Daily Archives: May 23, 2019

AI Won’t Solve Your Talent Problem!

Talent is Still the Biggest Issue Facing Procurement Today … so what are you doing about it? (Besides still cutting the training budget as soon as cashflow gets tight and delaying necessary system purchases because you can’t take a long term view.)

As SI has repeatedly said, Procurement Pros need to be jacks of all trades and (almost masters of all but in reality) masters of one (Procurement) (Trend #17), and that’s no easy feat when the skills and knowledge a Procurement pro needs to do her job effectively increases every year.

And new AI / Cognitive technology doesn’t decrease the skill sets and knowledge required, despite what one may think. In fact, it only increases it Why? First of all, do you have assisted intelligence, augmented intelligence, or a cognitive system that is as close to true AI (artificial intelligence) as one can get with today’s technology? And, more importantly, does your Procurement Pro understand what you have, what the differences are, and what the respective limitations are.

If the solution is just assisted intelligence, then it’s an automation solution (RPA) with some expert knowledge encoded to handle typical situations with certain assumptions. If the assumptions are invalid, will the software detect them? If the situation goes beyond the realm of typical, will the software detect it? And even if the software does, will it be able to do anything without expect guidance? An example of assisted intelligence is an automated auction where the platform automates the sourcing of an item or service designated for auction among pre-approved bidders and goes from demand specification to final award without human input. But will it detect if the bids are complete? Within expectations? That bidders are bidding on the right product or service? Maybe the buyer assumes shipping included, but the bidders aren’t including shipping, and since the system only has a ceiling, it doesn’t know that the bids are way too low, and awards to the lowest bidder, that is actually the highest as the bidder is the furthest away and has the highest transportation cost.

Same goes for augmented intelligence. However, with augmented intelligence, the software goes beyond simple RPA with fixed expert rules — it is able to analyze a lot of parameters and pick the closest matching scenario and associated workflow. For example, an opportunity analyzer that takes into account current market pricing, supply availability, bidder responsiveness, current market trends (upward and downward), projected demand, etc. and advises the buyer on the type and timing of the sourcing event as well as the best workflow. But what if the market pricing is a week out of date and the market price just jumped up 20% (due to a fire in a major supplier’s plant) and reversed the trend? That changes everything, but the solution may not detect it and instead advise the worse sourcing event.

Cognitive platforms that continually monitor the situation are better, and if they learn from the actions the expert users take over time, better still, but they still can’t cope with an exception al situation they haven’t been coded for, or trained for. For instance, even if they detected that last minute spike in pricing that reversed the pricing trend and, thus, changed the optimal sourcing strategy, will they understand why the spike happened and the best alternate strategy? Or will they default back to the recommending the default strategy in a situation where costs are increasing … e.g. switching from auction to multi-stage RFI with optimization-backed analysis? Neither is right in this situation. In this situation, its extend the current contracts with your non-affected suppliers, increase the number of units, and lock in supply early, even if cost is higher.

In all these situations, only a knowledgeable, experienced, and sometimes expert Procurement Pro is going to be able to make the right decisions … and a novice relying on the systems is going to make the worst, and most costly, decision imaginable.

There’s no true AI, no all knowing software, and no replacement for a real expert.

The reality is that, at the end of the day, these systems make your experts more efficient — and multiply their productivity — they don’t replace their expertise.