Category Archives: AI

How AI Enhances 10 Common Procurement Challenges Part I

A recent CIO article drew my ire because it claimed that AI Overcomes 10 Common Procurement Challenges as it oversimplified the problems and overstated the benefits of AI. Let’s take them one-by-one.

Procurement Takes Too Long, Slowing Innovation: According to the article, AI-driven platforms can generate RFPs, accelerate sourcing, automate approvals, and reduce cycle times … which is mostly true. Properly applied, AI can accelerate sourcing, reduce cycle times, and automate approvals … but not all approvals. As for RFP generation, that’s very limited — LLMs can generate RFPs with a simple prompt, but not necessarily a good RFP. The best RFPs are designed by humans (and then automation, which may or may not use AI, can pull in data from supporting documents as needed), and as for acceleration, it depends on the project — it can’t speed up supplier qualification where humans need to inspect the products and verify the requirements.

Moreover, a rush to AI can make things worse, and not better. Letting AI generate an RFP that misses a key requirement in terms of required certifications, performance criteria, production capacity, etc. can entirely invalidate an RFP process and lead to months of wasted effort if no human realizes that this key requirement was missed until an award is offered and a request for the certification, capacity, etc. is delivered and a “sorry, we don’t have / can’t do that” is returned.

Legal and Budget Complexities Create Bottlenecks: Budget tracking systems and rules-based automation allows for instantaneous budgetary approvals. Contract negotiation software can automate redlining, compliance checks, etc., but cannot handle a complex negotiation for a complex project where each side has a lot of requirements and multiple parties to satisfy. AI speeds up the technical drudgery, but not the human interaction.

Moreover, if you turn over negotiations to software, you have no idea what the end result will be. If you let it negotiate based on market data, and the cost data is off, you could be committing to a bad deal. If you let it predict timeframes based on how it expects prices to rise/stay high, but it’s off by two years, it could lock you into a three year deal when you only need a one year deal. And so on.

CIOs Need to Upskill Their Teams in AI and Cybersecurity: Just because “AI” can simplify processes with guided intelligence, that doesn’t mean the team is upskilled in the process. The reality is, there is no incentive for users to learn anything if they think the system will guide them in everything they need to do.

Thus, if you over invest in AI, especially the kind that guides users in every task they have to do, and works quite well on the basic tasks they have to do daily, and doesn’t screw up the first half dozen or so moderately complex tasks, the user will believe the system is almost flawless, start to trust it implicitly, stop questioning it as time goes on, start believing there is no need to learn anything else because the system knows it, and, over time, stop thinking. And then, instead of performance improving, it will decline … and that decline might be accompanied by a major financial loss if a bad contract is signed or major risk ignored.

Data Inaccuracy Leads to Poor Procurement Decisions: While it’s true that over three quarters of organizations struggle with unreliable data, AI doesn’t magically fix the problem. It can help with cleansing, validation, and procurement trend analysis, but ask any spend analysis vendor who has tried to apply an LLM to unclassified vendors about the classification accuracy (which tends to top out around 70%) — good data still requires manual cleansing and classification, especially where the system reports good confidence. It can definitely help, but it doesn’t take the onus off of the human.

In other words, if you believe that you can plug in a magic AI black box ad that it will fix your data, you are gravely mistaken. Sure it will tell you that it has cleansed, classified, and validated all of your data, but if it’s only 70% accurate, it’s only made matters worse if you trust the data 100% and don’t know what 30% is inaccurate. When you base your decisions on data, and the data is bad, you are bound to make a bad decision. The question is, how bad. You don’t know. And that’s a big problem!

B2B Software Selection is Increasingly Complex: Moreover, despite the claims, AI-powered vendor analysis doesn’t really help that much — see Pierre Mitchell’s crazy conversations with DeepSeek-Rq. Note how it not only recommends inappropriate vendors, but also recommends vendors that don’t even exist anymore … it can help you discover potential vendors, but you still need human reviews and deep pricing intelligence (from expert SaaS optimizers).

Trusting AI to select your software is worse than trusting an analyst firm map! And we know all of the problems those maps contain. (First of all, they only mention the same 10 to 20 vendors year after year, ignoring the dozens of other vendors that might be more appropriate for you.) AI cannot understand your needs, cannot truly map needs to requirements, cannot truly map requirements to features, and cannot truly assess how relevant a solution is, and definitely can’t assess how well a provider’s culture will match yours.

