Category Archives: AI

Sorry Garry, but you don’t need a Decision Ladder … you just need Busch-Lamoureux Exact Purchasing

because your decision ladder is baked in at the core! Even though we had no clue about it until you made your LinkedIn Procurement Decision Ladder post!

We reached the same conclusion you did — that decisions are not equal, especially in sourcing, and the cost of failure (and recovery, not reversibility — as recovery is never fully possible once a contract is signed, an order is made, or a shipment received, unless, of course, a Force Majeure event happens before any of that occurs) is paramount in how you handle the category in question.

Depending on the criticality of the category, and where it lies between low risk, low complexity, and low impact and high risk, high complexity, and high impact, you’re either going fast and fully automated with a high tolerance for failure (as one bad decision costs little and can quickly be recovered from) as per the first rung of your ladder or slow and methodical with decisions delayed until they are defensible and auditable at the top rung of your ladder (after all, you do have to climb up from the lower left of the lowest octant to get to the upper right of the highest octant if you are living in the Busch-Lamoureux Exact Purchasing Pocket Cube) or somewhere in between in the other six categories depending on the cost of failure and the cost of recovery.

When you know where every category falls, you know exactly how much planning, defensibility, and auditability is needed and, more importantly, how much human involvement. This makes it clear where you can play with experimental AI and where you can’t risk any decision not made by a human expert. (The machine should be used to do any and all analyses that are known and come to mind, but in high risk, high complexity, and high impact categories which have a high cost of failure and a high cost of recovery, as IBM wrote back in 1979, the machine should never make a decision because it can never be accountable for one — as that accountability always falls to you. And the courts globally are [becoming] in agreement with that.)

Exact Purchasing Helps You Survive the AI Era

In the ERP era, it was typically 60 months (i.e. 5 years) to project failure.

By then, you were (long) gone (from the role, if not the company) before the project was done. If it failed, you didn’t even know.

In the SaaS era, it was typically 18 months to failure.

In the SaaS era, you were in a different budget position in a different budget cycle and no longer responsible for the project by the time the project was done. If it failed, it wasn’t you. It was the person who replaced you.

In the AI era, it’s 1 month to failure! (And you’re going to fail. Project success rates are 6%, compared to the overall success rates of 12%.)

In the AI era, you make the decision, and before you know it, the system fails and you’re being held accountable because you’re still there, still in the role, and the project team hasn’t changed on either side of the equation.

You f6ck up, everyone knows its you, and only you (because you were in charge end-to-end), and you’re blamed for the loss.

The CFO hates you because you wasted money. The COO hates you because operations are worse than before. The CEO hates you because you made his favourite consultancy/provider partner (whose CEO plays golf with him on the golf course) look bad. And, worst of all, your team hates you as they have yet another system that doesn’t work, that they have to work around, on top of having to clean up the huge pile of sh!t it made when it was implemented. (To be expected. It’s probably just a reskin of the A.S.S.H.O.L.E. anyway.) Because there is no money left to fix it, and won’t be for three years because you overpaid through the nose to get it.

That’s your reality, unless you take extra, extra steps to make sure it doesn’t happen. Steps you don’t know to take because you don’t understand what you need, why you probably don’t need AI, and if you actually do, what (limited) AI you actually need and how to make it work in general, not just in a glorified demo based on buying butt wipes for your elderly care division.

The only way you’re going to know what steps to take is if you understand what you need.

The only way you’re going to understand what you need is to work through the Busch-Lamoureux Exact Purchasing framework category by category and outline what is required for each step of the source-to-pay+ process, then work through the (assisted) software selection process (with an expert advisor) to identify what types of solution you need, and then work through each of those solution types to determine if, and where, AI should be used, what kind of AI, and how to verify it in scripted demos (on data sets and requirements you provide and control) before you select any solution. Without going through all these steps, you’re guessing what you need, being blinded by the hype, and getting diverted to the new hotness when we both know it’s always the old busted hotness that saves the day. ALWAYS!

Vendors Steal Crappy Ideas — Please Don’t Encourage Them

Last year Joël Collin-Demers, The Channel Master, wrote a post encouraging vendors to steal his ProcureTech startup idea. Unfortunately, that idea involved the proliferation of sh!tty LLM technology and way too many vendors took him up on it.

