Africa Studio - stock.adobe.com

Don’t bank on Google Bank, bank on banks becoming Google-like

The journey to Google bank is happening as the UK 's Lloyds Bank signals its intent to move to a platform inspired by the workings of the Google engine room

After three years working at Google in what he described as being “in the kitchen of a master chef” for a software engineer, former Google tech lead Paul Taylor is now serving up the fruits of his learnings to the banking sector.

Taylor is co-founder and CEO at Thought Machine. The fintech company has developed a cloud-based banking platform that could create the closest thing to a "Google Bank" possible, save Google itself having a mid-life crisis and becoming a bank.

Back in 2014, when he left Google where he  headed up text-to-speech, he set up Thought Machine, a firm applying the engineering principles of Google to provide banks with an alternative to their legacy systems.

The company, which launched in 2014 currently has 125 staff based in London. Its customer growth over the past 12 months will see it increase its headcount to about 300 by the end of this year.

Thought Machine has already had an impact. Lloyds Banking Group plans to move 500,000 Intelligent Finance division customer accounts from its legacy IT onto the fintech’s cloud-based platform, known as Vault. After that, who knows how far the banking giant will go with it?

Digital-only UK challenger bank Atom is another customer looking to migrate to Vault. Atom is 40% owned by Spanish bank BBVA, which has an option to buy the rest of the shares. If the implementation is a success at Atom, BBVA, which has over 50 million customers, could consider it an option for its other businesses stuck on legacy systems.

“Our platform is designed for tier one banks and is built for scale,” said Taylor. Thought Machine has already run a trial and proved it could run a bank with 100 million customers.

Google’s influence

Google’s influence over Thought Machine comes from Taylor’s time at the internet giant, which honed his engineering skills. He was an academic at Edinburgh University in the 1990s and focused on search technology, artificial intelligence (AI) and machine learning.

He then moved on to set up two companies, and one of them, Phonetic Arts, was acquired by Google in 2010, which became the text-to-speech system for Google. This is the system behind the driving directions and voice search from Google.

“Google didn’t have an expert in that area so they made an offer,” said Taylor. He became Google’s head of text-to-speech.

Three years at Google gave Taylor an insight into how the company goes about its business of tech. “I considered myself a pretty good engineer when I got there, but Google takes it to a different level. It’s like working in the kitchen of a master chef,” he said.

The perfect storm

Fintechs like Thought Machine are not being created out of the blue. There is a burgeoning market for what they offer. After decades of building layer upon layer of middleware on top of reliable but inflexible mainframes, banks have been forced to accept that they have to change the IT that underpins them.

In total, about half a dozen banks globally are engaged with Thought Machine, with plans to migrate to Vault. This year looks set to be the year they make the switch.

This is groundbreaking in the banking sector, where until now the fear of migrating to digital platforms has outweighed the perceived benefits.

Read more on banking IT challenges

Banks of all sizes need to get the tech right if they are to prosper in the future. For example, they need to process transactions securely and in real time with as little effort as possible on the bank’s customers' part. Customers think of Google when asked about how they want to receive services, and banks are on journeys to providing Google-like experiences.

But while the customer-facing end of Google offers the kind of services the banks dream of, the engine that underpins them is the true holy grail of banking.

It is not just about funky functionality like the bank knowing who you are and what you like; it is about automatically switching to a new datacentre when things go wrong, making software upgrades with no downtime and doing this without the need for human intervention. “I learnt a lot of the tricks at Google,” said Taylor.

For example, Taylor said all of Thought Machine’s code is in a single repository, it compiles multiple times a day, is automatically tested and is deployed every 10 minutes. “So we do a full deployment of the full stack every 10 minutes,” he said. All of this is automated.

Thought Machine only programs in two languages: Python and Go, which was designed at Google. “When you look at legacy systems, bits are written in different languages and when you add it all up that leads to problems,” said Taylor.

Google continuously upgrades running systems without human involvement. In comparison, when banks upgrade systems, it involves large IT projects, communication with customers, planned downtime and quite often unplanned downtime.

Replacing a 40-year-old legacy

This might be a utopia for transaction engines like banks, but the problem for banks comes from the complexity of their legacy systems. Thousands of systems are interlinked multiple times as part of different banking products.

Many are written in different languages, some of which are only understood by a small group of ageing engineers. Then there are layers of middleware making everything tick. The result is extreme difficulty adding new functionality and a maintenance bill that drains the IT budget.

Commentators usually cite 80% as the proportion of a bank’s IT budget that is spent on maintenance. Lloyds said its planned move to the Through Machines platform could save £750m in annual IT costs.

Imagine how pay-as-you-go utility pricing based on usage sounds to the average banking CIO: music to their ears. This is what Thought Machine is promising.

