Back in late 2018 and early 2019, before the GENizah Artificial Idiocy craze began, the doctor did a sequence of AI Series (totalling 22 articles) on Spend Matters on AI in X Today, Tomorrow, and The Day After Tomorrow for Procurement, Sourcing, Sourcing Optimization, Supplier Discovery, and Supplier Management. All of which was implemented, about to be implemented, capable of being implemented, and most definitely not doable with, Gen-AI.

To make it abundantly clear that you don’t need Gen-AI for any advanced back-office (fin)tech, and that, in fact, you should never even consider it for advanced tech in these categories (because it cannot reason, cannot guarantee consistency, and confidence on the quality of its outputs can’t even measured), we’re going to talk about all the advanced features enabled by Assisted and Augmented Intelligence that are (or soon will be) in development (now) and you will see in leading best of breed platforms over the next few years.

Unlike prior series, we’re identifying the sound, ML/AI technologies that are, or can, be used to implement the advanced capabilities that are currently emerging, and will soon be found, in Source to Pay technologies that are truly AI-enhanced. (Which, FYI, may not match one-to-one with what the doctor chronicled five years ago because, like time, tech marches on.)

Today we continue with AI-Enhanced Sourcing that is in development “today” (and expected to be in development by now when the first series was penned five years ago) and will soon be a staple in best of breed platforms (and may be found emerging in development beta versions of some platforms). (This article sort of corresponds with AI in Sourcing The Day After Tomorrow that was published in January, 2019 on Spend Matters.)

TOMORROW

Automatic Strategic Sourcing Events

Just like tomorrow’s Procurement platforms will automatically identify products/services and (sub) categories that should be pulled out of the tail and inventory/catalog/one-time req buying and pulled into a strategic sourcing event, tomorrow’s sourcing platforms will create automatic events from them. Furthermore, tomorrow’s sourcing platforms will automatically create the entire event using the default category strategy (possibly adjusted to the current market conditions, see the next forthcoming capability), automatically pull in the (organizationally approved) suppliers, automatically pull in any questionnaires or documents that need to be completed by the bidders, automatically pull in supplier profile information and current prices (where available), and, if you set the flag for “no review prior to event initiation”, automatically send out the RFX, which could be the first in a series of RFXs/e-Auctions in a multi-round event. If the event is multi-round, after each round it can analyze the responses and any supplier who provides all of the necessary information (and makes the cut price/quality/risk/carbon/etc. cut) makes the next round. It will auto-execute the next round and keep going until the event has been completed and an award recommendation is made. Then, depending on the setting (auto-award, human review), it will either compute a recommended award and notify a buyer to approve, modify, or reject the award, or automatically send the award to to the suppliers for acceptance (with a contract for high-value or strategic products/services or a PO for lower value, more tactical offerings).

From a tech perspective, all this needs is the ability to analyze spend patterns and demand trends (trend analysis) to identify categories ripe for sourcing, product classifications to match to the category strategy, and product-supplier pairings to pull in the suppliers (and associated data), with current and preferred suppliers getting priority if there are too many. The rest is just workflow automation until the initial responses are returned. Then, it’s just analyzing the data with respect to expectations and tolerances, and either recommending an award based on the strategy, organizational priorities, and organizational constraints, or sending out the next round requests (deeper RFIs, price updates, etc.) to those suppliers who provided complete, satisfactory, answers according to business rules. This is just analytics, optimization, and good ol’ math coded with human intelligence (HI!).

Market-Based Sourcing Strategy Identification

Today, the best platforms support category-based sourcing strategy identification where the platform can identify the standard, best-practice, strategy based on the category and items, determine whether or not the strategy is likely to be relevant given available market data (supply availability, historical price variants, current market prices, etc.), and make a go-no recommendation to the buyer. Tomorrow, these platforms will be able to first analyze all of the market information, supplier information, product information, carbon information, risk information, and compare that to current company performance an demand and identify the right sourcing strategy for the event, making sure to dynamically align the category (which can include adding or dropping items and services) as required.

From a tech perspective, all this needs is access to extensive market data feeds, a large history of sourcing event and results with associated market data (relative to the supply vs. demand imbalance, price trends, demand trends, major risk factors, etc.), pattern analysis that correlates successful events (with results < market price) with market conditions (supply > demand, prices steady or falling, low market risk in the supply base –> e-Auction; supply >= demand, prices rising with inflation, low to moderate risk –> RFX; supply projected <= demand, prices rising above inflation, moderate risk –> renegotiate with the incumbent(s) before the contracts expire), pattern analysis of the current market conditions compared to historical patterns of success, and the selection of the best match. All trend analysis, correlation/(k-)means analysis, tolerances, and, you guessed it, math! Then you just kick off the category-attuned sourcing event as above.