Come back Thursday for Part II!

We Finally Know the Source of the AI Buzzword Bullsh!t!

The Agentic Software Service Hyper Optimized Learning Engine custom built for drowning the World Wide Web in soundbite and buzzword marketing bullsh!t centered on AI, or the A.S.S.H.O.L.E. for short! (With fervent thanks to the esteemed Arthur Mesher for delving deep into the depths to uncover the source of this madness!)

Technology Project Failure is at an all-time high, boosted by the recent AI failure rates (which are on the rise as almost half of AI initiatives are being scrapped in process, see CIO Dive), and while the hype should be subsiding (and shifting to the next hype cycle), it’s now hitting us harder and faster in what should be its death throes than any hype cycle that has come before.

The AI marketing onslaught is coming so hard and fast that it’s impossible to imagine how so much new soundbite, buzzword, FOMO, and FUD content can be produced so fast and so overwhelming to the point that it seems humanly impossible. And that’s because it is. It’s not coming from humans, it’s coming from the A.S.S.H.O.L.E.. As we have indicated in our previous posts on Gen-AI LLMs, one of the valid uses for Gen-AI is mass content digestion, search, summarization, and generation.

It appears that one of these systems was customized to ingest all of the initial human-generated AI BS and trained to spew out marketing soundbites, social media posts, articles, and other forms of web content ad nauseum and to continually ingest new content on the subject to create even more content, including AI-generated BS content from other AI systems that tried to copy the original A.S.S.H.O.L.E..

And even though it doesn’t matter, since apparently every LLM can be trained to emulate the original, the only question that remains is, who currently owns the source engine, what LLM was it originally built on, and what LLM is it running on now? This is obviously the industry’s best kept secret. I hope someone who has gotten to the bottom of this will let us know the full story of the A.S.S.H.O.L.E.. Considering the intellectual and financial pain and suffering it has caused, we deserve to know the truth!

For those interested, since I’m sure LinkedIn will disappear Art’s post if it hasn’t already, here’s the original. (And the Gartner rant ain’t half bad either!)

The Best Way to Survive the AI-Powered Apocalypse? Go Old School!

If you’ve been following along, you know that a great purge is coming on two fronts. All the pundits agree on that! On the first front, a large number of vendors are going bye bye, as we’ve been telling you since our first post on the Marketplace Madness. On the second front, they took ‘er jobs. Except it’s not they, it’s AI.

So doesn’t this mean that if you want to survive the days ahead that you should find the most advanced AI provider that isn’t going to get purged in the near future, adopt the tech, replace as much staff as you can with AI, find a way to survive the hardship, and come out ahead when everyone decides that what they have to do?

Well, for the vast majority of the analysts and pundits, it is exactly what you should do — and do it right now. It’s AI overload all the time. And just when most hype cycles start to die down, this one gets a second wind of hurricane proportions.

But, in fact, it’s the last thing you should do. In fact, you should implement a Gen-AI ban and Agentric AI ban immediately, and identify classic ML-powered AI augmented intelligence tech that can supercharge your team, acquire it, and train your team on that immediately. Because you can get the same results as any Agentric AI can get if you employ the right classic ML-powered human-driven AI technology with the right algorithms, analytics, optimization, etc. Sure, a human might be a little bit slower than an algorithm that can work 24/7/365 without a break, but human who is appropriately skilled and trained will make up for this with something the AI doesn’t have, true intelligence.

You see, the thing about Gen-AI and Agentric AI is that it works great until it doesn’t. As per our recent post, Gen-AI is full of problems. In a recent post, we noted that, Gen-AI can:

  • get you sued
  • increase the chance you will be hacked
  • result in Million/Billion-Plus processing errors
  • shut down your organization’s systems for days
  • help your employees commit fraud

And those are the good side effects from its hallucinations. There are much worse side effects that can happen. If you refer back to our posts on the valid uses for Gen AI and the valid uses for Gen AI in Procurement

  • the embedded biases, that you might not even be aware of, could result in decisions diametrically opposed to what you are expecting
  • when it computes two options that are equally likely to generate the same end result for the company relative to the KPI it is using, there’s no guarantee it will select the right option — and there’s always a right option, especially if one option for cost savings is a longer term contract so the supplier can upgrade equipment and the other option is forcing the supplier to cut an already razor thin margin 50%
  • the hallucinations eventually become real, as the systems get so advanced that they not only create super realistic evidence to back up their recommendations, but take over your entire systems in the background so that you don’t know that a web request to verify a claim is actually still being processed by the AI that is now running in the background
  • it starts negotiations and cutting contracts you haven’t even authorized yet
  • it becomes you … and you get blamed for all its mistakes