Taylor said banks are desperate to solve the problems of legacy technology: “How do you really get technology to scale? How do you stop it from becoming ossified and stuck? There is no silver bullet but there are multiple practices in there.”

The automation of tasks carried out by by people is an important component. “For example, at Google, there are next to zero manual processes for things like testing and maintenance. We were a team of about 20 engineers releasing systems that hundreds of millions of people were using every day,” said Taylor.

Search requests

Google processes billions of search requests a day and Google AdSense automates advertising charges and payments. “So it can be done,” he said. “It is not easy for banks to go from where they are today to where they want to be, but if they get there, many of their problems are solved.”

But given the challenges of planned upgrades, there are good reasons why bank leaders are reluctant to embark on major customer migrations, which are unknown territory for them, and a job on the line for those that lead.

Recently, Paul Pester, CEO at TSB, had to step down after a botched migration of customers to a new banking platform. The meltdown cost TSB £300m, not including the loss of customer confidence.

But banks can’t hold back anymore. “Legacy systems are definitely getting replaced because we have customers that are doing this,” said Taylor. “Banks are not going to move quickly but the pressures on banks are increasing year by year and ‘business as usual’ is just not an option.”

But the longer they leave it the harder it gets. He said banks went in a different direction to tech firms, but now they are looking to join them.

“There was an evolutionary fork in technology about 20 years ago. Google’s open source, cloud, agile and continuous deployment was one route and the banks took another. There was a divergence in how the systems work and the gap is getting bigger and bigger – and it is becoming more and more painful for banks to stay where they are,” said Taylor.

Why now?

Taylor said the banks he has spoken to are very keen to learn and are open minded. “They have already agreed that long term they are going to move off legacy systems and now they are just talking about which path to take and when.”

However, it is not just about the savings and efficiency improvements that modern tech can bring to banks. The driving force also comes from customer expectations, which have changed drastically.

The banks have to make everything easy and not just dress it up with a funky customer interface. This involves processing in real time. “It is fine if they have an app but it also has to make sure that other processes, such as changing your name or address, are also easy,” he said.

The regulators are also getting concerned about the downtime being experienced by different banks. “There are more than you think, but they all kind of blur into one,” added Taylor.

It wasn't always like this. “A bank could have gone down for a weekend in the past and everybody would have been fine. Now, if it is down for five minutes, Twitter lights up. It is also about expectations.”

The risks of IT downtime

Regulators understand the risks of IT downtime at banks. The carnage at TSB is not something rule makers want to see every other month, but if regulators want banking IT to be better, with less downtime and more visibility, they have to encourage banks to migrate, said Taylor.

“You can’t just say ‘I am very angry your system failed’. You need to give them a way out by enabling them to move to new platforms.”

He said Google-like processing would make life much easier for financial services regulators. For example, it offers real-time transactions, so there is no back-office settlement delay.

Thought Machine can do stress tests in one hour, whereas the banks usually take weeks, but it is the big unknown that holds everyone back.

Giant leap for banks small step for tech giants

Taylor said that while he has enjoyed working with bankers because they are open and receptive to what Thought Machine is doing, he has been shocked by the backwardness of IT in banks.

“I already knew it wasn’t the leading edge, but they don’t even have APIs (application programming interfaces). If you want to connect to a database they have to put a huge program in to do it.

“They have followed their own path about how to do it and I think they have now woken up a bit late to the real benefits of things such as continuous deployment and agile,” he said, adding that although banks get it, they still underestimate how much better it is.

On Thought Machine’s part, the firm had to learn about banking, and by breaking it down to core functions, this did not take long.

“We had to learn the language of banking, but you learn quickly,” Taylor said. “In essence, a bank is a simple machine, borrowing and lending money, but there are lots of technical terms. However, once you peel all that away, it’s a pretty straightforward technical problem and we have to make sure we bridge the gap.”

There has been some success bridging this gap. Thought Machine customers are ready to replace legacy systems and most of them have begun their journeys.

Digital challenger banks

The most common route taken by customers is initially building digital challenger banks of their own. “All of our customers are big banks and they now want to build standalone digital challenger banks,” said Taylor.

However, this is not the same as developing a mobile app and re-branding it. “All banks have got digital apps, but that is not enough, because they are still sitting on the legacy platform. The next step is to build a separate bank which will be a new one all the way down, with a separate treasury function, payments and credit cards,” he said.

Some banks do that because they want to enter new markets, and others want to attract a new demographic.

“But most banks are using that as a stepping stone. Get a digital challenger live, see how it goes, and get confidence within the bank, the customer base and at the regulators,” said Taylor, adding that most of Thought Machine’s customers are in this category.

Read more on Mobile apps and software

CIO
Security
Networking
Data Center
Data Management
Close