Category Archives: AI

The Dark Ages Were Bad …

… and, after most of western society was likely still recovering from the long term devastating effects of the volcanic winter of 536, that probably set us back 1,000 years in the grand scheme of societal development and civilization advancement.

… but that’s a minor setback compared to what’s in store for the Age of Retardation that is coming!

But let’s back up. Consider this recent article on LinkedIn by Karl Waldman on this Medieval Lesson: Cutting Skilled Workers Hurts Long-Term Growth where Karl discussed why the age of great cathedrals came to an end.

It had nothing to do with lack of wealth — there’s always been wealth, all that changes is who controls it — or a lack of interest — the Christian religion has consistently held more than its fair share of dominance through Europe from the building of the first great cathedral until the present day (and whenever it loses control in one country it finds a new one to take over). It was lack of skill.

As per the post, the European cathedral builders developed an ornamental tradition so specialized it took decades of guild training to master. When the Black Death killed a third of Europe’s population, the skilled tradesmen disappeared because the training pipeline that produced it had been destroyed.

Now think about what we’re doing today.

We’re pretending AI can do the work of experienced professionals and cutting them left, right, and centre. We’re pretending we don’t need junior workers (because they do the tasks that AI seems to do okay) and not hiring. We’re walking all of our institutional knowledge out the door, as well as our ability to react and fix exceptional situations with creativity (that will break AI when they arise), while ensuring there’s no one around to absorb even a morsel of that knowledge and skill.

We’re not only replicating the end results of the black plague at a rate that’s even faster than the black death spread across Europe (it took about 7 years with the first 4 being the worst) — and not only are we destroying all of our capability to build tomorrow’s businesses, but we are throwing away all of our capability to even maintain today’s businesses if something goes wrong! After all, our current staffing levels are minimal, and most of the people we have left are in cognitive decline thanks to the AI they are being forced to use for “productivity” reasons.

When the next unstoppable pandemic hits, and wipes out all of our silver haired experts with no skilled talent to replace them, we will enter the Age of Retardation and our global society will collapse faster than the Aztec Empire. (And if you don’t know how fast one of the greatest civilizations in Central America fell, maybe you should brush up on your history!)

You Need Automation. But You Don’t Always Need Agentic and You Almost Never Need Gen-AI!

In a previous post we dove into how analytics must drive source to pay, because most of source to pay should be automated and touch free as most of the source to pay process is straight forward (and capable of being automated for the last decade), non-strategic, and low to medium value.

Strategic Sourcing is an activity that should be focussed on high risk, high complexity, and/or high value categories and occasionally focussed on medium risk, medium complexity, and/or medium value categories where there is incomplete information or insufficient product/category history, atypical turbulence in the market, or highly particular requirements that just came into effect as a result of new regulations. That’s a minority of products/categories, not a majority.

Procurement should only be focussed on significant exceptions. And, with proper, modern, systems with proper e-document integration and exchange, most of the documents should be arriving in standardized digital formats, and most of the processing should, thus, be fully automated. And most of what is non-standard will be PDF in relatively standard formats that LLMs will be able to process to 95% accuracy and only require a few human verifications and field completions. The days of 20 people invoice processing team should be long gone, as the tech, even for standardized PDFs, has been in production by the leading players for over 8 years. Invoice discrepancies can be auto-identified, suppliers auto-notified, suggested corrections auto-included, one-click acceptance emails/screens for the suppliers included, and most contingencies accounted for. Only in the rare situations where suppliers refuse to accept a correction, invoices are in very non-standard or handwritten format, payments don’t go through, etc. should a human need to get involved. However, 95% to 99% of all documents and transactions that flow through Procurement should be 100% automated.

But most of this doesn’t need experimental Agentic AI or Gen-AI. Classic RPA will do just fine. For most of the rest, Adaptive RPA, with a bit of Machine Learning / Auto-Suggestion based on human-based exception processing, will do the trick nicely. If you look closely at current generation (A)RPA, Machine Learning, Optimization, and Predictive Analytics and walk through the full source-to-pay process, there is very little that can’t be automated without Gen-AI LLMs or experimental Agentic Systems. Sourcing — there are many standard (seven step) processes that can be completely automated based on data analysis, data-based risk assessments, goal definitions, and optimization. RFX (including e-Auctions) can be fully automated and, from the time you specify a product/category to source, everything can be automated to the award (including the demand pull/calculation from other systems).

When it comes time to contract, if you have standard templates or a large clause library, the system can automatically create the contract from the template and RFP responses, integrate DocuSign, and auto-execute it. If you don’t, or if you have to use the supplier’s paper, then you might use an LLM to create a draft for human review and/or analyze the supplier’s paper for terms, pricing (to make sure it matches the bid) and potential risks, as well as suggested revisions, before you sign. Gen-AI/LLMs unnecessary, but useful on a point-basis if you don’t have a good historical equivalent of a solution like Coupa Exari or iCertis.

Supplier onboarding can be fully automated with RPA powered dynamic workflows and third party data ingestion, as can risk and compliance analysis — no modern Agentic solutions needed.

Then we get to automatic invoice monitoring and point-based re-orders, receipt creation from inventory integration, and invoice processing in e-Procurement which has all been around for at least a decade. Automated approvals subject to tolerances, rules and pre-approvals — as well as predictive analytics on payments for new or one-time suppliers/orders or (slightly) out-of-tolerance invoices can automate the entire invoice-to-pay process.

We can get through the entire process on best-of-breed, classically oriented, RPA tech with some machine learning that processes human decisions in exception management, alters or augments the rules (and guardrails), and auto-processes the same type of situation next time. We quickly get to 95%+ throughput for any task that should be mostly automated, and a top human employee with BoB (A)RPA solutions and some augmented intelligence packages for analytics and research becomes 10 to 20 times as productive as they would have been in the past.