Real-Time Strategy Alignment in (Automatic) Strategic Sourcing Events

However, tomorrow’s AI-based sourcing capabilities won’t stop there. The platform will monitor all relevant market (related) conditions as the event progresses, compare all of the responses to those that were predicted/expected, and if, at any point during the (automatic) event something is too far off, it will automatically pause the event and either, depending on system configuration, alert the buyer that a shift in strategy is required (and what the new strategy it should be) or simply shift the event as appropriate (if possible; in the public sector, not always possible, but in the private sector, usually possible).

From a tech perspective, all this needs is trend and outlier analysis, pattern matching, and, you guessed it, math.

SKU Recommendation and Replacement

Tomorrow’s platforms will get better at identifying replacement SKUs not just in indirect (paper with similar thickness, weight, and gloss when the differences are inconsequential from a business point of view), but direct as well (compatible processors, with the same form factor, number of connections, compatible clock rate, and sufficient L1 cache). This is difficult because you need a lot of specification data, and most applications need it appropriately structured in a format no other application supports in order to process it. But, despite the focus on the Gen-AI bullcr@p, semantic processing is continuing to advance and as more and more validated database are built on each product and service type, and more specifications are added to each product and service type. As a result, these applications are getting better and better at helping to identify acceptable alternates with slightly different, but compatible, specs that can help Procurement and engineers find more cost-effective alternatives, including new tech that will have a longer shelf life.

As this tech continues to improve, it will be able to not just look at SKUs, but subassemblies, such as processor-controller board-memory combinations, that can be switched out to provide more cost effective alternatives with better reliability, risk span, or quality. This will be the result of not only a better understanding of each subcomponent, but the interaction requirements and overall processing power capable of handling the combinatorial explosion needed to automatically identify new potential subsystems, and not just components, automatically.

EOL Recommendation

Many niche PLM systems will already do this, but tomorrow’s sourcing systems will do this not just from a traditional “tech curve” perspective, but also from a Procurement and Supply Chain perspective, balancing life-span with price trends, material supply, market risk, and carbon impact. If a current product requires a large concentration of a rare earth mineral or metal (in short supply) or an ingredient that can only be grown in a few places in the world, and a new product comes along that requires less (or none) but still provides the same use (or at least a suitable alternative for consumption in the latter case), then it makes sense to switch over as soon as the cost is appropriate. Similarly, if one product is only available from a risky supplier or a risky country (with rising political or market instability) or has an unnecessarily high carbon cost, switching out could also be a priority.

Using trend analysis on demand and (future) cost, risk projections, and carbon costs, tomorrow’s sourcing systems will find the optimal inflection points (using analytics and optimization) for switch over and make early end-of-life recommendations so Procurement and Engineering can plan early for the switch-over and schedule the appropriate sourcing events for the appropriate timeframes (and ensure contract lengths are optimal). And, again, no Gen-AI needed!

SUMMARY

Now, we realize some of these descriptions are dense, but that’s because our primary goal is to demonstrate that one can use the more advanced ML technologies that already exist, harmonized with market and corporate data, to create even smarter Sourcing applications than most people (and last generation suites) realize, without any need (or use) for Gen-AI, that the organization can rely upon to reduce time, tactical data processing, spend, and risk while increasing output and overall organizational performance. It just requires smart vendors who hire very smart people who use their human intelligence (HI!) to full potential to create brilliant Sourcing applications that buyers can rely on with confidence no matter what category or organization size, always knowing that the application will know when a human has to be involved, and why!

Back in late 2018 and early 2019, before the GENizah Artificial Idiocy craze began, the doctor did a sequence of AI Series (totalling 22 articles) on Spend Matters on AI in X Today, Tomorrow, and The Day After Tomorrow for Procurement, Sourcing, Sourcing Optimization, Supplier Discovery, and Supplier Management. All of which was implemented, about to be implemented, capable of being implemented, and most definitely not doable with, Gen-AI.

To make it abundantly clear that you don’t need Gen-AI for any advanced enterprise back-office (fin)tech, and that, in fact, you should never even consider it for advanced tech in these categories (because it cannot reason, cannot guarantee consistency, and confidence on the quality of its outputs can’t even be measured), we’re going to talk about all the advanced features enabled by Assisted and Augmented Intelligence that were (about to be) in development five years ago and are now available in leading best of-breed systems. And we’re continuing with Sourcing.

Unlike prior series, we’re identifying the sound, ML/AI technologies that are, or can, be used to implement the advanced capabilities that are currently found, or will soon be found, in Source to Pay technologies that are truly AI-enhanced. (Which, FYI, may not match one-to-one with what the doctor chronicled five years ago because, like time, tech marches on.)

