CORONAVIRUS RESPONSE: Dear Procurement, AI won’t save you, but rules-based automation might!

This Spend Matters analyst brief has been posted outside the PRO subscription paywall to share information that could help businesses take on the coronavirus disruption.

In the last few years, a number of vendors have been pushing artificial intelligence. Some vendors have even been pushing AI-based suites as the future of sourcing and procurement. And for a time they had a great argument. There are too many low-value, straight-forward, simple and/or tail-spend categories that are not getting appropriately sourced in an average organization that doesn't have enough people power or hours in a day to properly address all organizational spend in a strategic manner and identify the range of savings and opportunities available to the organization. So why not let technology take over some of this spend, especially where it can't do any worse than what is being done now (because it can always just replicate what was done last time in a worst-case scenario), but, on average will do better than not having the category managed at all.

Given the coronavirus-related chaos going on now, with some products virtually unavailable, some potentially unavailable in the near future, others having dramatic price fluctuations (up and down), and quality always unknown, you'd think now more than ever you'd want a solution that could just take some categories off your plate and source/procure them for you. And while that would be great, now more than ever, you can't. You see, all of these "AI" solutions are based on classical machine-learning algorithms that have produced standard responses based upon standard situations identified from large historical datasets where a significant amount of the time the data is consistent — price is relatively static over a period of time or rises over time in accordance with a predictable trend line, supply/demand imbalance is relatively consistent, there are no major shake-ups in the supply base or customer base or new major product introductions that significantly impact the supply/demand imbalance, etc.

Compare this to the current situation where supply availability (depending on which factories / mines / etc. are online and which supply routes are open) and market demand changes (depending on the necessity of the good and/or the available business consumers) by the day, where prices are all over the map, where suppliers and competitors are shutting down (and maybe coming back online), where alternate products are all of a sudden viable, and so on. There is no historical data that describes this chaotic coronavirus situation, and no standard responses in the platform's repertoire. There's no statistically relevant response and anything the platform did would be essentially random and with the platform's inability to apply true intelligence to the situation, the platform would be just as likely to choose the worst option as choose the best option.

In other words, not only is now NOT the time to deploy AI-based sourcing and procurement technology — and it's not the time to over-rely on any existing AI-based sourcing and procurement solution you might already have. You can use the analysis capabilities where you have access to daily information feeds and look at the recommendations, but definitely do not use any auto-sourcing/auto-procurement. Apply human oversight to every key step of the process until the global economic situation settles down to a new normal that has been consistent long enough to produce enough data for the algorithms to learn from and adapt to. Everything is an exception now, so there are no consistent rules.

On the flip side, you have no time for any tactical, semi-thoughtless tasks that can be automated, or at least automated the majority of the time and only escalated to human review in exceptional situations. Now that we're all stretched thin, working remotely, having to spend most of the day in online meetings, we need automation more than ever. In particular, we need all of the classic automation that has been used for years as well as the automation that is underlying modern AI-based sourcing and procurement programs, but manually (and not system) controlled through rules and semi-automated processes with exceptions and forced go/no-go confirmations by humans. What's the difference?

Classic Automation Examples

Automatic Invoice Processing

When an invoice comes in, it is parsed to identify the supplier, the items, the unit counts and associated costs, then it’s matched to a PO (or, for a recurring purchase, a contract with a purchase schedule) and if the supplier name matches, the items match the PO or the contract, the costs match and the units match a received (but unmatched) count in the inventory system, the invoice can automatically be approved as payable and sent to the AP system. If something doesn't match, then it is kicked up to a human for review. In an average organization that makes proper use of POs, contracts and payment schedules, this will allow 85% to 90% of invoices to flow straight through to an AP system.

Form-Based Supplier Onboarding

When a supplier wants to be considered for an RFP or part of your broader network, they self-register and the system walks them through a series of questions using wizard-like interfaces to collect appropriate data, load associated data from any third-party networks or databases you have access to, and create a classification scheme for the supplier. It then puts the supplier in a queue for approval as a network supplier, or at least a supplier that can be considered, and identifies the elements that need to be verified by a human.

Traditional Opportunity Analysis

Analytics is run against your historical spend store and it highlights your top n suppliers, categories, geographies and divisions — as well as your top m suppliers, categories, geographies and divisions with spend NOT under management so you can identify potential opportunities and tackle them in order.

