Artificial intelligence, we are told, will change the workplace of the future and make many of our jobs redundant. As is often the case with future gazing, the reality will in all likelihood be much more prosaic, and the utopian dream of a labour less future seems unlikely to emerge. That said, a profound shift is certain to engulf us and the work of purchasing will likely be affected.
In some ways, AI is already here. Our own website OpenOpps.com uses artificial intelligence to classify our tender records. We’re also exploring AI to make our site's search engine more powerful and relevant. We have recently worked on a tool for the Competition and Markets Authority that uses machine learning to identify behaviours and indicators that are associated with cartel bidding.
We're also starting to look at predictive analytics, so we can start to predict how competitive a bid might be, or whether savings might be made. I’m sure buyers would be keen to know how small changes to their tender documents might deliver increased savings, or how a better lotting strategy might encourage more SME bidders. If you can dream it, given enough data, AI can start to predict it.
AI will give buyers much more power, much more insight. What if you knew that contracts with this specification went over budget more often than not? What if you knew that anything above a certain level of indemnity would prevent every small business from bidding on a contract?
The impact will almost certainly help to manage risk. Analysing hundreds of thousands of prior contracts would allow buyers to profile their contracting on the basis of a range of risk factors, predicting the likelihood of these events occurring in each contract.
But before we get too excited by a rich new vein of techniques to deliver lower risk, lower cost contracts, we need to be ready for those that might use AI in less helpful ways.
What happens, for instance, when we start to predict which companies will win a tender? Once this technology is able to analyse enough data, it is certain that an investment firm will seek to predict the fortunes of listed companies.
However, what happens when the same algorithm is taken up by suppliers? Surely an algorithm that predicts winners might reduce competitiveness, as suppliers choose to avoid bidding on contracts they perceive they cannot win.
The same algorithm in the hands of buyers might have a similar effect, as buyers use it to shape their requirements to fit their favoured supplier and provide them with a tender that will deliver the contract to them. Using a good algorithm, won't create a crass, easy to spot piece of corruption. Instead algorithms will create a sophisticated interplay between hundreds of different factors. To any casual observer a tender will have been fairly competed, but the outcome will have a depressing certainty. The future of corruption may be almost impossible to spot.
Governments spend $9.5trn a year with their suppliers. Using AI to spot corruption, drive competition and reduce costs is already starting to happen, but AI will come with its own responsibility. How we wield that responsibility is up to us.