computer parts

Google caused a stir when it open sourced its TensorFlow software back in November 2015, and the technology is starting to make its way into the mainstream.

The machine learning software library is the next generation of DistBelief, which was internally developed by the Google Brain team at the search giant for a multitude of tasks such as image search and improving its speech recognition algorithms.

TensorFlow is a deep neural network, so it learns to perform a task through positive reinforcement and works through layers of data (nodes) to help it determine the correct outcome.

Read next: How Google plans to bring AI and machine learning to the enterprise

By open sourcing the TensorFlow library of machine learning code, Google is facilitating the simpler construction, training and deployment of complex deep neural nets. TensorFlow doesn't exactly give every developer the ability to harness machine learning but it does provide both a Python and C/C++ API to link into a developer’s program.

This sort of machine learning work used to be exclusively in the realm of research, but with the open sourcing of software like TensorFlow the enterprise is increasingly able to utilise its powerful capabilities on their own data, operating it on cheap cloud computing.

The TensorFlow libraries make it far easier to incorporate self-learning elements and AI features like speech recognition, __computer vision or natural language processing into applications.

Read next: Machine learning APIs and frameworks 2017: 14 machine learning tools for data scientists and developers

TensorFlow certainly isn't the only deep learning library out there, but just as there are other search engines besides Google, it is widely regarded as the best out there. As alternatives go there is Torch, built by Swiss researchers, and Caffe, which came out of The University of California Berkley and has been built on by Facebook for its recent release of Caffe2.

In terms of limitations Google still only allows TensorFlow to be deployed on a single machine, granted with multiple GPUs, which can limit the scale businesses can use the tool. There are naturally workarounds for this, but they require expertise, time and money.

TensorFlow case study: Ocado

According to the TensorFlow website some of the biggest companies in the world are using the software library, such as Airbnb, Airbus, Dropbox, Snapchat and Uber (although they might not be using it in the most appropriate way).

However, finding actual case studies can be an uphill battle. One company that isn't shy about its use of TensorFlow is UK online supermarket Ocado.

The data science team there is using the libraries for everything from routing algorithms for its robots to move around warehouses, to improving its existing features like demand forecasting, which is currently based on decade-old linear regression models, and predictively suggesting items to add to your basket depending on past shopping habits.

Ocado actually first used TensorFlow to manage the flow of emails coming into its contact centre. Before TensorFlow the contact centre was dealing with emails on a first-in basis, with no sense of prioritisation. This would become an issue if bad weather hit and the volume of emails went up by three to four times and where issues with delivery would be much more pressing than something like website issues.

Dan Nelson, head of data at Ocado Technology told Computerworld UK: "So the business wanted to solve how to get a better grasp of what emails to deal with first, and that presented itself as a natural language problem," Nelson said. "So we wanted to intercept email as it arrived, understand the sentiment of it and also what they were talking about."

He says that the problem was too bespoke for Google's off the shelf natural language processing offering, so they turned to the TensorFlow library to solve the issue.

Ocado is a Google shop anyway, using the search giant's Big Query in the Google Cloud for much of its query and storage needs. Nelson said that deploying TensorFlow with Google certainly makes things far easier. "If we didn't use any cloud provider you would have to provision some pretty serious hardware, but they will still execute there. You don't have to use Google, but the fact that Tensor came out of there certainly helps for that process," he said.

Read next: Ocado to replace barcode scanning with AI 'vision' to speed packing processes

Speaking about the tool more generally Nelson had a dose of realism to offer: "Essentially TensorFlow allows you to access through the libraries different learning models. You need to train and adapt them, but it accelerates your learning.

"TensorFlow doesn't solve the problem, but gives you the toolkit to abstract away from academics of a convolutional neural net and use one to solve your problem," he said.

After years of lagging behind other industries, cloud computing projects are now becoming commonplace in the finance sector. But while such initiatives have generally centred around non-critical systems such as email and customer relationship management (CRM), banks are set to start entrusting core systems to public cloud providers too.

According to John Schlesinger, chief architect at Temenos - which sells banking software to some of the world’s largest lenders - the majority of new core banking projects launched by the end of this decade will be in the cloud.

“We think that by 2020 most core banking initiatives will be in the cloud,” Schlesinger told Computerworld UK at the Temenos Community Forum in Lisbon this week. He claims it is a case of ‘when’ and not ‘if’ banks adopt the technology at the heart of their business. Temenos itself has already made headway with a small number of regulated and lightly-regulated micro-finance banks that are currently using its software as a service platform.

At this stage the number of banks running core banking in the cloud is small. But while adoption is currently low, it is an increase from just one percent the previous year, and Schlesinger expects that this will grow quickly.