Today we continue with AI-Enhanced Sourcing that was in development “yesterday” when we wrote our first series five years ago but is now available in mature best of breed platforms for your Sourcing success. (This article sort of corresponds with AI in Sourcing Tomorrow Part I and AI in Sourcing Tomorrow Part II that were published in January, 2019 on Spend Matters.)

TODAY

Event-Based Category Alighnment

As per our Procurement series, a good AI based platform continuously analyzes (i.e. re-runs an analysis on a monthly basis) every product or service for inclusion against every organizational category and comes up with the most logical mix for the procurement organization based on likeness, current supply-base, spend-mix, and other existing parameters.

However, when it comes time for sourcing, the category should be appropriate for a sourcing event. This depends on volume, available supply base, and the category strategy (see the next item).

When it comes to sourcing, the AI will look at not only the product specifications, but also ensure there is a sufficiently large supply-base, with supply availability, spend-mix, and price trends. It will do this based on key material analysis (to identify additional suppliers in the market not yet supplying the organization), identification of market offers and volume disclosures from third party distributors vs. organizational need and overall percentages, analysis of spend vs. typical sourcing event sizes using simple (k-means) analysis, and price trends using basic curve fitting/projection. Nothing fancy.

Based upon the demand (volume), available supply base, supply availability, spend mix, price trends, and defacto templated sourcing strategy, the platform will recommend the event proceed using the standard strategy and template, proceed with modifications, or not proceed (alerting the buyer it’s not a good time, not a good event, or a new strategy is needed). It’s all traditional analytics, a smattering of machine learning, a sprinkling of pattern matching, tolerances, and confidence calculations. Nothing super fancy. The recommendation(s) will depend on a number of factors that revolve around the market conditions at the time. Current prices. Available supply base. Category dynamics in the consumer marketplace. Etc.

Category-Based Sourcing Strategy Identification

In our prior series, we indicated we’d have market-based sourcing strategy identification, and while that is in development, we’re not quite there yet. Market-based strategy identification requires a lot of data — market, supplier, marketplace, (anonymized) community intelligence, past event data, and past data from similar situations … the global marketplace has been so dynamic in recent years that we haven’t seen anything like it since pre-2000 … which was before the introduction of mass-market sourcing / procurement / modern supply chain software and we just don’t have the data.

That being said, for the majority of commodity categories, a number of leading firms have developed one or more standard sourcing strategies for the category and categorized the market conditions under which the strategies work. Modern sourcing platforms will run all the analytics against the specified demand ranges, supply vs. demand imbalance, historical price variances (since the last event), current market prices, check the thresholds, compute the match percentage and confidence, and then recommend go, go with changes/caution, don’t go — all using straight-forward trend analysis and mathematical calculations — no Gen-AI needed!

Real-Time Market vs. Response Monitoring and Automatic Pauses/Updates

As the responses come in, the application will not only track bids vs open market prices (and current prices), but compute the averages and if the bids coming in are worse than expected, alert the buyer. In a multi-round scenario, or RFQ-powered auction, the trends will be analyzed and if they are not as expected, the buyer will be alerted. In both cases, if something is off beyond a tolerance, which will adjust over time as buyer feedback on go-no go is collected, the event will automatically be paused if necessary. This just requires simple calculations against means and expectations. Good old math, a few business rules, and some workflow automation is all that is required.

Suggested Award Scenarios

Even if the platform doesn’t contain (true) strategic sourcing decision optimization [SSDO] (and see this recently updated article on Questions to Ask Your Optimization vendor for the requirements for a true SSDO solution), most modern platforms will recommend one or more award scenarios that take into account cost, business constraints, risk and carbon. It’s just a lot of combinatorial mathematical calculations and basic analytic verifications.

Carbon Impact Analysis

Using standard models for carbon production based on available data by industry, country, and when available, factory, modern platforms will use standard models and formulas to compute the carbon footprint by item, based on the supplier, the source location, and the location it is going to (and even take into account logistics based carbon production). It will do this for every item you’ve purchased, every item you’re considering, and show you the carbon impact of different award decisions vs. the status quo. No Gen-AI required! (Just a lot of formulae and data!)

SUMMARY

Now, we realize some of these descriptions, like yesterday’s, are also quite brief, but again, that’s because this is not entirely new tech, as the beginnings have been around for a few years, have been in development and discussed as “the future of” Sourcing tech before Gen-AI hit the scene, and all of these capabilities are pretty straight-forward to understand (especially with many of the fake-take and Gen-AI providers marketing these, or similar, claims, even though they are not entirely realizable within their platforms). And, if you want to dive deeper, the baseline requirements for most of these capabilities were described in depth in the doctor‘s January 2019 articles on Spend Matters. The primary purpose of this article, as with the last, was to explain how more sophisticated versions of traditional ML methodologies could be implemented in unison with human intelligence (HI!) to create smarter Sourcing applications that buyers could rely on with confidence.

