to provide better customer experiences with AI?
With the use of AI, you can provide 24×7 chatbots, and recommend new products to customers based on their shopping or browsing habits. You can train the chatbot on your existing FAQs and help documentation (which customers don’t read), and guide them to the pages on your website that they just cannot find. You can assist humans with providing more information about your customers, predict what they may want to ask, and identify frustrated words and tone. But, just throwing in an AI tool will not magically make it work.
Is your goal to gain a market edge with AI?
This could be a lofty aim, but unless you have an idea of what edge you actually want to gain, just puttin AI in to your business is not going to show you the light.
Is your aim to analyse and get benefits from large reserves of data that you hold with AI?
Now we are getting somewhere! If you have plenty of data, such as previous customer interactions, telemetry, historical records – they can all be used by a RAG AI to assist with decision making and innovation.
Starting in AI
Once you have an idea of what you want to get out of it, then you need to start with how you will control and understand it. AI is a powerful tool, but it makes mistakes and will need verification and oversight. Ethical use of AI means being responsible, ensuring that areas such as privacy, bias, accountability and maintaining information security are all considered. This needs to be pervasive and inherent, not just a layer of a check.
AI Transparency policies
For your customers and staff, you need to be upfront with where your data comes from, where it is stored, and what your policies and controls are. This can help understand biases, limitations and allows users to take responsibility for their use of AI systems. With a clear understanding through transparency, ethical and security breaches can be identified and then addressed.
AI Fairness controls
If your AI is providing decision support, you need to ensure that discrimination and diversity are considered, with a view to ensuring that bias can be detected and corrected. This can also include privacy, that personal information from one person is not revealed to another. Training data needs to be carefully screened to ensure that it is not biased – see Dr Joy Buolamwini’s experience.
AI privacy and security
AI will learn from the information provided to it, and this can include personal or corporate information that is uploaded to it, or made accessible to AI through other means.