“We believe that by 2020, which will be after PSD2 and after Instant SEPA, that the trickle will become a tsunami,” he said.

The first steps are already being made in the UK. Newly launched OakNorth was one to be the first UK bank to run its Mambu core banking systems in the cloud last year, migrating the platform to Amazon Web Services. Digital retail challenger bank Monzo, which built its core systems from the ground up using modern infrastructure software tools such as containers, also moved its core systems. Meanwhile, Metro Bank recently migrated its core infrastructure into a private cloud managed by Rackspace.

The approach makes sense for smaller challenger banks, lowering infrastructure costs and helping removing one of the barriers to entry in the market. “What cloud does is make everything cheaper. The on-premise story dramatically would lose out to the cloud story, given the economics of manufacturing of machines.” 

“A new initiative today like Atom Bank or Starling Bank may not be in the cloud, but by 2020 it would be,” he added.

For the UK, regulation is no longer a barrier that prevents data from being held in the data centres of AWS, Azure and others, with the UK’s Financial Conduct Authority publishing guidance for lenders last year. “We don't think in the UK there is a regulatory problem, the problem is we are a very consolidated banking market, and for a bank to be in the cloud it would have to be a new initiative.”

Big Four held back by legacy infrastructure

For the Big Four UK banks the situation is different. Should they wish to move core systems out of their data centre in future, the main barrier is their reliance on complex legacy infrastructure, built up over decades in many cases.

“The big banks have a huge problem because for them it is a huge project to move their core banking,” Schlesinger said. “I think the only way they can do it is a build and migrate strategy and what they build should be built on top of infrastructure as a service. So over a fifteen-year period migrating core banking into the cloud.

“They’ll start the process in a couple of years’ time and they won’t finish it for a decade. So the data centres will be there running the existing systems but on a legacy basis.”

Cloud flexibility

There are a number of potential benefits of moving core banking to the public cloud, alongside lower costs. The flexibility of cloud computing helps lenders move more quickly to compete with nimble finch competitors, rather than relying on legacy infrastructure. There are also benefits around security for many banks, with the hyper scale cloud providers able to employ huge teams of security specialists that dwarf the number available to all but the biggest financial firms in order to protect system.

There are also numerous options in terms of routes to the cloud - IaaS, PaaS, SaaS and public, private, managed - and these will appeal to different banks depending on a range of factors, said Schelsinger. This includes the individual bank’s size and the geographies they operate in.

“I don’t think all the banks will do the same things,” said Schlesinger. “So for a tier one initiative, they will buy infrastructure and they will build platforms and software as service on it. That is the Tier 1 bank at home - so Chase in the US, or HSBC in the UK [for example]. But a Tier 1 bank out of geography looks more like a Tier 2 bank. I see Tier 2 banks renting platform as a service, and actually we are seeing that already in some of the RFPs we are getting.

"For all other banks, that's the Tier 3, 4 and 5, they will be buying software as a service, running very, very low entry costs."

In other news at the Temenos’ 18th annual European customer event, the Swiss vendor announced an update to its MarketPlace platform, launched last year to help connect bank and fintech firm services ahead of the PSD2 regulations.  Temenos has added new functions to the MarketPlace such as a Sandbox platform as a service non-production environment to make it easier to test MarketPlace services.  

Qualtrics has gone from the Silicon Slopes of Utah, building software for university researchers, to a platform of tools used by the world's biggest brands to survey their customers and employees.

As the story goes, Qualtrics started from the Smith family basement in Provo, Utah. Ryan Smith, cofounder and current CEO, started building research software as a project with his seriously ill father, a marketing professor at Utah's Brigham Young University (BYU), back in 2002.

Qualtrics started by targeting what Ryan's father knew best, the academic sector, something Smith joked about being "a horrible business model" because universities "have no money".

The original idea though was to create a Software as a Service (SaaS) tool that would allow anyone to conduct research quickly and easily. Then, as the world dealt with the fallout from the 2008 financial crash, Smith and his small team saw that customers were using the platform for something else: they had started to adopt the technology for customer experience management. In other words they were doing market instead of academic research on the platform.

Around 2010, as purse strings tightened and consumers became less brand loyal, businesses had to start competing more in the customer experience space. So although Qualtrics wasn't originally built for that, a trend was starting to form, and the team went about architecting the platform around this customer experience.

What is the Qualtrics Customer Experience Platform?

Smith told this founding story again on stage at the Qualtrics Converge Europe conference in London this week as a means of mapping how the company has evolved, as is par for the course in the tech world, into a platform – the Experience Management (XM) platform.

