The big picture
Big data, analytics and AI have permeated into every aspect of our lives from the financial institutions we bank with to the supermarket where we pick up our daily groceries. We speak with Shafik Khashouf of SAS, which is helping companies leverage the po
SAS vice president Shafik Khashouf on helping firms leverage big data, analytics and AI
BIG DATA and the predictive analytics models built around it were impacting our lives long before we were even aware of it, operating nondescriptly in the background.
In the 1970s, for example, mail order companies would use big data and algorithmic analytic models to determine which customers they should send their costly and limited supply of mail order catalogues to in order to extract the maximum yield. If you received a mail-order catalogue, you were probably the target of a well-mapped statistical model.
In 1976, tech entrepreneur James Goodnight set up Statistical Analysis System (SAS) in Cary, North Carolina, where it is still headquartered, to help businesses like those mail-order companies develop the required analytic tools.
“When we started 40 years ago, we needed to educate customers on the need to use algorithms for decision making,” says Shafik Khashouf, vice president, professional services and delivery, South Europe, Middle East and Africa at SAS.
“Customers back then were quite resistant to the idea of being aided by computers or even outsourcing their decisions to algorithms. This has now completely changed. We don’t need to educate the market anymore. The market is hungry and in desperate need of AI and big data solutions.”
How exactly do firms like SAS leverage those copious amounts of raw data into realworld solutions for companies? Khashouf explains that a life cycle of data-to-value typically goes through three stages: The data management stage, the discovery stage and lastly the deployment phase.
“At the data stage we help manage the data to provide optimal data quality. At the discovery phase we provide software and methods to draw insights from the data. At the last stage we deploy models to draw predictions from the data you’ve trained the system on,” he says.
The data stage is crucial because SAS can provide its decades-worth of accumulated field knowledge and software tools to help companies determine what data they should be collecting (and what they can filter out) in order to achieve their goals.
The discovery phase traditionally used simplistic linear models or linear regression models to draw insights from the data at hand. Not any longer, though. That’s where companies like SAS have stepped in by integrating what Khashouf refers to as “explainable AI” models into the mix.
Here’s an example of it in a real-world scenario. Let’s consider credit scoring, which used to be a purely statistical model.
If your application is rejected, you are entitled to an explanation from the bank as to why you have been rejected.
But often, the old linear models could not provide those explanations because the decision-making process was inadvertently programmed to arbitrarily reject an application based on maybe a person’s gender or nationality.
Enter AI, which is far more sophisticated than linear systems and correlates data across multiple data points.
“AI provides reason codes as to why you were rejected in order to make sure that the algorithms didn’t rely on biased reasons for those rejections.”
Sticking with the example of the credit application, a factor like your nationality will no longer be the most important criteria in determining your creditworthiness even if the most number of defaulters at the bank you are applying to for credit are indeed your fellow nationals.
“We provide technologies that help you to not only draw insight from that data but to interpret how the AI is learning from the data to reach those decisions,” says Khashouf.
The AI systems in place today delivers not just a decision, but are able to explain how it got there – something that linear models could never do. “We need models to provide explainability. This is where SAS is heavily invested in at the moment.”
Every time AI is deployed there’s always the possibility that it can go rogue. Khashouf cites an example of a Microsoft chatbot, which was released a few years ago and became racist based on the data it was learning, forcing the company to pull the plug on the technology.
“At the discovery phase, SAS puts into place supervised and unsupervised methods to make predictions of what is being learnt out of the data,” says Khashouf. This ensures there’s a balance struck and that the AI models are doing what they’re designed to do.
Worldwide, the International Data Corporation (IDC) expects spending on big data and business analytics to reach $220bn by 2020. In the Middle East and Africa, expenditure on big data analytics (BDA) is expected to reach $3.2bn in the same period with double-digit growth thereafter.
This growth in the big data analytics field can be attributed to the fact that it is being
By 2020, the expenditure on big data analytics (BDA) in the Middle East and Africa is expected to reach $3.2 billion
used by companies spread across sectors. In the marketing field, big data analytics is used to understand behaviour based on a customer’s transactional data, life stage, loyalty factors, and more.
In the healthcare sector, AI and machine learning are used to track patient records and suggest health plans, as well as indexing of insurance information and even detect healthcare fraud.
In the public sector, governments are leaning towards what is being termed as evidence-based policymaking in which governments takes the outcomes received from data subjected to AI models to help it formulate policies on taxation, state benefits, and so on.
Here in the Middle East, Khashouf says that every client they’ve interacted with already has an analytics programme in place or in the process of acquiring one.
“We’ve been working with our customers in this region not just to sell software to them, but also to help them build their competencies, update their skill sets and help them change their organisational culture so that they can rely more on data-driven decision making rather than gut feeling. We’re helping them to develop their human capital.”
With data scandals like Cambridge Analytica still fresh, has SAS and other companies like it faced a ‘data backlash’ of any kind? Khashouf reiterates that SAS doesn’t undertake data harvesting itself but works with data already gathered by companies.
“We have not experienced data backlash because we work with customers to draw intelligence on the data that they have collected. We’re not in the business of selling the insight from that data to other providers, aka the Facebook model. There has been a new regulation that has been enacted recently – for example, the General Data Protection Regulation (GDPR) in the EU – to address data privacy.”
Interestingly, Khashouf says that these new government regulations aren’t constricting SAS’s business model, but instead opening up new business opportunities.
“We are now providing software services to companies centred on customer data protection. We are helping them obtain the stipulated levels of consent from customers during their data collection and to be in a position to provide their customers with ‘the right to be forgotten’ service too.”
The UAE government has realised the benefits of big data access. Dubai authorities are promising access to a wide array o f data by 2021 to both individuals and businesses
As well as regulating the industry, governments from the EU to the UAE are also encouraging open access to data.
“The UAE government has realised the benefits of big data access. Dubai authorities are promising access to a wide array of data by 2021 to both individuals and businesses too. This data will undoubtedly drive the Smart Dubai 2021 initiative and accelerate the leader’s decision to be the happiest and smartest city in the world.”
While it’s easy to celebrate the deeper integration of AI into the data analytics business, there are serious ethical concerns around it.
“For example, when you’re driving a car and you suddenly need to decide whether to hit a pedestrian or crash into the curb, it is you who is making that decision,” says Khashouf.
"But in an autonomous car where it is AI controlling the car, who really is making – and therefore liable – for that decision? Is it the programmer, the company that’s implemented that programme or is it you for not overriding the system?”
“There is a lot of debate around where AI should begin and where it should end. Should it serve as a decision support system and leave the ultimate decisions to humans? The line is being blurred very quickly between what is decision support and where the system is being allowed to make decisions for you.”
Companies like SAS that find themselves in the middle of the AI ethics debate are doing all that they can to mitigate the ill effects of the technology.
“During the deployment phase we make sure that the predictions made by AI are safe, auditable and traceable. We test them with a wide variety of random situations so that we understand the first order, second order and third order implications of the decisions being made.”
And after over four decades in the business, SAS has had ringside seats to the evolution of the data industry.
“Access to cheaper storage and faster computing power, for example, has changed our industry to something completely different to what it was even three years ago,” explains the VP.
By working across an international set of clients, SAS has also seen what happens to those companies that rely on data analytics models and those who don’t.
“Companies today are using AI to acquire customers more intelligently. Those companies that have this data model in place are increasing their revenues faster than their competitors. Companies that ignore the competitive edge of data will lag,” says Khashouf.
“The benefits you can extract from AI are numerous – we’ve not even scratched the surface yet.”