In other words, ignore the Gen-AI and Agentric-AI technologies that are not the miracle cures they are promised to be. The miracle cures are the last generation ML-based AI technology that was just about to transform your operations under the expert fingers of your leading practitioners, not some probabilistic monstrosity that requires an entire data center to run to generate an output no one verify using a system no one understands. Hone your chops on those and you’ll get the results you need, without having to deal with unexpected, possibly catastrophic, failures along the way.

After all, when we told you about all of the great advancements that were coming in Source To Pay in our classic series (indexed here), none of it required Gen-AI to achieve!

Yes, Gen AI will Have to be Consumed By …

Orchestration along with Intake if any of these loud, overfunded, mostly useless (but, unfortunately, not mostly harmless) startups are going to survive!

Yes, the doctor said it and yes, it’s totally true.

So why this diversion? the doctor was recently asked a variation of the question by a very knowledgeable, observant, and forward thinking executive with a track record of getting it right (and growing companies) who wanted to know if he was grasping the situation accurately and likely correct about how this whole mess is going to shake out once the mass extinction begins later this year/early next year (where the doctor is predicting at least twice the typical percentage of failures, rivalling or exceeding that of the first mass extinction post the funding frenzy and market crash of 2008, as well as a large number of mergers that will happen just so companies can partially survive; and where THE REVELATOR is predicting less than one fourth of companies will make it through unscathed, because the space cannot support 666+ companies).

As the doctor has previously penned in Marketplace Madness is Coming Because History WILL Repeat Itself:

Stand-alone Intake(-to)/Orchestrate solutions, the current poster children of the space, will soon have a fall from grace (and only the smart will survive)! Call me Scrooge if you like, but there’s a logic behind why I’m developing a bah-humbug attitude towards most of these. And it goes something like this.

Intake

  • Pay For View: if modern procurement solutions are completely SaaS, then they should be accessible by anyone with a web browser, so why should you have to buy a third party solution to see the data in those applications? Wouldn’t it make more sense to just switch to modern source to pay solutions that allow you to give variable levels of access to everyone who needs access instead of paying for two solutions AND an integrator?

Orchestrate

  • Solution Sprawl: while orchestration is supposed to help with solution sprawl, it’s yet another solution and only adds to it. Wouldn’t it make more sense to invest in and switch to a core sourcing and/or procurement platform with a fully open API where all of the other modules you need can pull the necessary data from and push the necessary data to that platform?

I2O (Intake-to-Orchestrate)

  • Where’s the Beef?: Talk to an old Pro who was doing Procurement back before the first modern tools began to be introduced in the late 90’s and they’ll tell you that they don’t get this modern focus on “orchestration” and managing “expenses” and low-value buys because, when they were doing Procurement, it was about identifying and strategically managing multi-million (10, 50, 100+) categories where even 2% made a significant improvement to the bottom line, and way more than 10% on a < 100K category.
  • Where’s the Market? This is only a problem in large enterprises — right now, many of these I2O solutions are going after the mid-market who are eating it up because of ease of use, but as soon as they realize the emperor has no clothes, and there’s no support for real strategic procurement (yet alone strategic sourcing) and you have to go out and buy more platforms, what’s going to happen? The reality is that the mid-market is better served by a federated catalog management / punch-out platform, or next-gen marketplace (they’re coming, tech is cyclical like fashion, and it’s due) and will likely be better served still by a new breed of e-commerce B2B solutions for end-user Procurement.

Moreover, as the doctor has penned in many posts, Gen-AI is only useful for tasks that ultimately reduce to

  • large document/corpus summarization
  • large document/corpus query
  • language translation (including natural to system and system to natural)

That’s why the doctor listed so few valid uses in More Valid Uses for Gen-AI … this time IN Procurement!, and why most of those were utterly useless such as:

  • Create meaningless RFPs from random “spec sheets”.
  • Auto-fill your RFPs with vendor-ish data.
  • Generate Kindergarten level summaries of standard reports for the C-Suite.

In other words, on its own, each technology is mostly useless. (But not mostly harmless. On its own, consistently misused, Gen-AI is very harmful. See our other articles for a discussion of that.)