The journey is a familiar one: build the technology, adapt it to how customers are using it, and converge the platform back into an easy to understand package. Broken up into its component parts this consists of the customer experience, employee experience, product experience and brand experience products, all built on top of the core Qualtrics research engine.

The aim is to allow anyone within a business to gathering insights for things like customer or employee satisfaction, product testing, brand or pricing research; analyse the needs and trends, report it back and drive continuous improvement, all in a simple web-based platform.

Qualtrics wants its XM platform to be the repository for rich experiential data, instead of the static operational data businesses tend to run on. "So human factors, beliefs, emotions," Smith said. "Why things are happening and what will happen next."

Smith likes to benchmark Qualtrics against Salesforce, the SaaS leader in the customer relationship management (CRM) market. He claimed that the new experience management platform will be the "system of record for experience data".

The platform then has a whole host of smart features like sentiment analysis and statistical regression techniques to predict and push the highest priority cases to the right person to take action, not unlike Salesforce's Einstein features.

Read next: How Salesforce brought artificial intelligence and machine learning into its products with Einstein

Closing the experience gap

Smith joked on stage that you only have to look back a couple of weeks – cough, United – to see how devastating poor customer experience can be to a brand, and the importance of getting on top of what your customers think.

Smith called this the 'experience gap' and pointed to Bain and Company figures which show that 80 percent of firms believed they delivered a superior experience to their customers, but only eight percent of customers agreed.

Smith said that the Qualtrics platform, by bringing together the four core satisfaction elements of customer, product, brand and employees, helps companies close this gap.

"This is not just some new features," he said of the Experience Management platform. "We have been working on this for five years and how you manage those four core elements and how you delight customers by setting it up at every touchpoint and how you build iconic brands."

Qualtrics customer case studies

As the image below attests, Qualtrics has a seriously impressive stable of enterprise customers, and the pool is pretty vertical-agnostic.

Speaking on stage during Converge, Alison Windon from Allianz insurance said that following the arrival of a new group CEO, with a driving focus on the customer, the insurer "needed technology as a key enabler" to drive that cultural change.

"[Allianz needed] a tool to do heavy lifting for feedback, analysing it for key insights and facilitate key action planning at the client level," Windon said. "We needed to get the right information to the right people, at the right time, to drive impact. And we needed it to be flexible, self service and scalable globally. We found the right partner in Qualtrics."

After 18 months Allianz is now using Qualtrics across 22 geographies to get feedback from more than 12,000 customers.

Adidas took the opposite approach, in that it uses Qualtrics to take its expertise in customer experience to deliver a better employee experience.

Read next: How Adidas is bringing its customer experience approach to employee feedback

Qualtrics IPO?

Utah startups tend to raise less capital than those in Silicon Valley, which means they often grow at a slower, steadier pace. This genuine bootstrapping tends to give these companies a more genuine appreciation for their customers than the lip service you typically get from other B2B startups though.

Qualtrics now counts 8,500 customers globally and although progress has been a tad slower in Europe, it certainly has more of a presence than the average person might be aware of. After opening its first European office four years ago in Dublin, Qualtrics now has offices in Ireland, London, Munich and soon to be in Paris, supporting 1,200 European customers.

The money has followed, and in a major way, with the company securing a $180 million (£140 million) funding round earlier this month at a valuation of $2.5 billion (£1.94 billion) from Valley big hitters like Insight Venture Partners, Accel and Sequoia Capital, as it pushes towards growth and an eventual IPO.

Smith told TechCrunch in the aftermath of the funding round as much: "Going public is super easy to do. Just file the S-1 and we're out. It's about being public and how that works and getting the house in order to make sure that that's the case. We're going to be a great public company. We're going public."

Unibet parent company Kindred Group is turning to machine learning to recognise patterns of problem gambling across its suite of betting and casino websites.

Kindred Futures is the group's innovation wing, tasked with bringing in entrepreneurs and experts on how to "co-create the future of gambling," according to the organisation's website. It recently held a working group that brought in academics and figures from the technology industry to discuss how machine learning and AI might be able to help reduce problem gambling.

According to statistics from the Gambling Commission published by NHS Digital in 2012, 62 percent of all people gambled in that year, 0.5 percent of people identified as problem gamblers, and 4.3 percent of people were described as being “at low or moderate risk of developing problems with their gambling”. A combined report is scheduled to be published in spring this year.

Kindred Group already has a proprietary system in place – Player Safety Early Detection System, or PS-EDS – designed to raise red flags if players are thought to be at risk. The head of Kindred Futures, Will Mace, describes the existing system as "industry-leading" but also "rudimentary".

