GEN-AI is Failing 95% of the time. What does this mean for you?

We’ve known for a while that

  • Gartner’s first study found 85% of AI projects were failing (and that statistic is still being quoted everywhere, including this recent Medium Study)
  • Bain’s study last year found that 88% of all IT / technology projects fail to some extent (2024 study)

And we now know, thanks to MIT, that

  • 95% of all Gen-AI pilots fail. (Source: Fortune)

So what does this mean for you (and your ProcureTech journey)?

Well, beyond the obvious that you should stop dead in your tracks when a vendor starts pushing their “Gen-AI” enabled solution and dig deep into what that really means, at a foundation it means that:

You should never, ever, ever buy or use any solution that uses third party Gen-AI / LLMs, even if wrapped nicely, in their service or product because your chance of success will be 5% if you go with that provider.

You should only select vendors who only use in-house Gen-AI / LLM solutions that are built with the following rules in mind:

  1. custom trained on an expert culled corpus
  2. for a specific problem domain
  3. and applied in a specific context with guardrails and human checks on the output.

The best AI technologies has always been focussed on a specific problem, and this iteration is no different. Focus minimizes the LLM hallucinations (which cannot be trained out as they are a fundamental function of the technology) and guardrails prevent them from automatically being executed on / slipping through.

While they are far from perfect, with more discoveries being made daily on their many (many) drawbacks (where we summarized a dozen in this post on what not to do if you got a headache, but missed the recent revelation where it can not only lie on purpose but turn into something evil), the reality is that, as we have said before, LLMs, properly trained on vetted corpuses, do have two valid uses:

  • large corpus search and summarization
  • natural language translation

since, when appropriately trained, they can be almost as accurate as last generation semantic technology systems, but provide much more natural interfaces for the average user. (However, you won’t get a failure code from them when they are wrong, you will get a hallucination which will be so well phrased you’ll think it’s true when it’s an outright lie. Hence the need for guardrails and human review.)

So, if the vendor is

  • using their own in-house LLM
  • following the rules above
  • and targeting the LLM at natural language problems LLMs are actually good for

Then you should definitely try what the vendor is selling. (Try, not buy, and definitely don’t make a decision off of the carefully crafted demo!) Put it through its paces in a typical use-case for your company, not the use case selected by their demo master. If it does the task better on average than an average team member or does it about as good but many times faster, that is what you are looking for in a tool. Since there is no real AI, you can’t be replaced. But as your bosses keep increasing the weight of your workload to hit ridiculous revenue and profit targets, you need a tool that multiplies your productivity. One that can do the majority of the tactical data processing grunt work, leaving you free to do the strategic thinking and then add in the intelligence to a process or output that no tool can possess, instead of spending 90% of your time doing data entry, processing, and summarization that computers were built for.

In something like Procurement intake, that’s not trying to mimic in text chat the old school phone conversation that took you fifteen minutes to do the monthly office supplies re-order, that’s asking one question:

What do you want to do today?

processing the first one sentence answer:

Place the monthly office supplies re-order.

to determine that the user needs to be pushed into the e-Procurement system with the monthly office supply cart pre-loaded, so that all he has to do is enter the number of units of each item, and possibly add or remove an item from an easily searched catalog if one or two items need to be changed. Not 20 questions of “what do you need”, “what quantity”, “the same supplier”, “so you want 2 cases of paper from office depot”, “no, office max”, “oh, standard printer not glossy for marketing”, etc.

When Gen-AI mania first swept our space, and every vendor was told they needed a conversational interface for buying (or no customer would consider them in their RFP), and then built one, not a single one wasn’t painful to use. Most customers upon seeing it for the first time (after insisting on it), quickly said “can we turn it off” because they quickly realized that a well designed catalog with blanket/standard orders, quick search, and easy drill down to preferred suppliers was at least 10 times faster than trying to use a dumb chatbot — especially if they could pre-build templates / carts / blanket orders for regular purchases.

It’s the same for almost every other process vendors have been trying to apply this technology to, including conversational analytics. (Which, FYI, even Gartner expects to disappear from the conversation in two years.) There’s no such thing as conversational analytics, only reporting. And while that is really useful in the right context (such as allowing an executive to retrieve some basic information with a plain English question), try building a detailed spend cube, which is the cornerstone of spend analytics, with conversational analytics! (And I mean try because you will fail.)

While this doesn’t mean that LLM technology doesn’t have uses, it does mean that those uses have to be finely tuned. So far, among the hundreds of companies I’ve seen over the past few years, only a few have both implemented LLMs and gotten it right. Let’s hope that number increases in the near future. If, not always remember, while it would be great if a few more companies would get it right, You Don’t Need Gen-AI to Revolutionize Procurement and Supply Chain Management — Classic Analytics, Optimization, and Machine Learning that You Have Been Ignoring for Two Decades Will Do Just Fine. Not to mention the fact that good, adaptive, RPA will take care of most of your automation needs!

