Content Copyright © 2018 Bloor. All Rights Reserved.
Also posted on: Accessibility
Jennifer Roubaud is the VP for UK and Ireland of Dataiku, who market an enterprise analytics platform the Data Science Studio. The platform is fully featured offering connectivity, with connectors to over 25 storage systems, including traditional RDBMS like Oracle, and SQL*Server, and analytics databases like Redshift and Hadoop etc, so full connectivity for the vast majority of users. It then allows the data to be integrated, explored, and enriched with tools such as R, Python and SQL. Then the data can be exposed to machine learning tools such as MLLib, XGboost and Scikit-Learn to create a predictive model, the results, which are visual insights can then be integrated, again via a wide variety of interfaces to a broad range of hosts to provide real-time predictive solutions. So really comprehensive, it looks great, and seems to tick most of the boxes. But I was interested in how people actually get value from these smart pieces of kit, so we explored around the tool, rather than diving into the tool itself.
We started by exploring what was different about Data Science when compared to the traditional BI and data-mining stack that most of us grew up with. Jennifer pointed out that they address data from very different perspectives. The traditional BI stack is there to provide predominantly descriptive analytics. It tells you what happened and adds value by presenting the facts in such a way as to clearly tell the story of what happened and to provide pointers for the SME to ascertain why. Data Science is looking at the data to produce a picture of what is likely to happen, and the likelihood of that predication being reliable. As such it is more explorative, it is providing a recommendation, how that recommendation is used is going to determine the value that is gained, and is another major differentiator from traditional BI. This I found very interesting, and supported things I had heard earlier this week when I sat in on a webinar by Insurance Nexus in the States on “AI, Machine Learning and Chatbots Improving Insurance Profitability & CX”. There we had three senior AI professionals from large US based Insurers explaining what they were doing. They too emphasised the importance of deployment as critical to deriving value.
Jennifer and I discussed this further. From the popular press you could be forgiven for believing that jobs Armageddon is just around the corner and that AI will replace great swathes of knowledge workers alongside robots replacing manual workers. However we have very few examples of technology on its own achieving any great lasting change, so if we are to exploit the possibilities that machine learning can bring what is required. Here I found Jennifer’s insights fascinating. We hear a lot about creating a data culture, and to do that needing a change programme that is far more than just technology. The VPs of the Insurance companies echoed this saying that AI is not about replacement it is about augmentation, and freeing staff from repetitive low value tasks to take on the challenge of facing the disruption facing their industry and turning it into opportunities. Jennifer also spoke of the challenge of taking disruption and instead of letting that generate fear and inactivity, needing to capture the moment and look ahead at what can be achieved by thinking things anew, Jennifer explained that the role of the modern CDO is as much as anything to be a communicator of a vision. Their role is to empower people to do a better job. They have to communicate and they have to be agents of change. What is clear from those that know is that a small scale skunk works looking into machine learning and AI is not going to achieve what is required, it might address a few morsels of low hanging fruit but that is not enough. The CDO has to start at the top and invigorate the CEO and his team to see the reasons to start top down, and use the capability to understand the drivers behind their business, both internal and external, and instead of looking to just refine the wheel to ask is the wheel even the right thing for the challenge that is being faced.
If AI and machine learning is not just about technology, then clearly having machine learning as just another IT tool is insufficient, and we then discussed what skills are required. Again this was reflected in the discussion that was held amongst the Insurance Executives. You need new skills, broader skills. We could see that what was key was abilities to see what was being produced within the context of the business, a number on its own is never enough, what does it mean, what actions are required to leverage it. Key skills that are required are abilities to communicate and to collaborate, skills not often seen in IT departments today. Jennifer then went on to explain how their company has a vision of making the benefits of machine learning available to a far wider audience, including those who have no ability or wish to code, but who are fascinated by the data that reflects their business. We can see that as the existing data is exposed that people start to question very meaningfully the value of the legacy ways and the data that it has captured, its quality and its value. Again the Insurance panel highlighted the challenge they face with the vast majority of their systems being legacy system, which focus on functionality over data being an asset, and how that is going to have to change, if the companies are to meet the challenge of remaining profitable in a changing world.
We then discussed what is the potential for AI to revolutionise the way we do things. This was interesting as the Insurance panel said that it is too early to see yet what direction this is heading in, so they are just in an exploratory phase. Jennifer took a very interesting stance and highlighted the potential that exists for these technologies to change things within society, to recognise vulnerability, to coordinate services to serve the citizen to create a fairer society, and explained that to this end they make their software available to NGOs to encourage them to experiment.
We both agreed we are at the foot of a very steep upward curve of change and that in the next decade endless possibilities will exist if we can see the big picture and can adapt ourselves to the changes. There unfortunately we ran out of time, but clearly Jennifer and Dataiku are leaders not just in the technology, but also in the thought leadership to guide people through the disruption to achieve something meaningful beyond just selling technology. Hopefully we shall get the chance to talk again.