Back in late 2018 and early 2019, before the GENizah Artificial Idiocy craze began, the doctor did a sequence of AI Series (totalling 22 articles) on Spend Matters on AI in X Today, Tomorrow, and The Day After Tomorrow for Procurement, Sourcing, Sourcing Optimization, Supplier Discovery, and Supplier Management. All of which was implemented, about to be implemented, capable of being implemented, and most definitely not doable with, Gen-AI.

To make it abundantly clear that you don’t need Gen-AI for any advanced enterprise back-office (fin)tech application, and that, in fact, you should never even consider it for advanced tech in these categories (because it cannot reason, cannot guarantee consistency, and confidence on the quality of its outputs can’t even be measured), we’re going to talk about all the advanced features enabled by Assisted and Augmented Intelligence (as we don’t really have true appercipient [cognitive] intelligence or autonomous intelligence, and we’d need at least autonomous intelligence to really call a system artificially intelligent — the doctor described the levels in a 2020 Spend Matters article on how Artificial intelligence levels show AI is not created equal. Do you know what the vendor is selling?) that have been available for years (if you looked for, and found, the right best-of-breed systems [many of which are the hidden gems in the Mega Map]). And we’re going to continue with Sourcing. (Find our series on Advanced Procurement — No Gen-AI Needed! Yesterday, Today, and Tomorrow at the following links.)

Unlike prior series, we’re going to mention some of the traditional, sound, ML/AI technologies that are, or can, be used to implement the advanced capabilities that are currently found, or will soon be found, in Source-to-Pay technologies that are truly AI-enhanced. (Which, FYI, might not match one-to-one with what the doctor chronicled five years ago because, like time, tech marches on.)

Today we start with AI-Enhanced Sourcing that was available yesterday (and, in fact, for at least the past 5 years if you go back and read the doctor‘s original series, which will provide a lot more detail on each capability we’re discussing. (This article sort of corresponds with AI in Sourcing Today that was published in January, 2019 on Spend Matters.)

YESTERDAY

Workflow / Project Automation

Once a sourcing project is defined, which typically consists of identifying the required products and demand, the critical requirements of the supplier pool, the RFI, the RFP/Q, the evaluation criteria and weightings, the award rules, and the initial award offers, the entire project is easily automated using rules-based automation. Best-of-breed platforms will integrate fuzzy matching to identify additional suppliers who provide similar SKUs, RFI/P/Q templates which will automatically be pulled in and modified based upon the particular items in the category and organizational risk/compliance rules using semantic characteristic matching (traditional NLP will be fine), and built in “cherry-pick” algorithms that will compute standard award scenarios (lowest price, max 3 suppliers, geo-split, etc.) and create a default recommendation — which only requires math and traditional analytics.

Auto-Fill

For the better part of the past decade, the best platform auto-fills not just successive rounds, but auto-fills / pre-populates all of the supplier, item, and RFI data based on available information in all integrated systems — be it from past events, the supplier master, the forecasting platform, or market(place) data (for products).

This just requires rules-based automation and workflow with reg-ex pattern matching, and simple trend analysis and market data matching for price / demand population. Easy peasy on the tech ladder.

Outlier Identification

As we wrote years ago, it only takes one bad data element to make a good sourcing process go bad. Just one. One bid too low that takes a buyer down the wrong path. One risk rating too high that steers a buyer away from what would be their best supplier. One demand error that steers the best supplier away. But all of these “outliers” can be easily detected with traditional mathematical clustering algorithms used as the back-bone of machine learning — k-means, nearest neighbour, etc. — and identifying any values too far off the norm and then alerting the buyer to (have the supplier) correct them.

Rule-Based Auto-Award Identification

For simple scenarios where it’s always lowest cost, simple mathematical calculations can identify the supplier-item awards, and these can be limited to a max # of suppliers as then it’s just computing some combinations. No “AI” required.

SUMMARY

Now, we realize this was very brief, but again, that’s because this is not new tech, that was available long before Gen-AI, which should be native in the majority (if not the entirety) to any true best-of-breed Sourcing platform, that is easy to understand — and that was described in detail in the doctor‘s 2019 article for those who wish to dive deeper. The whole point was to explain how traditional ML methods enable all of this, with ease, it just takes human intelligence (HI!) to define and code it.