#70: ABAP+Joule, SAP Tinder, Bimodal Nerds
In this issue:
ABAP and Joule: It’s Complicated
We’ve heard a lot about Joule + AI + ABAP from the virtual TechEd stage but not the whole lot was shown. Fear not, Devtoberfest came to the rescue with Boost your Coding Efficiency: Explore Joule’s ABAP Developer Capabilities session giving us a preview of how Joule is going to take ABAPer jobs. (Kidding. It’s not. Not yet, anyways.)
It was a good session (once again, really appreciated informal presentation style) with Karl Kessler, The Godfather of ABAP himself answering questions in chat. Some of my thoughts/observations on this.
The demo example is a custom method to validate email address format and Joule doesn’t ask “hey, buddy, why are you doing this, isn’t there already standard functionality for this?” Maybe Joule is just embarrassed that standard functionality is still a function module? There could be also a class method (or 99 classes) for this but I’d expect Joule to know better, being a designated SAP tool.
At one point you can see that Joule is asked to provide EML code but comes back with SELECT… and then needs to be redirected. This matches my own experience with ChatGPT though, but at least ChatGPT doesn’t ask me for “AI credits”.
Most of the demo time is spent on showing generation of unit tests. This is a divisive topic, but my experience is that the “purse string holders” just don’t want to pay for unit tests. That’s just how it is. So I think this is the most “no brainer” use case but also the least valuable from the customer point of view.
Last but not least, no one should need GenAI to understand what a method with 4 lines of code does. I would love to see though how Joule would react to a more typical example of a “big ball of mud” ABAP report with cryptic (or plain incorrect) naming. JP
But What Does GPT Think?
I love Tobias Hofmann’s SAP development Tech Radar visualization tool for SAP technologies. And I wouldn’t say I love them, but I’ve become accustomed to analyst firms’ visualization techniques like Gartner’s Magic Quadrant and Forrester’s Wave.
SO LET’S MASH THEM ALL TOGETHER WITH AI! IN WAYS THAT DON’T REALLY MAKE MUCH SENSE!
I took 10 items from 3 of the areas that the Tech Radar rates: Technology, Tools, and UI. I created a program that performs a short Elo rating round (judged by GPT 4o-mini), pitting each item against the others on the two Gartner dimensions, “Ability to Execute” and “Completeness of Vision.” I then normalized the Elo scores in the group from .1 to .9, and plotted the results in Magic Quadrant-like plots. Here’s a small GitHub repository with the code I used, and .pngs of the Magic Quadrants I generated.
Yes, I know that Elo rating them isn’t the methodology that Gartner uses. But I wanted to find some way to make AI do the hard work! PM
Swiping Right on SAP Tinder
Weird posts have been making rounds on LinkedIn recently, accusing “some” SAP partners of acting like the used car salesmen. (Naturally, the companies the accusers work for are not at all like that, no siree.)
But sometimes a grain of value can be found even in LI detritus. There are just as many different SAP partners, unaffiliated freelancers or consultancies as there are different SAP customers. What is important is to find a good match for both parties.
Some wisdom can be borrowed here from the dating world. While there is no Tinder for SAP consulting (to my knowledge), just like you’d snoop around a potential mate’s social media profiles, it’s possible to gather intel on your next SAP project partner. Going into the relationship, it also helps to understand whether you’re looking for a commitment or a fling. And what your negotiables and non-negotiables are.
Last year, to prepare for our Tech Connect session, we ran an informal survey of SAP consultants and customers. On the question “What makes the customer relationship go sour”, the consultants overwhelmingly picked “Lack of understanding”. To arrive at that understanding, I think you need to have a good-faith dialogue and be prepared to ask and answer questions.
I’ve been on this planet long enough to survive quite a few bad SAP projects as well as bad dates. And almost every time it comes down to “wow, we really came here not even looking for the same thing”. Our survey might have been anecdotal, but it doesn’t make it wrong. JP
Bimodal Nerd Mode
Wade into an AI discussion on Hacker News. Mosey on over to our channel and watch a few videos. Roughly speaking, you will find two factions: Pitchforks And Torches Against AI Code, and Writing Code Is Freaking Magic Now. They’re strong enough that a bimodal distribution of AI code-writing opinions has formed. It may seem the middle is empty (THE ¯\_(ツ)_/¯ ZONE, above), but I don’t think it’s because there’s no one there. It’s just that those people aren’t really raising their voices.