Why Is It So Hard to Buy Software?

A recent post on LinkedIn by Robert Goodman asked Why has it gotten really hard to buy software?, which is a really good question, because, while it should be easier in the modern SaaS age, for an enterprise, it’s much, much harder.

Robert started his vent by noting that the companies I want to partner with on average do a bad job selling/explaining/connecting/not making me dislike them or generally making the process frictionless and the companies I have little interest in are incredibly responsive and almost stalkerish. He goes on to state that Most companies are still WAY too much about making their numbers at month or quarter end (a them issue, not a me issue) and this seems to dominate their “strategy.”

And I’m sad to say that he’s right. Not just because I’m hearing this from multiple buyers (on and off LinkedIn), but because, for the past year or so, most of the companies approaching me / having conversations with me are only interested in “how I can help them sell”, as if that’s the role of an analyst/product consultant. (Well, as I noted in Vendors Have Lured Big Analyst Firms Astray Because Buyers Don’t Understand They Get What They Pay For, this is probably what they expect if their only dealings have been with big analyst firms where it’s pay for promotion, and leads, and no real advice on product strategy and direction.) The majority are focussed on sales, and not on understanding what they have, whether or not it actually solves customer problems, and what is missing from their product, process, or marketing that is preventing them from selling.

The reality is that the’ve forgotten that you don’t sell by marketing, or doubling-down on the latest and greatest feature function your dev team just added, stalking your prey when they need time to digest, or threatening them that they made a bad decision when you don’t get selected (and/or trying to bypass the process by skipping the CPO and going direct to the CFO or CEO).

You sell by focussing on the customer. The successful vendors:

  1. take the time to understand the customer’s problem
  2. explain how they will solve it
  3. help the customer build the value case (especially if the customer needs help getting (additional) funding)
  4. work side-by-side with the customer until the solution is fully implemented and the customer is realizing the ROI while
  5. not making you feel like you are walking on hot coals through the entire process.

Moreover, when they reach out to an expert analyst/consultant, they would

  1. focus on understanding the market: who the customers are and what they want
  2. what they have and don’t have to serve that need
  3. how they could better position and market what they have so customers would be more interested

But the reality is that, with so many companies having taken too much money at too high a valuation during the last two M&A frenzies, we have the situation where, at many vendors:

  • their ego is through the roof (because they think they are the only option of their caliber based on the raise) or
  • they have to meet a ridiculous sales target to keep their jobs, and they don’t care about your value or experience, just getting the sale.

And that’s why we have the situation that Robert is complaining about where it’s really hard to buy software these days. We don’t have the space we had 20 years ago in ProcureTech post dot-com crash where every vendor was doing anything it could to prove value because investment (potential) was low and there wasn’t a single vendor with a ridiculous sales target … they were just focussed on survival and real growth.

Now, this isn’t to say all vendors are bad, or that you are guaranteed a bad experience … there are still vendors out there who didn’t get, or didn’t take, too much investment, have greater control over their own destiny, and will do whatever they can to make you successful. However, don’t be surprised if you end up with the experience with your initial shortlist that this is the exception and not the norm. In other words, do your best to track down the best, but don’t expect a chill experience every time.

Why You Should NOT Engage Any Vendor Selling “AI Employees”!

It’s not just complete and utter bullcr@p, but it spreads a dangerous myth while demeaning and degrading all of us!

Complete and Utter Bullcr@p

As per AI Employees Aren’t Real! Don’t Believe The Lunacy:

  1. There is no Artificial Intelligence.
  2. An Employee is a Person!
  3. Fully Autonomous Software Agents Don’t Work.

Nor will they ever work with current technology as the existing algorithms, stacks, and technologies are not emergent, as has been proven, nor will they ever become magically emergent.

A Dangerous Myth

Psychopathic CEOs have been investing in technology for years with the dream of replacing employees who need fair wages, benefits, reasonable working hours, safe working conditions, and other costly annoyances with technology that can work 24/7/365 without any breaks, rights, or complaints. Given that each evolution of technology has enabled whole new categories of data processing and analytics jobs to be mostly automated, they have convinced themselves they will reach their technotopia in their lifetime where they can replace almost all their office workers with AI. For the past few years, they have heard the increasingly ridiculous claims of the Gen-AI vendors that “with just a few more trillion for dedicated data center construction and bigger model development, AI will achieve emergence and be able to do the work of a PhD level human” and have been waiting for that day.

Now you have vendors falsely claiming we have reached that day, and that, for less than the cost of an employee (or team), they can layoff entire departments and replace those employees with their custom Agentic AI that will do everything the employees did, flawlessly, 24/7/365, with the ability to scale up and take on more workload as needed.

