AI often conjures up an image of an age when machines act like humans. Some find this exciting, others terrifying, but most find it too far fetched to give it serious thought.
Indeed, although the day in which an AI is capable of fully replicating the human brain is still a way off, advances in deep learning and neural networks are making exciting progress in replicating the cognitive operation of our brains and ways of thinking. Though human thinking cannot be fully mimicked, these tremendous advances in AI technology shouldn’t be ignored and their disruptive potential.
AI, automation, and decision-making
Anyone considering where their business will be in ten years should take this very seriously. At its most obvious level, it creates opportunities for more informed, intelligent business decisions and automating of common work. But it also changes the dynamic of business and nature of work. Decisions makers, operation managers and R&D leaders will have to get used to AI being a part of their business processes.
Example: Enfield Council
A recent example of an organisation implementing AI is Enfield Council. The council announced it would be using the Amelia AI software in a customer facing role, helping callers locate information and complete standard applications, and simplify some of the council’s internal processes.
Within businesses, AI is being used to spot fraudulent behaviour, improve customer targeting, and even diagnose illnesses and recommend treatment plans. We only expect to hear more such stories.
The importance of data science specialists
As well as automating repetitive customer services tasks, we also expect that large teams of tech specialists and programmers will be replaced by small teams of experts who can train AI platforms to perform business tasks. For example, much of the IT work currently outsourced, such as legacy code maintenance, could easily be completed by an AI in the near future.
This, in turn, will lower the barrier to entry, meaning smaller organisations can compete with big players by using AI to automate operations and tasks previously requiring people, resources and technology outside of their budget.
Example: Machine learning in finance
Data analytics is already transforming business in similar ways. But the limitations of data analytics platforms are that they focus on spotting correlations, which may or may not have causal connections.
For example, an off the shelf analytics black box may tell you that a series of credit card transactions abroad are fraudulent 80% of the time, but that's frustrating for the 20% who have their card blocked after a holiday spending splurge.
AI, machine learning and neural networks, on the other hand, work by processing data and improving themselves based upon it. They can look at data much more intelligently and learn which particular set of correlations are likely to be unrelated or outliers.
In our credit card example, it can, therefore, identify, whether in this instance, for this individual person, the transaction is likely to be actually fraudulent or not. Not just whether it correlates to a typical pattern of fraud which has been averaged across a diverse population.
Black box analytics platforms vs. AI
To be useful, black box analytics platforms rely on reliable, clean and normalised input data sets. This is a lengthy, largely manual process which requires significant statistical expertise. Artificial intelligence, however, can look at data and over time, learn to correct for outliers, duplication, heterogeneous data formats, etc.
This enables people who don't have immediate data skills to benefit from the insights held in their data and allows for AI tools to be created to process tasks in real time; from data processing to guiding customers through an application process, and identify the best possible responses and outcomes.
The upshot is yet another lowered barrier of entry for businesses. Where once only large enterprises could afford the skills and technology to generate the kind of business-changing insights and automation that data can offer, now much smaller organisations are also able to do so. It makes it much easier for cutting edge start-ups - in areas such as biotech, clean energy, healthcare as well as new digital businesses - to scale quickly and compete with larger players with much lower time and financial investments.
Of course, these tools still have to be built and trained, and this requires considerable specialist expertise. But once the model is established, the learning and data cleaning can largely be automated. The barriers to actually using and benefiting from data become much lower.
A startup could employ a small team of data scientists to build their AI platform in a scalable cloud service, then the business function could use it without the need for specialist skills. For heavy AI users, even this process can be partly automated - Facebook builds so many AI services, that it has built an AI that builds AI.
Preparing for intelligent AI and data science
For the IT and software industry, this is a big change they need to be ready for. AI will be able to do much of the programming and routine work currently done manually. Building and training AI models will surpass traditional programming skills.
We still need extremely smart, talented people who understand the data and how to prepare it for use, what it means for the business, and how to effectively integrate AI solutions into enterprises. But we will no longer need armies of them. The focus of IT skills will shift from programming systems to training them and using them to inform, and sometimes automate, business and operational decisions. A few very talented people will be able to bring huge value to the organisation, which can then be exploited by business people without extensive technical expertise. This marks a significant shift in skill-sets that businesses need to be ready for.
This is great news for innovators who will have access to the advanced IT capabilities of large companies for a fraction of the cost. But companies which rely solely on their scale as a barrier to entry for competitors will need to look at how their own data can keep them ahead of the game in a rapidly levelling playing field.