Modern Automation

Semi-Autonomous Invoice Processing

Eliminating up to 90% of invoices that need to be processed manually is good, but eliminating up to 97.75% of invoices that need to be processed manually is much, much better. When you think about it, most invoice matches fail because the supplier sends incomplete header information, doesn't reference the PO or contract, quotes the wrong price, bills for more units than were delivered (or accepted), etc. Most of this is easily rectified by including the missing information, or correcting the price or units and flipping it back to the the supplier (with notes that this was the PO or contract price, etc.) for acceptance, which will happen 80% to 90% of the time (when the supplier realizes their mistake or realizes you caught them trying to sneak an old price through). Why should these invoices get kicked up for manual review and correction when the system can handle it with automation? More modern systems take care of all this automatic processing for you and only kick it up if it can't fill in the gaps and reconcile the error, or the supplier rejects the suggested corrections.

Information-Based Supplier Onboarding

Most classic systems collect a lot of information from a supplier in a highly structured fashion, but only have (at most) one integration point and can't verify the majority of information. While this saves the time of the buyer entering the information, it saves very little review and validation time because so much has to be validated by human eyes and human hands that it doesn't save that much time. Modern systems integrate with multiple supplier networks, third-party financial / risk / sustainability data sources, government registries, etc. and allow the majority of the information to be automatically pulled and verified. The more information that can be verified, the less work for you.

Market-Based Opportunity Analysis

Your top X categories, or your top Y categories based on spend not under management are not your top Z opportunities, especially not in a turbulent market such as we have now. Your best opportunities are your top opportunities against current market conditions, which could among be your top X under contract (if prices have recently plummeted and the contract is near enough to termination that you could start using another supplier for the majority of volume) or your top Y not under contract (if now is a great time to lock pricing in). This is gauged by looking at average market prices, average community savings across the industry/geography over recent months among the global buying community, and running sophisticated comparative trend analysis of average organizational price vs average market price and multiplying that by projected volumes. Great systems will do this work for you and present you with your top Z opportunities in descending order of estimated value over the next 12, 24 or 36 months (given average organizational contract length) for your review, so you can quickly pursue those with opportunity.

The better systems will also go one step further and highlight categories coming up for renewal in the next 3 to 6 months or not under contract where current trends indicate prices are likely to increase significantly (due to recent shortages in supply or surges in demand) so that you can maximize cost avoidance by locking those categories in quick, especially if the losses would dwarf the savings of going after another untapped spend opportunity. And the best systems will look at cost avoidance vs. cost savings potential and suggest how to balance each set of opportunities to minimize organizational spend overall.

... but modern automation systems will go beyond traditional categories of automation and include ...

Automated Supplier Discovery and Recommendation

When a source of supply goes away, or is no longer sufficient to meet organizational demand, you often need to find new suppliers fast — but having to search web directories, Google listings and third-party networks is time-consuming. It typically takes weeks of person time to get to a shortlist for an RFI. But with modern technology that uses advanced machine learning and semantic processing techniques and scours product information booklets, third-party ratings and comments, and organizational product definitions / bill of materials, these tools can often come up with a very relevant shortlist to send a targeted RFI to in a matter of minutes.

Automated e-Sourcing Event Selection, Setup and Award Recommendation

With a dynamic analysis of current market conditions versus historical, the system not only will match the product with the closest event template in the system, but with the general strategy used in such a situation — be it e-auction, multi-round RFX, re-negotiation / extension with incumbents, etc., recommend that strategy, and adjust the event template to that strategy. With just a few clicks, the buyer can select the template and recommendations, tweak the event and push it live.

Similarly, when the bids/responses are submitted, the system can review the submission for completeness (all necessary data is there) and sensibility (quotes within expected ranges, documents are actually requested certifications, etc.) to minimize buyer verification time and, once all bids are submitted or the event is closed, automatically create comparison reports and award recommendations under different, standard, scenarios for review. If a user thinks an award recommendation is close, the system will make it incredibly easy to make minor adjustments and then auto-create award-notification letters and draft contracts (either through the use of predefined templates where the specific award information is attached or by pushing the award into the contract lifecycle management system).

In other words, while you want some ML/AI underneath to automate some analysis and make some recommendations, because everything is so out of whack, the focus is on automation of currently manual rule-driven processes, not on automation of sourcing and procurement.

Accept and embrace that, and your productivity might just soar.

See all of Spend Matters’ Coronavirus Response series here, and see all of our coronavirus coverage here.

Through this month, a Spend Matters' special PRO Expert Survival Pack is available to procurement practitioners only* at up to 50% off. The discount applies to PRO subscription content from our analysts and other services. — Learn more

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.