I’m sorry to say that it was the one post I wish he hadn’t written!

Too many vendors decided to steal his idea, as evidenced by the constant proliferation of “AI” vendors believing they can wrap, or cr@p, an LLM better than the giants who have collectively spent trillions and actually deliver value.

They can’t. That’s because LLMs are fundamentally flawed. Hallucinations are core, consistency is a pipe dream (and those pipes are so dirty even Mario can’t clean them out), and you still need a considerable amount of exceptional data to get anything remotely useful out of them.

All Deepseek proved was that you don’t need to spend millions (or billions) to build an LLM — open source code and your own rack in a data center will allow you to get the same quality of results (i.e. garbage) as a mega-model if you focus it to a particular task in a particular problem domain.

The models would be small, fast, and cheap, but, just like the big models, won’t work out of the box because they are not intelligent, aren’t deterministic, and aren’t even consistent. (And let’s not overlook the fact that a subsequent iteration on a task or document might undo something they got correct in the last iteration that you approved.)

As for his examples:

  • No RFX execution — draft creation, sure, but accuracy varies
  • They’re more likely to enable fraud than stop it (see many SI posts)
  • The contract insights they return may not be the most relevant ones (and leave you blind to million dollar risks)
  • They are just as likely to make up risks as detect actual risks with new suppliers … and accuracy will vary greatly based on the data available and what you plan to use the supplier for
  • Given that they can’t think, don’t understand logic, and can’t even do basic math (it has been proven, see Apple studies for e.g.), you should never use them for benchmarks (just for data extraction from hard to digest sources, providing Intern Indy reviews the data first)

Now, if you insist on riding the hype wave, knowing that failure is likely inevitable (with only 6% of companies seeing a return from AI investments), then this is the way to do it as you’ll waste the least money proving classic tech with augmented intelligence is the way to go (while doing the least harm to the environment).

Conclusion: it’s the brilliant way to go bust! 🤣 😭

China is Leading in AI!

And the real reason why? The courts are defending labour rights and NOT allowing companies to replace workers with AI.

As per a recent posting over on “The State Council Information Office (of) The People’s Republic of China” on April 30, 2026: (Source)

“A Chinese court has ruled in favor of a human employee in a labor dispute caused by AI replacement, which experts said may send a reassuring message to labor rights protection efforts in the age of automation.”

Furthermore, this was not the first time!

On December 26, 2025, the Beijing Municipal Bureau of Human Resources and Social Security released a set of arbitration cases for 2025, including a dispute triggered by AI-driven job displacement. In that case, the arbitration panel made it clear that 𝐀𝐈 𝐫𝐞𝐩𝐥𝐚𝐜𝐞𝐦𝐞𝐧𝐭 𝐝𝐨𝐞𝐬 𝐧𝐨𝐭 𝐯𝐚𝐥𝐢𝐝𝐚𝐭𝐞 𝐚 𝐝𝐢𝐬𝐦𝐢𝐬𝐬𝐚𝐥. It found that adoption of AI technology is a voluntary move to stay competitive and not one that is mandated or acceptable as a basis for human replacement and dismissal.

Furthermore, legal scholars in China are emphasizing that 𝐭𝐡𝐞 𝐜𝐨𝐬𝐭𝐬 𝐨𝐟 𝐭𝐞𝐜𝐡𝐧𝐨𝐥𝐨𝐠𝐢𝐜𝐚𝐥 𝐭𝐫𝐚𝐧𝐬𝐟𝐨𝐫𝐦𝐚𝐭𝐢𝐨𝐧 𝐬𝐡𝐨𝐮𝐥𝐝 𝐧𝐨𝐭 𝐛𝐞 𝐛𝐨𝐫𝐧𝐞 𝐬𝐨𝐥𝐞𝐥𝐲 𝐛𝐲 𝐰𝐨𝐫𝐤𝐞𝐫𝐬 and that while 𝐭𝐞𝐜𝐡𝐧𝐨𝐥𝐨𝐠𝐢𝐜𝐚𝐥 𝐩𝐫𝐨𝐠𝐫𝐞𝐬𝐬 𝐦𝐚𝐲 𝐛𝐞 𝐢𝐫𝐫𝐞𝐯𝐞𝐫𝐬𝐢𝐛𝐥𝐞, 𝐢𝐭 𝐜𝐚𝐧𝐧𝐨𝐭 𝐞𝐱𝐢𝐬𝐭 𝐨𝐮𝐭𝐬𝐢𝐝𝐞 𝐚 𝐥𝐞𝐠𝐚𝐥 𝐟𝐫𝐚𝐦𝐞𝐰𝐨𝐫𝐤.