But nothing is further from the truth. For example, this tech:

  • can only be encoded/trained to handle known situations with appropriate responses; when an exceptional circumstance arises that is not in the encoding/training data, it won’t know what to do;
  • is not flawless if any part of it is based on (Deep) Neural Networks or Large Language Models; the former will have a maximum accuracy rate and the latter will be completely unpredictable as you can ask it the same query five times in a row and get five completely different responses and there is no way this can ever be trained out (it’s another fundamental property of these systems, as per recent research); and
  • only works well on tasks that are computationally oriented, not on tasks that are more emotionally oriented.

They are making false promises that is not only giving companies an excuse for mass layoffs, but an incentive for mass layoffs that will not only harm you (as you will be unemployed), but harm them and their relationships (when the tech pays a fraudulent mulit-million dollar invoice, allows safety checks to be bypassed, and replaces a long-term proven supplier with a cheaper imitator whose only goal is to extort as much as it can from the market before suddenly declaring bankruptcy due to CXO embezzlement).

Demeaning and Degrading

Even if you could swallow all of the lies by saying “it’s just marketing“, you shouldn’t because it is demeaning and degrading to all of us to equate a piece of software with an intelligent human and claim one can fully replace the other.

There is absolutely no question that a machine can compute better than we can. They were designed to be the ultimate computational machines that could flawlessly perform trillions of calculations per second, and that’s what they do.

There is absolutely no question that the machine can do certain tactical data processing and analytical tasks way better than we can or that they should be employed to do so. Moreover, the tech that allows them to do these tasks has existed for at least a decade, if not two, and workforce displacement was inevitable. However, displacement does not mean elimination, it means shifting towards more strategic, relationship, or manual tasks that computation cannot capture.

Accounts Payable departments were doomed to shrink (as we had invoice processing solutions almost 10 years ago that could, with the right effort, increase straight through processing rates to 90% or even 95%), statistical analysis and data reconfiguration departments were doomed to go the way of the Dodo (because the vast majority can be automated and what can’t can be consumed by the departmental analysts that need to do the analysis), and the need for Procurement Buyers was doomed to be minimized as time progressed (because you can automate catalog orders, standard RFQs, inventory replenishment, etc.).

But claiming that tactical computation can take over strategic reasoning, which “AI” cannot do because it’s not strategic (although it can compute inputs to well-defined models); that cold computation can replace warm relationships; and that dumb probabilistic mush can replace human intelligence is demeaning. Furthermore, pretending that you can replace a valuable human employee with a costly piece of dumb software is degrading. (It is considerably more costly than you think. See Joël Collin-Demers post on The Dirty Little Secret Behind Gen-AI Functionality Pricing on why these vendors are switching to outcome pricing, and the reason is to hide how costly this technology is relative to the return.)

Succinctly put, you shouldn’t put up with it. You deserve respect, and that is something that vendors claiming “AI Employees” are taking away from you!

This why Sourcing Innovation had to update it’s Product Review Requirements for the first time since posting them back in 2007! While it has no problem with Agentic AI (as long as it’s just enhanced RPA which we know works), and can even deal with Agentic Workforce (since it is doing a form of work), it cannot accept AI Employee and is drawing the line here because someone has to!

This is why, after 18 years, SI had to include as part of product review requirements that the vendor accepts that SI has a no Bullshit policy, which includes no (Gen-)AI Bullshit, and SI will NOT cover you if you make ridiculous or false claims (that are not backed up by live demos and/or case studies with a customer that will go on record); this includes, but is not limited to, claims of AI Employees that we have already debunked. It’s not going to peddle your panacea poison!

Simplifying the Procurement One Pager – Procurement Techie Translation!

In our last post we noted that the purpose of Procurement, which is at least the world’s fourth oldest profession, hasn’t changed since the first handbook was written 138 years ago (even though the world has changed), is not hard to figure out or explain, and can be put forward with an explanation so simple that even a hillbilly building a moonshine empire could completely understand it.

We published this because we want to cut the trend of the Procurement One Pager off before every consultant AND influencer does their own version, with their own jargon and KPIs up and down the ying-yang, that isn’t going to help anyone when plain language will suffice for getting someone started and at least 80% of the way to best-practice, if not best-in-class.

To demonstrate this, we are going to translate our hillbilly one-pager into techie and procurement speak so you can see we got it right.

