Paddy Lawton at eWorld – Machine Learning and Procurement Benefits

At the recent eWorld Procurement and Supply event, Paddy Lawton, founder of Spend360, the analytics firm bought by Coupa earlier this year, spoke about The Inside Story: From Machine Learning to Procurement Benefit Realisation. The session was presented in conjunction with consulting firm Xoomworks, who are probably the leading implementers of large-scale Coupa programmes in the UK.

Lawton started by explaining why understanding cost is not as simple as it might sound for large organisations. Is the spend data accurate? Are the units of measure understood and consistent? What about currencies, let alone just understanding the many different ways in which “IBM” might be recorded in your payables ledger. That’s why executives spend a lot of time “trawling though spreadsheets”, just to try and get a clear view of spend basics - how much, with whom, by whom, etc. The whole issue comes down to classifying unstructured data, he said.

That’s where the potential for AI (artificial intelligence), machine learning and deep learning come into play.  In a recent survey, 80% of CEOs see data mining and analysis technologies as the most strategically important digital technology. And back in 2013, Gartner was saying that spend analysis was “essential for organisation with revenues of over $1 billion” plus. But now, Lawton claims, this technology is much more affordable and should be a priority for many more organisations.

However, he cautioned against getting over-excited about AI and so on. “We will always need humans to reinforce the learning inside the machine”.

For me, the most fascinating element of the session was his references back to the very beginnings of AI, a Dartmouth (USA) summer-school research project in 1955!  Just before the session, Lawton showed me the proposal for the research, and it was very prescient; it talked about neural networks, machine learning and so on.  Basically, the brilliant key people in the 1950s and '60s developed a large proportion of the underpinning theory that is still the basis for development today in this field  – there just wasn’t enough computer power available to make it happen at the time.

In the '80s and '90s, practical uses started to be developed – like early number plate recognition. But it was still fairly basic. Then the first “neural network” was developed in Canadian universities, linking together computers and providing the basis for “deep learning”, which needed huge computer power (and loads of data) to be effective.

So coming back to the present day, spend and supplier classification using deep learning can clean up data better than traditional methods. And as we move into the third-generation products, we are seeing “deep machine learning and intelligent algorithms”, as Lawton says. That means for instance, “understanding the words but also what lies behind them” in invoices and other documents - the system will “know” what a chair is, by understanding the whole sentence and the context, and learning from past examples.

For spend analysis, that is much more powerful than rules-based processes that simply say “if supplier = Grainger then category = MRO…”

What strikes me is how clever all this is in terms of the technology behind it, and yet in some ways the benefits won’t appear too startling to the blasé young category manager. This is not going to revolutionise procurement – rather, it is going to lead to spend and data analysis that is just more accurate and much faster to produce. Frankly, once we have it, everyone will take it for granted. But isn’t that the same for any revolutionary technology – how often do you consciously remember how amazing your smartphone really is?

But as always, the competitive advantage will come to those who know how to adopt these advances and really make use of them, against those who stick with Excel-based spend cubes. And if you get the chance to see and hear Paddy Lawton speak, it’s well worth taking it – he’s one of those rare people who is a genuine “geek”, a deep technical subject matter expert, but also a persuasive and entertaining communicator!

Share on Procurious

Discuss this:

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.