That’s the real future of Procurement. Small, top-talent teams (mentoring small emerging top-talent teams) doing the work of teams five to ten times their size, doing it better, and delivering more value than anyone would have believed possible with best-of-breed tools. Not error-prone, hallucinatory, agentic systems that work well in demos and a few select categories, and go all over the place in reality (and then try to hide their mistakes like Nick Leeson [who single-handedly collapsed Barings Bank] until they do a modern equivalent of the 2005 J-Com trade and cost you hundreds of millions of dollars on your key billion dollar product line).

So while you need to modernize at all costs, you don’t need to go full Agentic on unproven solutions. Get 90% of the way on tech that has been proven where you can control the automation level until you get comfortable with automation and learn where you can safely hand tightly boxed “decisions” to the machine (where well-defined calculations would determine your decision the majority of the time) and where you can’t. Otherwise, you’ll just end up being another member of the 94% AI failure camp. That’s not a statistic you want to be part of, especially given the cost of this tech today (and the increased cost tomorrow as energy grids start to break and the compute costs for modern AI tech goes through the proverbial roof).

Another Reason To Avoid AI: NO ECONOMIC GROWTH COMES FROM AI!

A recent study by Goldman Sachs, summarized in Fortune, found no meaningful relationship between AI and productivity at the economy wide level/.

Think carefully about that. 450 Billion, which is more than the GDP of over 100 countries, was sunk (and I mean sunk) into AI last year — for the net result of ZERO economic growth. For 1/6 of that, every college in the US could be free — and you’d have 20 Million smarter adults with no student debt dragging them down, causing them stress, and zapping from their productivity. For 1/12 of that, you could eliminate all the hunger and food insufficiency in the US. For 1/50 of that, you could re-open Alcatraz and provide a King with his own special castle and his own moat.

In other words, there are so many better things that could have been done with that money — including training your people to be more productive, modernizing processes for efficiency, and building deterministic tech that actually works at 1/100 to 1/10000 of the compute power in a data center that’s already powered up.

The only company “winning” is Nvidia, who provides the chips, which means that most of the money is going to its factories in Taiwan and South Korea, and those are the only countries that are actually winning while Americans, who were laid off in droves last year, get poorer, colder and hotter, hungrier and thirstier (as AI sucks up all the energy, which is now not available for heating or air conditioning, and all the water for cooling, which is now not available for drinking or farming).

Think about that the next time you think an overpriced clod or chat, j’ai pété wrapper, even if hyped up as an AI Employee by the A.S.S.H.O.L.E., is going to solve all your problems. Especially since all the Age of AI has done for us is make us dumber, poorer, and less prepared for what is to come next than any age that has come before.

The One Big Benefit Of NOT Going AI …

You don’t have to worry about your AI vendor going toes-up when power costs go through the roof and your AI vendor can no longer charge pennies for compute when its costs rapidly become dollars and it can’t pass them on due to contractual commitments to existing clients (or to new clients who won’t pay dollars for computations that might return hallucinations).

The new generation of AI tech — Gen-AI LLMs / AGI — requires way more compute power than the last generation, 100 to 10000 times more on average, for most requests. Grids are stretched and beginning to break. We’re at the point where only nuclear can power the data centre needed for a modern Gen-AI/AGI offering. And, as per Koray Köse’s recent article on AI leadership is about who controls the power, U.S. nuclear plants operated at 92.3% capacity last year. OUCH!

THERE IS NO ENERGY LEFT!

You can’t build a new nuclear plant overnight — if you can even build one at all anymore! Last year, DOGE’s Firing Fiasco at the NNSA stretched an already stretched organization even more. Many returned to work, but not all, but budget cuts likely left them without the capacity to even properly monitor existing aging nuclear infrastructure, yet alone approve more plants.

And it’s not even clear how much know-how is left in the US to build new plants. The Vogtle Units 3 and 4 in Georgia were the first units built from scratch in over three decades. The experience and expertise isn’t there to safely build these plants en-masse.

And the last thing the US wants to risk is another meltdown. Three Mile Island wasn’t a Chernobyl, but all it takes is a rushed private sector job with a lack of proper oversight and testing and one small mistake to trigger the next meltdown on US soil.

In other words, the power isn’t there for more AI.

So those organizations that can do without modern AI, that can use classic solutions with fit-for-purpose last generation AI that requires a fraction of the power and can run on already strained, non-nuclear, grids will be the big winners when the power squeeze hits and the Big AI players start dropping like flies.

AI is Exacerbating the Need for Global Data Centres NOT Controlled By US Firms!

A recent post by Joël Collin-Demers on why Your LLM Doesn’t Need a US Passport pointed out two very important facts that you’re probably not aware of but should be:

1. Your company is feeding sensitive data to US-based LLMs every single day.

2. The US CLOUD Act lets American authorities demand data from any US-based provider REGARDLESS of where their servers sit in the world!

In other words, you’re giving the USA full access to all of your proprietary and confidential data anytime they want it — in full breach of your data localization laws if you’re NOT in the US and in a country with such laws (and if you’re not in the US and don’t yet have data localization laws to adhere to you will soon have such laws to deal with as a result of the US global over-reach for your data to feed its AI).

This is not just an AI problem (which, if you think you really need, you have other non-US options if you are not a US company as per Joel’s extensive list), it’s an overall SaaS/SaS problem. If you’re not a US company, you need to make sure that not only your data, but all of your applications (including, but not limited to, AI) are hosted in non-US owned data centres off of US soil without safe harbour agreements.