Bells and Whistles Lead to Cells and Thistles! Part I

I’ve been hearing and seeing two things lately:

  • smaller vendors telling me they’re struggling to close / losing deals because their potential customers are telling them they are leaning towards the bigger/splashier vendor with more “Bells and Whistles”
  • vendors getting lots of (and possibly too much) funding are focussing heavy on bells and whistles (because they believe it will help them sell fast and meet the significant, and often unrealistic, sales targets of the investment firm[s])

So I have to address this because if you select a solution based on bells and whistles, expect to end up in a cell while your fingers bleed from the thistles that prick you every time you try do something that’s not the primary, simple, use case. This then leads to usage of your purchase declining quickly, causing you to end up with another piece of shelfware before you get even halfway through the initial contract subscription.

Now, to help you understand I’m not making this up for clicks and shares, we’re going to dive into some of the most common bells and whistles and why, in the best case, they’re a complete waste of money and, in the worst case, the thorn prick will end up being so painful that your team will simply stop using the software. We’re going to do this across source-to-pay.

Intake

Everyone’s gotta have it because, apparently, it’s brand new and Zip invented the category a few years ago. This is, of course, complete bullsh!t because

  1. Coupa has had full intake since Procurement Independence Day in 2006. You just didn’t know because when Robbie took over the Coupa factory, licensing switched from enterprise size to per user, and the cost was too much to give non-Procurement users a license so you didn’t see the intake that was there. (And to that effect, if Zip is pricing per user, how is this better?)
  2. Zycus had one of the first modules dedicated to intake, iRequest, over a decade ago

Moreover, everyone has to have Gen-AI intake, because, apparently, using a few drop downs to figure out what a user wants is too difficult for the average iZombie who has been zombified by ChatGPT over-dependence.

While this is useful to parse an initial request from a user — do they have a question, want a report, or want to make a requisition (and if so, is it from a catalog or do you have to go to market) — beyond that initial parsing (which will never achieve an initial parsing accuracy beyond 90%, so at least 1 in every 10 requests will have to be rephrased for the system to get it right), it goes from useful to painful to useless.

Here are a few examples:

  • Policy Question: as we’ve said before, Gen-AI has two strengths: large corpus search and summarization and natural language processing; in this circumstance, if the success rate is 90%, it’s mostly useful, but to be honest, if you know the right terminology, elastic search will work just as well, retrieve the exact clause, and never hallucinate a response. Summary: sometimes useful.
  • Catalog Buy: good luck with this because you’ll spend at least 5 minutes trying to order a box of gloves, and I kid you not — see this post where we illustrate how it will take you five minutes to explain to the Gormless AI what you want when you could place an order in a well-defined catalog in 15 seconds with 3 words, 1 number, and 4 clicks. Nothing will drive your users to maverick purchasing off of Amazon, BestBuy, and even Walmart faster than a Gen-AI intake portal. Summary: painful.
  • Analytics Request: while it’s great for a simple “how much am I spending on supplier S” (especially if that total exists in a spend cube), the fact that Gen-AI is very bad at math (to the point that you can ask the same question twice in a row and get a different answer), and not so great at parsing very specific requests such as how much did we spend with supplier S on steel products last quarter which makes it all but useless — even Gartner, the ring leaders of the Gen-AI lovefest, have predicted that “conversational analytics” will completely leave the ProcureTech vernacular in two years. Summary: useless.

… but this is just the beginning. We’ll continue with the Source to Pay process in Part II.