Artificial intelligence and intelligent procurement

Cost control has never been higher on the agenda. Accounts teams in pharmaceutical businesses are well acquainted with the need to spend hours completing detailed analysis of extensive data, but what they might not know is that there is an alternative: artificial intelligence.

Machine learning has now become a useful tool for tackling cost analysis. Here at Tungsten we are interested in exploring this further, looking at how businesses can be better informed about procurement options using the latest developments in technology.

To investigate, we’ve gone into partnership with Goldsmiths University and in April last year established the Tungsten Centre for Intelligent Data Analytics (TCIDA). Led by Mark Bishop, one of the UK’s leading researchers in machine learning, the project is already delivering tangible results.

Spending habits were highlighted as a real cause for concern last year, when a report by Lord Carter revealed that the NHS had been erratic in its procurement choices. Latex gloves were being bought for £5.44 in one hospital, but £2.39 in another. Aprons bought for £4.22 in one hospital cost £2.51 elsewhere.

Lord Carter, chair of the NHS Co-operation and Competition Panel, examined the ledgers of 22 leading hospitals in the UK. Following his investigations, he called on the NHS to get a handle on its purchasing of everyday health consumables, to meet the global best practice standard of a catalogue of 6,000 to 9,000 product lines with price variances of 2 per cent. At the time it had 500,000 lines and price differences above 35 per cent.

This standard he outlines, of a manageable number of lines with little price variance on each, is something all businesses and organisations should strive for. How to go about achieving this, however, can be tough.

Our e-invoicing technology, which automates the financial side of the procurement process to make it more efficient and uniform for businesses like GlaxoSmithKline, can help. Businesses that use it to process their invoices have digital access to every invoice ever filed, and therefore what they’ve bought at what price.

Given that an accounts payable department of a company may handle thousands of invoices per month, the sums to be saved are considerable. Using e-invoicing also means happier suppliers; they can keep track of their invoice status, so don’t need to ask or chase the buyer for updates, and can access early payments.

These computer-led systems open up a huge amount of data that can help improve future business, but this data is so vast it is impossible to analyse manually. Each entry simply cannot be checked with the naked eye, and therein lies the flaw. While the entries can be searched with current computing tools, variations such as spelling mistakes and typos are hard to catch without a human eye. For example, a computer won’t know that a biro costing 10p is standard, but one costing £100 is an anomaly. A decimal point in the wrong place makes a big difference.

Artificial intelligence offers a more efficient solution, we believe. With machine learning, we can train computers to cross-check items so that when a pen appears at 100 times its normal price, flags can be raised and the mistake rectified by a real person.

At the TCIDA, we are currently working on technology whereby computers can assist the buying decision process. By checking internet prices against existing deals, we can offer a guide on what the right cost looks like.

Using our current spend analysis capabilities, we have helped one pharmaceutical company to find $80 million worth of potential savings, by analysing spend on 475,000 products. The technology of the future will not only help other businesses in the sector see where better buying decisions could have been made, but become a vital tool when making new decisions.

We believe artificial intelligence has a lot to offer and so far we have only scratched the surface. Despite what you see in science fiction films, it’s about more than robotic dogs. The possibilities it represents for businesses are endless, or at least further than the human eye can see.

This article first appeared in The Pharma Letter.