1. AI gets a human perspective

While the dystopian film version of fully autonomous AI replacing humans is entertaining and interesting to muse on, the reality is, AI can’t function without a human perspective. However, as AI becomes more ubiquitous and used in more industries and situations, we’re likely to see a greater commitment to, and understanding of, ‘augmented intelligence’. A concept that Gartner defines as: “a design pattern for a human-centered partnership model of people and AI working together to enhance cognitive performance, including learning, decision making and new experiences.”

As we see more human intervention in AI, and use of larger data sets based on human experiences, we’ll see improved accuracy and personalisation of experiences. And where in some circumstances, this improved accuracy is nice to have from a user experience perspective, in others – such as healthcare – it is essential to get it right.

2. Foundation models become the bedrock of AI

In 2023, we’ll see foundation models further transforming AI implementations, which will exacerbate the need for governance, compliance and testing. Foundation models are based on deep learning algorithms that have been trained with giant datasets made up of everything from images, words and even voices. In essence, they are based on a broad set of unlabelled data that can be used for different tasks, with minimal fine-tuning. This may dramatically accelerate AI adoption, but the need to ensure that the correct data is being used ethically and equitably increases as more data becomes available. Building in compliance and ongoing testing as part of the development process can help teams regularly assess whether the data is accurate or not and can even help to prevent bias. This is especially critical for AI use cases in banking, healthcare and other areas where faulty information can result in major problems.

3. Hyperautomation drives composable applications

As more cloud native and composable applications become prevalent in 2023, hyperautomation will become a reality. Composable applications make it easier for businesses to deliver new services assembled from other apps and services very quickly. Shopify is an example of this technology. It’s essentially a service made up of reusable functionalities including APIs, payments and SEO. It has everything required to deliver the full ecommerce experience created from different parts assembled to make a new digital experience. But a lot of testing is required to ensure all the composable applications function in a cohesive way.

As these applications become more commonplace, the need for hyperautomation – effectively automating any processes or functionality you can – will increase. Hyperautomation will enjoy more success with extensive testing and monitoring. This is necessary because users will access the application from different points and their expectations of how to use the application will differ. Going forward, automated testing at scale augmented with human driven testing will be needed to avoid fragmented technology and ensure the quality of the experiences.

4. New AI use cases for 2023

AI is becoming more prevalent in many industries but expect to see AI playing a bigger role in frontline healthcare next year. This will be particularly true in the UK and Europe where there’s a shortage of doctors and general practitioners. During COVID chatbots were used to diagnose symptoms, and in the future we’ll see more instances of AI filtering patients through to the correct medical professionals, once the initial diagnosis has been identified.

Travel is another area where we’ll see more adoption in the form of AI assistants. These automated assistants will help to create a more consistent travel experience. They will manage the entire process from arriving at the airport to boarding a flight, to booking a taxi upon arrival at your destination and checking you into your hotel. They will even provide you with information on where to visit and where to eat during your stay.

 5. Data is still the life-blood of AI

As AI becomes more entrenched in our daily lives, companies will be aware of the implications of not having enough or not validating the quality of the data in their AI/ML projects – poor user experience, potential for bias, and ultimately the possibility of failure of an entire project – and their costly outcomes. Many organizations find AI implementations much more difficult than expected because they underestimate the work that goes into training models properly. It’s not a one-and-done process and developers need to continually request changes to the data being collected as the needs of the data model become clearer. To ensure this happens, organizations will need to build the cost and time needed for proper data collection and testing into their project plans to provide customers with high-quality AI experiences.