Recently I went on atrip to a dock. Not any dock, of course, but The Dock, Accenture’s recently opened design and development centre for what it calls 'Industry x.0' thinking, experimentation and development.
The objective was to get a heads up on the type of work being undertaken there, not least because one of the hot topics for the second half of last year was Industry 4.0. This obviously begged the question of why Accenture was not following the already established nomenclature, to which the answer was that it might Industry 4.0 right now, but a 5.0, 6.0 and onwards would surely follow along in short order.
The company sees the role of The Dock being to identify and anticipate what can be achieved with the coming technologies. This can be either off its own bat, or through involvement with client businesses in their applied researches that, hopefully, may lead to working solutions.
Its operations are divided into six main areas, though there is much cross-over between them. These are its own R&D resources; a Ventures operation which handles the development of partnerships and investments; Laboratories, where tech developments are incubated and prototyped; Studios, where the prototypes are developed into contender products or services; Innovation Centres,where client businesses can work with Accenture teams to demo and scale operations; and Delivery Centres to work on the industrialisation of worthwhile innovations.
The IP rights to what comes out at the end can be assigned on an `it depends’ basis. If the development has potential application horizontally, even in just one market sector, Accenture will aim to hold on to the IP. If it is a specific development for the specific needs of a specific customer, then the customer will usually end up with the IP.
AI et al means, start talking
This means, to be sure, that there is a wealth of hi-tech software and gizmos being created, developed and tested to be seen all around the place, and the building itself readily shows off some interesting ideas on architectural agility and operational flexibility. But in the end, as industry starts to succumb to the growing vendor hyperbole of `get your AI/Automation/et al here and get it now to beat the competition’, something more subtle emerges as an underlying message.
The key learning to be found in The Dock is really quite straight forward: talk lots and don’t rush to apply AI, automation, machine learning and the rest. This is not least because there is much in these early days of such untested technologies that will attract the law of unintended consequences – for good or ill.
Here are just a couple of examples of that which emerged during the discussions with Accenture staff at The Dock. On the good' side is the Parcel Integrity Project, involving an extended RFID device that incorporates accelerometers and orientation devices which can report not only a parcel’s location but also on its physical status: has it been dropped or is it in the correct orientation. Any failures will be traceable.
However, the use of such a device could readily be extended to whole containers as well as individual packages. These can suffer rough handling, and even fall overboard. Logistics service providers are already looking at this potential. But they have added another application: the internal cooling systems fitted in some containers are now a theft target housed in an easy and often poorly protected target. So the plan is to use the `parcel tags’ on the cooling systems themselves.
On the 'ill’ side, the company was approached to develop a machine reasoning system for propensity modelling for fraud investigations. The goal was to identify who should be investigated in insurance, tax and benefits frauds. The project failed however, because available data was too sparse. In addition, some techniques for gathering information could not easily be explained and therefore could not be justified sufficiently to get approval. It proved impossible to rank people for investigation against understood parameters.
This, however, proved to be a good example of where talking through subjects and processes, together with the wider implications, is an essential start point. Using sparse data may be acceptable if the task concerns a less serious application, such as the placement of an advertisement (a common practical application). But if it is investigation for fraud the rules are significant different, and the evidence/justifications have to be to a much more stringent level.
Find a data scientist
A prime area for discussion for any business, therefore, is the quality of the data used to drive any process. Accenture’s answer is that businesses need to discuss this issue with data scientists, particularly when it comes to identifying which data sources are the important or valid ones. If the important aspect of AI is the `decision’ – the result that drives the next stage of a process, be that small or large – then ensuring that the right data sources are selected as sources is a vital step. This is also a point where it can be said that humans still have to be in the loop.
It also becoming necessary that AI can explain itself in some way. While there are times when a process can be automated, there are also times when a human in the link is essential, and getting the touchpoint for the human interaction with the data is a critical factor.
A good example of this is in tele-medicine. Here there is a complex mixture of data science, medical knowledge, technology support (ie instrumentation etc) and user confidence. There is a tendency to lack confidence in the results and diagnoses using AI systems and therefore an urge to employ more tools etc as a cross/double/triple check on the results. This is where the Studio capabilities come into play because it gives Accenture the opportunity to build working examples of planned AI tools and services and work with end users to identify what works, what does not work, and what is still missing.
The company also works with clients on internal issues such as the inevitable staff re-skilling that the new technologies require. This is now a particular issue as a McKinsey report published towards the end of last year suggested 375 million people will have their job changed by tech, and up to 800 million could find their job 'disappeared'.
The company has created a program, Skills to Succeed, which is heavily focused on re-training. It also gets involved in developing thinking on what it means to be a responsible business, and it plans to publish material on this development program sometime this year, though it states it is not intending this to be a 'Bible of Automation’.
There is also the issue of managing the speed of change, not least because some of the changes look attractive to businesses but the 'glamour’ may yet prove to be superficial, with underlying unintended consequences that are not always apparent. But if discussion with users can help identify even one or two of them that can be a real benefit, not least by getting the user into thre mindset of thinking in such terms.
Away with old complacencies
Digitalisation is obviously also good for building and enhancing customer experience, but companies need to understand the importance of this. One problem with the old complacency is that many of those companies have stopped listening to customers, or even their own staff, sometimes years ago. Now they have to learn to listen to customers and staff again, and learn how to interact with them rather than shouting at them through a variety of channels.
There is a belief at The Dock that there is a secret sauce in every company that stops them thinking about how they can use automation to do what they were doing three years ago and instead make them courageous, willing to change, agile enough to make the changes quickly – and to keep making them.
ID2020 Alliance– a United Nations-led global partnership to create an ID program for refugees that have no identification papers with them. How do they prove who they are by systems that use digitalisation to create a personal, portable identity. Over one billion people globally, including many millions of children, women and refugees, lack any form of oﬃcially recognised identiﬁcation.
Without an identity, individuals are often invisible—unable to vote, access healthcare, open a bank account, or receive an education—and bear higher risk for traﬃcking. Without accurate population data, public and private organizations struggle to broadly and accurately deliver the most basic human services.
As a member of this Alliance, Accenture is involved in the development of the biometric identity system, the use of blockchain as the transaction system, and the development of the required database. Much of the prototype has been designed at the Dock, bringing together designers, technical architects, developers.
The prototyping process is mainly about demonstrating the art of the possible, and the company accepts the issue of scaling this to a global operation has still to be faced. Even the issues of co-operation and interoperability between different countries and continents have yet to be surmounted.
The reality, however, is that it can take several years for refugees to substantiate their ID and their story, and they are unable to work in that time. This is itself a good example of the complexity of issues that come attached to many AI/Automation/Machine Learning projects and why just about any such project will require much more talking and thinking than any IT project before.
This will need to be approved by and operating in close conjunction with the immigration authorities of every country around the world, such as the UK Home Office. Each will need to approve the project, and the development program will need to adapt the technologies to accommodate their individual aspirations and misgivings. That will involve a good deal of talk.
Hardly `hot news’ of course, and almost into the realm of being blindingly obvious, but when I get yet another press release from a vendor in the AI space asking me to tell you, dear reader, to get on with buying AI products (preferably theirs, of course), all I can envisage is the emergence of AI horror stories late this year when the law of unintended consequences starts to have a field day on some early implementations. Early adopters often get bitten.
So taking some time to kick things over – internally and with third parties – is probably a good idea. You will always be told that being first with something will beat the competition. But not if they get it right and you are still digging your hole