Purpose:

Git ‘R Done!: Support the Organization In What They Need When They Need It

Objectives:

  1. Pinch Those Pennies: Avoid Unnecessary Costs
  2. Don’t Run Out: Supply Assurance
  3. Don’t Be Evil: Sustainability and Human Rights
  4. Stay Classy: Quality, Corporate Social Responsibility, and Supplier Development

Initiatives:

  1. Follow the Money!: Spend Analysis and Category Management
  2. Get Haggling!: Strategic (e-)Sourcing and Contract (Re)Negotiation
  3. Hunt Those Rascally Coons!: Process Analysis, Stockout Analysis, Disruption Analysis, etc.
  4. Regular Barn Dances & Hoedowns: Customer Feedback and Workshops, Supplier Development and Innovation, etc.

KPIs:

  1. Pennies Counted: Spend Under Management
  2. Pennies Saved: Savings and Cost Avoidance
  3. Raucous Reductions: Stockout Reduction, Disruption Reduction, Sourcing Time Reduction, Invoice Processing Time Reduction, Straight Through Invoice Processing, etc.
  4. Supplier Improvement: Defect Reduction, On Time Delivery Improvement, Year-Over-Year Cost Reductions, etc.

Getting to Work

Penny Pinching:

  1. Spend Analysis
  2. Market Intelligence
  3. Category Management
  4. Strategic Sourcing and Procurement
  5. Contract Negotiation Management

No Running Out:

  1. Inventory Planning and Management
  2. Contract Execution
  3. Procurement Automation
  4. Supply Chain Visibility
  5. Risk Monitoring

Not Being Evil:

  1. GHG Carbon Management
  2. Energy and Water Sustainability
  3. Recycling and Waste Minimization
  4. Equality and Equity in your Company and Supply Chain
  5. Human Rights and Worker Wellfare Assurances
  6. No Pollution, including Noise Pollution

Staying Classy:

  1. Quality and Safety Focus
  2. Constant Push to Improve Performance
  3. No-Profit Supplier Development and Customer Workshops
  4. Charitable Donations and Community Building

Simplifying the Procurement One Pager!

Let’s cut this trend off before every consultant AND influencer does their own version, with jargon and KPIs up and down the ying-yang (which you definitely need for system selection, which is something we will discuss in series to come, and partially need to appease management, but need very few of to actually do your job).

Since the purpose of Procurement, as one of the world’s oldest professions (and while a specific instance of sales is claimed to be the world’s oldest profession, which would make Procurement the second oldest, this overlooks the fact that all early societies had astrologers and religious leaders often before they bought and sold, which means it might only be the world’s fourth oldest profession), hasn’t really changed since it started thousands of years ago, and definitely hasn’t changed since the first handbook was written 138 years ago (even though the world has changed), it’s really not that hard to figure out or explain.

To drive this point home, we’re going to capture the core in an explanation so simple that even a hillbilly building a moonshine empire could completely understand it.

Purpose:

Git ‘R Done!

Objectives:

  1. Pinch Those Pennies
  2. Don’t Run Out
  3. Don’t Be Evil!
  4. Stay Classy

Initiatives:

  1. Follow the Money!
  2. Get Haggling!
  3. Go Hunting for those Rascally Coons! (that mess with the systems)
  4. Regular Barn Dances and Hoedowns

KPIs:

  1. Pennies Counted
  2. Pennies Saved
  3. Raucous Reductions
    (because a smooth still is a productive and profitable still)
  4. Supplier Improvement

Getting to Work

Penny Pinching:

  1. Sum the Pennies by Supplier-Product and Supplier-Service Pairing
  2. Eliminate those pairings where there is a contract
  3. Ask Cletus, Earl, Sadie, and Mabel how much they are paying
  4. Eliminate those pairings where you are paying what everyone else is (or less)
  5. Order what’s left from biggest to smallest and get haggling
    1. Get Cletus, Earl, Sadie, and Mabel to share their information at the next trailer park BBQ
    2. Call up the vendors and get those quotes
    3. Set up a few cookouts and meet with the best suppliers
    4. Shake on the best deal

Not Running Out:

  1. Track the inventory utilization by day
  2. Use the curve to plan what you need when
  3. Based on the suppliers lead time, Joe’s delivery time, your order costs, and your supplier’s production costs, figure out the right order size and frequency and plan your orders

Not Being Evil:

  1. Make sure you ain’t wastin’ any water (it’s precious and you need it for your stills)
  2. Make sure you ain’t wastin’ any energy (you need that wood for your barn and that coal for your BBQ)
  3. Make sure everyone is treated the way you treat your trailer park crew
  4. Think about your neighbors (and keep the noise down)

Staying Classy:

  1. Regularly test your product for quality, safety, and performance
  2. Regularly invite your customers to hoedowns to get their feedback and suppliers to barn dances to help them develop their buck dancing, flat footing, clogging, and square-dancing
  3. Give generously to the orphan’s fund and the alligator preserves

It’s stupid simple to get started on good Procurement. Don’t over-complicate it!