"It's a set of rules to do with behaviour – chasing losses, chasing wins, significant changes in your deposit pattern, that sort of thing," Mace explains. "But there can be loads of explanations for those – maybe you bet a couple of quid here and there, then all of a sudden you start betting hundreds of pounds. This would probably raise a flag, and we'd look to see if we can understand what's going on."

At present, the responsible gambling department make joint decisions on when they should intervene. Players aren't aware of the system until somebody contacts them.

"We record all transactional behaviour on our sites for obvious regulatory and operational reasons so when those things coincide and hit a rule we decide what to do about it," he says. "[But] if you were playing on our sites you wouldn't know anything that was happening, and then when a flag was raised you wouldn't know a flag was raised until we decided it was the right thing to do to get in touch with you."

Mace hopes that a machine learning-led approach that combines different data sets could augment the existing platform and even help automate the process to some degree.

A problem is that there are no set guidelines in place – everyone's circumstance is different, their habits will be different, and if they were really set on circumventing risk-reducing measures they could simply go to another website or walk into a bricks-and-mortar betting shop.

"User behaviour takes many forms – the interesting thing in this particular case is there's no single definition of problem gambling," Mace says. "What's a gambling pattern to me might be a problem but to you it might just be your normal form of entertainment. There is no 'if you do X and Y' you have a problem, everyone is different – that's why the more rudimentary methods have their shortcomings."

It's early days for Kindred Group, and Mace does not yet know what the platform will look like or if a machine learning integration will be successful. But the company is now in discussion with one of the startups it invited in for talks and is in the design phase.

Addressing problematic behaviour for online gambling is something the gaming companies will have to pay closer attention to whether they like it or not due to regulations from Europe, according to Mace.

"The regulators are increasingly insistent upon companies having the right, responsible gambling strategies, attitudes and approaches in order to both obtain and maintain a licence," Mace explains. "That's relatively new."

But for a truly effective approach to combating problem gambling, there would need to be some degree of collaboration across the betting websites. This might look like regulators pushing the operators to run a cross-website semi-anonymised way to pool their knowledge.

"It's definitely a challenge to an individual company like us trying to detect problem behaviour when you can just as easily bet with someone else," says Mace. "There's all sorts of challenges that we're taking steps towards seeing what we can do, and hopefully developing momentum from the industry to come together on it.

"Until there's collaboration across those sites nobody's going to have a complete picture of your betting behaviour, and therefore a lot of people who do have a problem are going to go unnoticed even if you bring in non-betting behaviour sources."

Adidas is taking its approach to gathering and responding to consumer feedback and applying it internally for employee feedback and satisfaction with a dynamic, monthly, mobile-first approach.

When director of people analytics Stefan Hierl started at Adidas Group, he found himself wondering why the company was using sophisticated resources to monitor consumer feedback but "archaic methods" internally. "That didn't make any sense to me," he says.

So Hierl went about moving away from the monolithic employee engagement survey of 80 questions, followed by months of analysis and reporting cycles. Now Adidas wants to create more of a constant feedback loop for its employees that better resembles the way it engages with customers.

Working with customer experience software vendor Qualtrics, Adidas is able to capture this feedback from a dedicated mobile app to gather net promoter score (NPS) data from employees.

The new approach is called "People Pulse" and is a fully branded mobile app for internal use, which runs on monthly cycles to capture both open text and 1-10 rating survey responses in a format that takes just five minutes to complete. This is important for a company like Adidas that has a lot of young employees who, according to HR Director Tony Cooke, "won't read email".

The flexibility of the Qualtrics platform means that Adidas can adapt the questions in People Pulse depending on changing business factors, like a change of CEO for example, on the fly.

Employees will be asked how likely they are to recommend working at Adidas and then two open-text questions on what is best and what could be improved about working there.

The respondent will then instantly see the results. "We provide real-time feedback to stakeholders and employees for confidence and trust that we are listening," Hierl says.

Adidas wants to continue to capture this sort of feedback at the appropriate point, and going mobile is a smart approach for this. This mirrors the consumer feedback model, where Adidas wants to get feedback at the appropriate touchpoint, such as straight after a purchase or delivery.

For employees this could be for feedback around the onboarding experience, which the company will request immediately via the app. "If you systematically capture this data you get more insight," Hierl explains.

The Qualtrics platform then performs text analysis on the results and clusters them according to keywords and sentiment. Managers can then easily digest the feedback through their own branded mobile app called Leadership Experience.

Hierl says that the results of People Pulse, internally known as the people score, is now considered a top KPI alongside revenue and share price.

Hierl would like to see the data capture side catch up to the modern day even further, going beyond form filling to more conversational interfaces like chatbots or AI personal assistants like Amazon's Alexa or Apple's Siri.