Why are there such different takes on AI-assisted programming? I can think of a few reasons, but I don’t know how to weight them.
Some of the haters may work in super niche environments that are not well-represented in AI’s training data. I think there’s a kind of middle ground that some languages and frameworks occupy, where ChatGPT or Claude more know the syntax and rules but hallucinate many APIs and utilities that don’t exist.
Some of the Kool-Aid drinkers spend disproportionate time in prototype mode. As a Kool-Aid guy myself, when I’m trying something new, a 70%-good-enough AI answer that I can iterate on is way more valuable than an intimidating blinking cursor.
Software development is an intense kind of cognition, and I think people’s trains of thought for code writing are wildly different from one another. You experience a unique mind-universe, and for many people that universe just doesn’t work well with prompt-and-iterate AI help.
My prediction is that AI Kool-Aid development will become the norm. Not because it’s necessarily better, but soon all code development tools will have AI features baked in. It will be difficult not to drink the Kool-Aid. PM
Old Woman Yells at Interfaces
If you were expecting this story to be about things like Integration Suite or IDocs, then sorry to disappoint: it is about the ABAP interfaces, the IF… and ZIF… stuff. You know how when you just want to look at what the code does and you double-click on the method, but then SAP is like “WhAt cLaSs dO YoU WaNt tO LoOk aT?” How the heck am I supposed to know?! Just show me the code! Yes, THOSE interfaces.
While some developer sheeple just go around mindlessly adding interfaces everywhere, it’s great to see people like John Mikael Lindbakk fearlessly speaking out against the oppression. “Interface hell” indeed.
Sure, some interfaces are the necessary evil (e.g. SAP Gateway development). And the “constant” interfaces get a pass too. But others… Bleh. The interfaces might have a strong fan base in ABAP but I just don’t like those vile things. JP
There’s Yeast In Them Numbers
RISE helps SAP's quarterly numbers look pretty good. As reported by Stuart Lauchlan, CEO Christian Klein said "I'm so, so happy to see that, finally, RISE is the kind of methodology offering that I always aspired it to be." If my paycheck was tied to good SAP numbers, I'd be over the moon.
Reading through quarterly results and CEO comments reminds me that the hands-dirty, elbow grease developer/techie/analyst/practitioner world that I live in is far from the pure, idealized, magical world of huge software license fees. Possibly the clearest piece of this dichotomy is reflected in what ASUG's late July customer survey showed: 49% of survey respondents' S/4HANA migration costs were over budget. And I've heard from a couple places that support costs from partners who provide application management services have started to climb when they're tagged to S/4HANA systems. Don't take that as scientific polling. More like…wrapping gossip in an EXTREMELY PROFESSIONAL newsletter.
As someone in the consulting game, I feel secure that there are at least two more years of intense S/4HANA work to be done (and probably a lot more work to catch up on projects that were delayed for S/4 migrations). But all this teaches me why most consulting firms I've talked to over the years have a not-so-secret desire to exit the consulting game altogether in favor of licensed products. Collecting bucketfuls of recurring license cash is a lot better than turning wrenches on stuff someone else made. PM
Nerdletter Talk is back! Watch our UNHINGED conversation leading up to these stories.
Pssst… There are some free tools available for SAP professionals (and beyond). Björn Schulz is just giving them away! Check out his Show&Tell vid.
Meet Jelena and Paul in person at ASUG Tech Connect conference (also official SAP TechEd “stop” in the US) on November, 12-14 in the sunny West Palm Beach, FL! You will want to attend Paul’s presentation about… wait for it… AI! And then join our Unscripted conversation about the answer to Universe and everything. Add these to your agenda!
What do we run on? Mostly caffeine. Support the nerds by buying us a cup or two of coffee. Thank you for your continuous readership and support!