While the world is obsessed about how machine learning can make cars drive by themselves fintechs are offering businesses more mundane, yet more rewarding, creations using the technology.
For example Previse, a fintech founded in 2016, offers a platform that harnesses machine learning and huge historical data sets to identify the chances of invoices not being paid on time, and automate the payment of the 95% that are not identified as being problematic.
With trillions of dollars in payments made to suppliers globally each year it is a massive opportunity to make things more efficient.
According to Previse co-founder and CEO Paul Christensen: “All B2B across the world uses payment terms and it’s the most outdated, inefficient and insane way of doing business. But it is the way the world works.”
So an opportunity to change this was identified by the Previse team, which set out to make B2B payments as instant as B2C. “Could you imagine if you went to Starbucks, bought a coffee, and said ‘I’ll pay you in three months,’” said Christensen.
Christensen, who worked on the fintech team at a large bank for 15 years, left to start his own business with cofounder, which became Previse. He said late and slow payments are a very common problem that every business knows.
The first thing the founders did when they had the idea was to test it out. “The problem was clear as there is $125trn in B2B payments across the world and businesses are struggling to get paid.”
He said the team’s “ah-ha moment” was when they realised nobody is looking at the data that corporates have on the payments they make, “Every large corporate in the world has a goldmine of payments data in its ERP system,” he added.
The idea was that using big data and machine learning tools they could mine that data to predict the very few invoices that cause a problem.
The platform, through its bank backer, would pay the invoices immediately, automatically, unless it thought it might not get paid.
“The first thing to do was get our hands on a huge data set to test if it was actually possible,” said Christensen.
Previse worked with a small consultancy firm which had data on the spending of big corporates all over the world. “We mined that data and built a bunch of algorithms and tested it all, which took about six months.” This proved that it actually worked. By looking at historical data on invoices it could predict which invoices are going to cause a problem.
Then it was time to test it in the real world. “We got six large corporates to let us give it a go.” It worked for them and Previse raised its seed investment and hired the developers and data scientists it needed. It took about 9 months to build the platform
Machine learning is at the core of the platform, running algorithms on the data sets of the large corporates
Christensen said: “It is really low hanging fruit. I go to all these conferences and everyone seems focused on stuff like autonomous cars [but not many] talk about things like this.”
Machine learning assigns a probability of being paid to an invoice by looking at every feature on the invoice and compares it with past history. It is focused on the buyer’s behavior.
The calculation is for the bank behind Previse so it can decide whether it can pay it off immediately. About 5% get rejected, said Christensen
The corporate signs up with Previse, and the company sending the invoice is offered the chance to go through the Previse platform to get paid straight away. If it does a small commission is taken from the invoice and shared between Previse, the bank paying it immediately and the company making the ultimate payment, which is the corporate customer of Previse.
While the supplier pays a small commission but is guaranteed to get paid immediately, while the corporate being invoiced has happy suppliers being paid on time.
Previse currently has offices at Kings Cross London and Glasgow, where a lot of the data science is done, Previse has 33 staff. It will have 80 by the end of this year.
It also has representatives in the US, Germany and Vietnam. These are currently just individual members of staff but this will be expanded this year.
The company sells direct and through partners such as accountancy firms, banks, and even other IT suppliers. This includes companies like PWC, RBS, and Oracle. “We are small and want to focus on the data science and not have a huge workforce.”
Christensen said overcoming inertia and bureaucracy within corporates is the company’s biggest hurdle.
“The biggest challenge was getting large corporates to adopt it. It doesn’t matter how compelling the product is, the enterprise sales cycle takes a long time,” he said.
Previse was recently named by Tech Nation as one of the 30 fast-growing tech companies joining its Upscale 4.0.
Read the previous fintech interviews
Part 14 Finastra, Part 13 InstaReM, Part 12 Eucaps, Part 11 AimBrain, Part 10 Meniga, Part 9 TrueLayer, Part 8 InvestCloud, Part 7 ClauseMatch, Part 6 Rebuilding Society, Part 5 Honcho, Part 4 Akoni, Part 3 Wrisk, Part 2 CreditLadder, Part 1 Taina Technology