The data scientist in the age of AI and IoT
- 15 November, 2017 06:35
Technologies are changing the jobs of data scientists from being crafters and programmers of structured data, into an orchestrator of complex automated capabilities
Seven years ago, Dr Thomas Davenport said the data scientist is “the sexiest job of the 21st century” and also the “most in demand”.
He has not tempered his views on the importance of the role, but says the data scientist job has to change to continue to be viable.
“Data scientists now need to make use of newer, more automated tools for machine learning and artificial intelligence and so on,” says Davenport, who teaches analytics/big data in executive programmes at Babson, Harvard Business School, and MIT Sloan School.
These technologies “can radically improve the productivity of an individual data scientist,” says Davenport, who is also a co-founder of the International Institute for Analytics and senior adviser to Deloitte Analytics.
“That changes their job from being crafters and programmers of structured data, into an orchestrator of complex automated capabilities."
He notes there is still a shortage of data scientists, though the situation “has got a lot better”.
Davenport pioneered the concept of 'Competing on Analytics' that he dealt with in a book with the same title, published with co-author Jeanne G. Harris 10 years ago.
What then has changed since his book was released? Davenport states that different analytical technologies as well as data sources have come of age over the past decade.
He believes there are four eras of analytics, from traditional or Analytics 1.0 (internal and descriptive reporting), to the big data era of the early 2000s (Analytics 2.0), to the data economy as analytics became core to the business (analytics 3.0), and the automation era (Analytics 4.0).
The fourth phase is about autonomous cognitive and artificial intelligence-related activities.
“I think it has replaced big data as the most exciting thing happening in the world of analytics," he says.
“That means in many cases, the models are going to be created with some degree of automation and the operation is going to be not by humans necessarily, but by machines making decisions and [taking] action.”
Into the analytics economy
An important development has been the advance of the “analytics economy”.
“It is a SAS term,” he says, referring to the analytics vendor. “It means a large number of decisions and actions within organisations are based on analytics, while much of your success is based on your ability to manage, understand and analyse data.
“I think it is not that different from what I called 'Competing on Analytics' 10 years ago.”
During that time, he defined the capabilities of analytics competitors.
They were really good at managing data, had high quality data, were able to integrate their data effectively and some began to use external data, took an enterprise approach to analytics (“They did not have a lot of little silos with no sharing of expertise or technology”) and had strong leadership.
These leaders appreciated the value of data-driven decision making and really made the resources available to build up analytical capability.
He says the key to his original model was having targets, or clarity where to apply analytics. It could be around customer engagement and loyalty, or for risks.
"Eventually, you can do multiple things, but you need to start with initial targets," he states.
“When I wrote 10 years ago, everybody had software, a statistical package of some sort and an enterprise data warehouse,” he says. “Now it is a much more complex technical environment, so you need to manage that well.
“Companies that put all of those capabilities together, would really succeed in the analytics economy.”
The companies that tend to be best in it are those that have a lot of data, like online firms, financial services firms and telecommunications, he says.
They are able, among others, to get to the best customers and offer the right price, optimise old processes and create new ones, embed data and analytics into products and services, use analytics for innovation and not just to cut costs, and predict and respond to changing business conditions, he says.
The only risk is if you are unwilling to learn new skills.
How about small and medium enterprises (SMEs)?
Davenport says for these businesses, there is a lot of free software and external data available, and hardware is much more accessible in the cloud than it ever was before.
“The things that tend to be in relatively short supply for those companies are awareness and imagination, simply because they don't tend to have people who are given the time and resources to go out and investigate new technologies,” he states.
“If you are a small business, you are focused in many ways on getting the product out the door, giving customers what they need today, not thinking 'where is my business going to go in the future?'”
He has come across SMEs who do investigate what they can do with analytics and have started to put sensors into their products and think about AI. “But there is not just very many of them."
“And that is one of the factors that has made for greater inequality for both corporate and individual levels,” he says. “The rich got richer.”
Walmart, for instance, became a retail giant in part because it invested heavily in information capabilities and was able to manage its business much more efficiently and effectively, he says.
Many startups, meanwhile, are taking advantage of open source technologies for data analysis and management.
“Now, particularly for any online business, it is so easy and so important to get a sense of who your customers are, what are they looking at, what are they interested in. Entrepreneurs are beginning to be more data and analytics oriented,” he says.
“It is much easier to create entrepreneurial ventures now, because so much of the capability is widely available.”
He cites the case of Harry’s and The Dollar Shave Club, which are causing a great deal of concern to manufacturers that have massive R&D budgets and own manufacturing plants.
The two companies contract manufacturing to China and sell online. They can charge a lot less for their razors, so they have profited at the expense of the big players, he says.
“There are some new possibilities for SMEs in part because of widespread availability of technology and also because of the commoditisation of manufacturing. Some other key processes of supply chains can be outsourced.
“It is an exciting time to be an entrepreneur, certainly.”
An issue that worries him is “the uncontrolled proliferation and bad management of data”.
He believes ultimately data about individuals will need to belong to those individuals and they should have the right to determine how their data is used, though that is really not the case today.
In most places, he says, individuals don't have the right to determine - whether it is their telco, financial or shopping data - what happens to it.
Davenport also talks about the concern of people losing jobs due to automation.
He says there is also a lot of technologies available - machine learning, automation, deep learning - and the challenge for companies is matching these to use cases.
When asked about RPA (robotic process automation), he says, “It is arguably the dumbest of the smart machines. It is a step above Excel macros.”
Though in terms of capability, he says it tends to have the highest return on investment and the most rapid implementation of cognitive or semi-cognitive technologies.
“It is getting slowly smarter,” he says. “Vendors are adding more cognitive capabilities."
He believes the biggest impact of RPA is on outsourcing firms.
There has been unrest in some Indian outsourcing firms where people are getting laid off when companies started using the technology. “It is quite unfortunate because those companies have been a route to the middle class for a lot of people in other countries.”
He says the affected outsourcing firm will have to rapidly learn how to add value to smart machines. “How do we configure and install lots of RPA tools?”
Some banks he has worked with intend to eliminate offshore outsourcing as much as possible through technologies like RPA, but it is early days.
“The jobs that tend to be automated are the highly structured and repetitive ones. So again, you want to get out of that role if there is way to do it.”
In regards to automation, he is often asked, 'what does it mean for human jobs?'
He considered this while researching the book, Only Humans Need Apply: Winners and Losers in the Age of Smart Machines with co-author Julia Kirby, that was published in 2016.
“I don't think we are going to see massive unemployment,” he says. “The only risk is if you are unwilling to learn new skills.
“If you are unwilling to learn new skills, I think it is quite possible you won’t be able to work effectively alongside smart machines and add value to them and so on.”
This happened with the arrival of automated textile machinery in New England, where he lives. “Apparently there were always plenty of jobs for people who understood how these machines worked and who could configure them, install them and improve them and fix them.
“It does mean having to acquire a lot of new skills,” says Davenport.
“The same thing will be true for all the kinds of smart machines that we are talking about today.”
Divina Paredes interviewed Dr Thomas Davenport at the 2017 Analytics Experience in Amsterdam which she attended as a guest of SAS.
Send news tips and comments to divina_paredes@idg.co.nz
Follow Divina Paredes on Twitter: @divinap
Follow CIO New Zealand on Twitter:@cio_nz