  • Intake is useless on its own because capturing an input is worthless if you can’t do anything with it
  • Orchestration is useless on its own because it’s yet another piece of SaaS you need to maintain that provides no value beyond linking two or more pieces of software together that could both be linked direct through their APIs (since it couldn’t link the software in the first place if it didn’t have APIs)
  • Gen-AI is mostly uses on its own as most of its valid uses are in CLM or RFP query (not creation!), which is only a small part of the S2P cycle

However, if you put it all together, and do it right, the whole may be more than the sum of its parts.

If it’s all expertly glued together:

  • Gen-AI creates a natural language interface where a user can make any type of request, not just a purchase request, that is translated to a standardized system format
  • Intake can process those formats, ensure completeness (relative to the needs of the different enterprise applications and modules that are integrated), send complete requests to the orchestration module, get back the responses, and feed them through the Gen-AI interface to translate them to natural language before being fed back to the user
  • Orchestration links all the applications in a way that directs the request to the right application, or application chain, ensures it gets properly processed and executed and ensures the right results get returned to the right applications in the chain and, ultimately, the user … providing, of course, it’s enterprise wide back-office orchestration, NOT just Procurement!

Which means that the only way any of these players are going to survive is if orchestration gobbles it all up AND does it right.

Gen-AI Won’t Work For Procurement … And Neither Will Agentric AI if the foundation is Gen-AI!

Right now every vendor is pushing “AI”, and the vast majority of that “AI” they are pushing is a Gen-AI LLM, and often that is just a wrapper of a third party Gen-AI LLM, like Chat-GPT (which only the French know how to pronounce properly).

And they are pushing this as a cure-all for all your procurement ills. It’s the new magic elixir. The new panacea. But, in reality, it’s the ultimate silicon snake oil, because it almost works. And it makes you feel really good when you use it. In medical terms, it’s not a treatment, it’s a psychedelic that takes all your pain away (until it wears off that is). But, just like the spoonfuls of LSD that allowed Bender to become the Iron Chef, it will only last long enough for the vendor to win the contract from you, and then it will start to fade. Until it fades completely when you need it most and fails you utterly when you need to figure out how to deal with a border closing that just happened, a critical raw material shortage due to an unexpected natural disaster, or a trade war no one saw (but should have seen) coming.

This is because, as we keep telling you, Gen-AI, which was built as a predictor technology to predict what block of text, in natural language, should follow an existing block of text (using chain-of-compute), based on training across a very large corpus of existing documents. It’s no more, no less. That’s why it’s only good for tasks that can be reduced to large document search and summarization. (And natural language translation tasks, because it understands basic semantics and can easily be trained to translate to and from any machine language you train it to.)

However, this doesn’t help you with any task that requires actual computation! It’s not analytical data processing, it’s not optimization, and it’s definitely not advanced machine learning for advanced mathematical pattern detection. These are the majority of your tasks and the tasks you need to do to analyze a situation. Buys should be based on the lowest total cost of ownership at the maximum acceptable risk level. Sales predictions, and thus demand, should be based on tried and true mathematical trends, not hunches or market hype. Basic invoice processing should be against business rules for validation, approval, and payment, and that should be primarily based on rules-based automation.

Note that none of these core technologies you need to solve the majority of your problems are AI, as we pointed out in our recent article that said you don’t need Gen-AI to revolutionize procurement and supply chain management. Not to say that these technologies can’t be enhanced by the right application of AI — for example, AI could predict the optimization paths most likely to arrive at the optimal answer, the right curve fitting algorithms to match the trend lines, and the right outlier analysis to identify missing, off, or fraudulent information.

Real solutions come from real tried-and-true AI technology developed over years, or decades, that was designed to solve a specific type of problem, not generic text processing technology that was not designed for the problem, has no understanding of the problem, and will make stuff up in an attempt to solve the problem (which is referred to as a hallucination, but is not a bug, but a core feature of Gen-AI / LLM technology).

This is also why Agentric AI built on Gen-AI won’t work — you can’t automatically build an RPA sequence from a chain of compute that could be completely hallucinatory, and you certainly can’t rely on it to solve your problem.

This doesn’t mean there isn’t a use for Gen-AI, it can be trained to be a natural language interface to these other tools that will work reliably the vast majority of the time (say 95%+ if trained over time), but the use is definitely NOT what you are being promised.