In 2017 we wrote about how Artificial Intelligence (AI) can benefit and enhance Robotic Process Automation (RPA), and earlier this year we looked at some ways that AI can help with the actual implementation of RPA. It’s clear that AI has a big part to play in the success of RPA (in what people are currently calling Intelligent Automation) as well as providing value in its own right. But people generally don’t know how to find AI, how to buy it, and how to deploy it. The good news is that this is getting easier and easier, as the use of AI becomes much more democratised within business.
There are four fundamental ways that businesses can access AI. The first is the simplest and currently the most popular; bringing in a software vendor that has already designed, built and tested a solution. The benefits are clear – the client doesn’t have to learn, or recruit, any specific AI skills and the value is delivered relatively quickly. All that is required is installation (or connection, if it is SaaS) and configuration / training. The tricky bit, of course, is making sure that the vendor’s solution matches your requirements exactly. If it does, you are good to go, but if it doesn’t quite match then you may need to look at other, more tailored, solutions.
At the other end of the scale, there is the competitive-advantage-wielding bespoke solution. This would be when the business has a unique set of data and an idea that no one else has done before. This sort of approach will require AI developers and data scientists, probably from a consultancy and probably very expensive. The prize though, can be very big, with the potential to transform a market or even create a completely new one.
Between these two extremes are a couple of other options, where some distinctiveness is required but where you don’t want to start from a blank sheet of paper. The less-than-optimum version of these two is where you go to your incumbent outsourcing provider and use their in-house AI solutions. This can work, but generally it will be like fitting a square peg in a round hole (and using the provider’s resources to do it with).
Probably the most favourable solution will be to use one of the few AI platforms that have come to market in the last few years. The biggest, and most popular of these, are provided by Google, Amazon, Microsoft and IBM. Each has a range of algorithmic services, such as ‘speech-to-text’, ‘Q&A’ or ‘Face Recognition’, that can be accessed through an API. The client pays a small charge every time the API is called. You will still have to build some software around the algorithms, but this solution provides a good balance between individuality and standardisation.
As well as the tech giants’ platform solutions, there are now a few independent vendors, such as Kortical, H20 and Polymatica, that can provide a range of different AI services. Most are focused on the data analytics end of AI, but some include Natural language Processing capability as well. These will be the interesting players to watch over the next year or so.
So, accessing artificial intelligence capability is set to become much simpler with the advent of more and more AI vendors and the new AI platforms. It’s never going to be plug-and-play, and you will likely have to use consultancy advice at some point on the journey, but the ability for businesses to exploit AI alongside their RPA capabilities is now easier than ever.