By Alan Crameri Technical director at Barrachd, part of Capita
However, it is not a new concept! AI’s roots are in the ‘expert systems’ of the ‘70s and ‘80s, computers that were programmed with a human’s ‘expert’ knowledge in order to allow decision-making based on the available facts. What’s different today, and is enabling this revolution, is the evolution of machine learning systems. No longer are machines just capturing 'explicit’ knowledge (where a human can explain a series of fairly logical steps). They are now developing a ‘tacit’ knowledge - the intuitive, know-how embedded in the human mind. The kind of knowledge that’s hard to describe, let alone transfer.
This machine learning is already all around us, unlocking our phones with a glance or a touch, suggesting music we like to listen to, and teaching cars to drive themselves. Underpinning all this, however, is the explosion of data that’s happening at the same time.
So AI is a journey. And the journey to AI starts with ‘the basics’ of identifying and understanding the data. Where does it reside? How can we access it? We need strong information architecture as the first step on our AI ladder.
There will be 50 billion smart connected devices in the world, all developed to collect, analyse and share data. This data is vital to AI. Machine learning models need data… Just as we humans ‘learn’ our tacit knowledge through our experiences, by attempting a task again and again to gradually improve, ML models need to be ‘trained’.
By the year 2020, it’s estimated that every human being on the planet will be creating 1.7 megabytes of new information every second!
Of course, some data may be ‘difficult’ – it might be unstructured, it may need refinement, it could be in disparate locations and from different sources. So, the next step is to fuse together this data in order to allow analytics tools to find better insight. The next step in the journey is identifying and understanding the patterns and trends in our data withsmart analytics techniques.
Only once these steps of the journey have been completed can we truly progress to AI and machine learning, to gain further insight into the past and future performance of our organisations, and to help us solve business problems more efficiently. But once that journey is complete - the architecture, the data fusion, the analytics solutions – the limits of possibility are only contained by the availability of data.
As experts in data and advanced analytics, Barrachd’s solutions span industries, creating efficiencies and finding insights in business, customer and citizen data. Let’s take an example that is applicable to most organisations, the management of people. For our talent analytics solution, we identified and sourced data from a number of systems – employee and payroll data, absence records, training records, exit interviews, performance ratings, employee engagement surveys.
Next we fused it together to give a complete ‘picture’ of an employee’s interaction with the organisation. This data is then displayed in a number of interactive dashboards that are used by managers of each business area to track their people’s performance, relative to division and organisation averages.
Managers can instantly visualise how they’re performing, and which areas to focus on for improvement. The next stage, however, is to use AI models to ‘predict’ those employees who might need some extra support or intervention – high-performers at risk of leaving, or people showing early signs of declining performance. By utilising statistical modelling techniques, organisations can predict attrition and spot the reasons behind it, allowing them to take action.
Satisfaction, retention, and interaction - increasingly businesses look to social media to track the sentiment and engagement of their relationships with customers and consumers. Yet finding meaningful patterns and insights amongst a continual flow of diverse data can be ‘difficult’.
At Barrachd, we’ve created a social media analytics solution that’s already being used in the financial services, retail and gambling industries, to analyse how customers and consumers view and react to the companies and brands they’re interacting with through social media.
Here, the data is external to the organisations concerned, but our solution collects and interprets targeted data from across different social media platforms, combines and merges it, and extracts relevant meaning, to create an information architecture ‘behind the scenes’. The next stop on the AI journey enables powerful analysis of trends and consumer behaviour over time, allowing organisations to track and forecast customer engagement in real-time.
Social media data isn’t the only source of real time engagement. Customer data is an increasingly rich vein that can be tapped into.
Disney is already collecting location data from wristbands at their attractions, predicting and managing queue lengths (suggesting other rides with shorted queues, or offering food/drink vouchers in busy times to reduced demand).
Infra-red cameras are even watching people in movie theatres and monitoring eye movements and facial expressions to determine engagement and sentiment.
The ability to analyse increasingly creative and diverse data sources to unearth new insights is growing, but the ability to bring together these new, disparate data sources is key to realising their value.
We are working with one of the largest UK police forces to implement an information management solution, fusing data from 18 different areas to provide a unified platform for data analysis, allowing them to view spikes in crime, better distribute resources to manage and prevent criminal activity, and to manage risk and vulnerability.
With access to reliable and accurate data at their fingertips, staff and officers have the tools for better decision-making, not only to fight crime, but to create a more efficient workforce too.
A valuable commodity when forces are being tasked with doing more with less. With this journey underway, the next step is to “predict” the future – when and where crime is likely to happen, or the risk or vulnerability of individuals, allowing the police to direct limited resources as efficiently as possible. Machine learning algorithms can be employed in a variety of ways – to automate facial recognition, to pinpoint crime hotspots, and to identify which people are more likely to reoffend.
AI models are good at learning to recognise patterns. And these patterns aren’t just found in images, but in sound too. Models already exist that can ‘listen’ to the sounds within a city, and detect the sound of a gunshot - a large proportion of which go unreported. Now lamppost manufacturers are building smart street lights, which monitor light, sound, weather and other environmental variants.
However, there is one underlying factor that occurs across every innovative solution – now, and in the future. Data quality. IBM has just launched an AI tool designed to monitor artificial intelligence deployments, and assess accuracy, fairness and bias in the decisions that they make.
In short, AI models monitoring other AI models. Let’s just hope that the data foundation that these are built on is correct … at the end of the day, if the underlying data is flawed, then so will be the AI model, and so will be the AI monitoring the AI! And that’s why the journey to advanced analytics, AI and machine learning is so important.
As technical director at Barrachd, part of Capita, not only does Alan offer strategic leadership and guidance, but he brings over 15 years’ experience in designing and delivering planning, predictive, data warehouse, BI, dashboarding and advanced analytics solutions.
Alan has an MBA from Lancaster University and a degree in Material Science from Clare College, Cambridge.
Read more insights from Alan on our website.
Connect with Alan
Email: Alan.Crameri@capita.co.uk or connect on LinkedIn.