This is the thinking that will allow for actual progress and development.

AI is not intelligent, humans are still needed, and progress will be made when we stop accepting the BS that AI can replace us and instead only listen to and work with companies that state that appropriately designed, implemented, and/or restricted AI can augment us in our jobs and make us 3, 5, and even 10 times more effective — enabling us to be super human workers.

It might be too late for the US, but if Chinese courts continue to make rulings that indicate that 𝐜𝐨𝐦𝐩𝐚𝐧𝐢𝐞𝐬 𝐰𝐡𝐨 𝐛𝐞𝐧𝐞𝐟𝐢𝐭 𝐟𝐫𝐨𝐦 𝐀𝐈-𝐝𝐫𝐢𝐯𝐞𝐧 𝐞𝐟𝐟𝐢𝐜𝐢𝐞𝐧𝐜𝐲 𝐠𝐚𝐢𝐧𝐬 𝐦𝐮𝐬𝐭 𝐛𝐞𝐚𝐫 𝐜𝐨𝐫𝐫𝐞𝐬𝐩𝐨𝐧𝐝𝐢𝐧𝐠 𝐬𝐨𝐜𝐢𝐚𝐥 𝐫𝐞𝐬𝐩𝐨𝐧𝐬𝐢𝐛𝐢𝐥𝐢𝐭𝐢𝐞𝐬, it won’t belong before China is truly dominating the world (since the US will have no competent employees left when everything goes to hell).

Ontologies Could Have Saved Us — But in the Age of Gen AI, They Might Just Ruin Us!

What is an Ontology?

Philosophically, an ontology is the study of being, existence, and/or reality that is designed to investigate not only what entities exist but how they can be categorized.

In computer science and, more specifically, the data age, an ontology is a formal, machine readable, specification of entities, their properties, and their relationships within a domain that is used to structure information in a way that systems can share and structure it.

In the early days of semantic technology, an ontology was used to structure data in a meaningful way to allow sophisticated models to process, and make sense of, natural language with relatively high degrees of accuracy. It was usually expressed in a formal ontology language that allowed for detailed entity, relationship, part of speech, and even concept definitions. They were often defined in such a way they could be organized into interconnected libraries which formally organized knowledge into large, connected, corpuses that could be deterministically processed (hallucination free) and completely understood by any application that was capable of processing the language the ontologies in the library were encoded in.

And this was the true beginning of the semantic web, which was also known as Web 3.0, which was still in its infancy in the 2010s, but starting to take off by early (early) adopters (with almost 2% of web domains containing semantic markup circa 2014).

But then five things happened.

1. SaaS exploded, and so did the need for data, and the ability to consume it in standard formats.

2. GPT-1 was released in 2018 and the Gen-AI craze began shortly thereafter, leading us down the hallucinatory hole of incessant inanity that every consultant thought could power everything.

3. This led to the agentic craze, which increased the demand for data (and the desire to consume it in structured formats).

4. Every SaaS provider, and their dog all of a sudden needed multiple, steady, streams of data in standard formats to power their agentic applications.

5. In response, every data provider responded by adopting a simple data standard, calling it an ontology, even if all they were serving up was average scope 3 carbon data by country and factory type.

And now the term has no meaning since it’s the term used by every SaaS vendor and data supplier to essentially describe their data file structure. No formality. No relationships. No underlying structure that allows the machine to actually reason. Just another random data file blended into the data soup that feeds the hallucinatory engine that will tell us to go over the cliff like lemmings (and lead countless to their deaths as they cognitively surrender to what the AI tells them to do).

What could have been our saving grace (if Web 3.0 research had continued and true ontologies of ontologies had been created) might soon be the source of our demise as Gen-AI blends together mismatched data with flawed reasoning and produces the digital equivalent of toxic waste.