
Comedy Value
To start things off on a light-hearted note the agenda said the event was on Level 1, but when I got in the hotel lift I found there was no level one. The event was actually on level two.
There are three rooms where the presentations occur – Ballrooms 1,2 and 3. So let us say the agenda directs me too ballroom one. There is no number outside of each one. Outside each is the same identical sign saying, “The Grand Ballroom”. That is not even a physical sign, it is a digital sign so presumably you could change each one to say whatever you wanted.
In the agenda be it hard copy or on the phone they have the sessions listed thus
1A – Something | 1B – Something Else | 1C – Something Else Again |
If you are facing the three rooms, you will find that 1C is in the room on your LEFT and 1A is in the room on your RIGHT. As a result, many times during that conference I walked in a room, got puzzled that what was on the screen was not what I was expecting, realised what had happened and then walked back out again.
Would an AI have been to arrange things better is the question?
A map hidden in the corner tells you what the numbers of the ballroom are.
Often at conferences they have vendor stalls, and in return for someone trying to sell you something you get pens, or socks, or hats, or some such. This time there were no vendor stall at all, no sponsors apart from SAP. Though apparently PWC are giving shed loads of cash to SAUG to fund their AI workgroup.
Off we go!
One thing about conferences it is vital that everything starts on time, otherwise all the sessions over-run, and the problem gets worse as the day goes on. It is like going to the Doctor. If you have the first appointment of the day you are not going to have to wait, if you have the tenth appointment you might have to wait ages as the first nine appointments all over-ran. In the “Mastering” conferences they ring a bell five minutes before a session starts meaning “get in here” and start on the dot. This conference did not start bang on time, it was six minutes late (could have been a lot worse). When I am presenting I always start exactly on time, and when I am facilitating I insist my speaker does the same.
Opening Remarks
The most important announcement (and no-one outside of Australia will care much about this) is that this year there will be an SAP TECHED in Melbourne Convention and Exhibition Centre, 18-20 November 2024. There has not been one in Australia since about 2004.
Tim Wilkes – SAUG AI working group.
Tim Wilkes was also the compere for the whole event and a very good one at that. A lot of events do not have a “Master of Ceremonies” as it were and are the worse due to that gap.
Going back to the AI working group, the idea is that every single company in the SAUG does not have to invent the wheel itself, especially given it is often puzzling to know what AI is for in a business sense, and everyone has a lurking suspicion their competition is way ahead of them here.
SAP is fully supportive of this effort and often shares it’s plans. AI services will be supplied via the BTP as might be expected. The long term goal is no more TCODES you just tell something like “Joule“ what it is you are trying to do.
Jonathan Fogarty on AI.
With multiple references to 80’s pop songs, and TV programs like “Westworld“.
AI is a curious cocktail of Hype and FOMO – AI is going to solve every single problem humans have in the entire world, and when it is finished it is going to wipe out humanity.
One SAP example (not of killing everyone, but of solving a problem) is a scheduling tool for field service vans, to get all the vans from A to B in the shortest possible time, taking into account factors like traffic congestion, presumably learning to predict traffic congestion patterns and constantly refining its algorithms. I say presumably because that is me guessing rather than anything that was said.
What was clever often how this conference started was how every one of the first few sessions lead logically onto the next (often you have three of four keynotes on utterly disparate topics) and now in this AI session there were call-outs to some of the sessions later in the day which were deep-dives into bullet points being made in this summary presentation e.g., how data is important.
There was a bit on what skills are needed by humans to use AI properly? Domain knowledge, communication skills and “mindfulness“ of all things.
His final reference was from the HHGTTG and was DON’T PANIC!
The compere said one of his observations was “Only SAP could make AI seem boring“ and that was intended as a compliment.
Mark Pozdena – Powerlink Queensland
This is a state-owned electricity company (Queensland Government) that has a grid which covers four times the area of Germany.
They clearly do not have much of a budget problem, and quite clearly have a very detailed long-term IT strategy, all involving cutting edge technology, mostly from SAP.
Traditionally, just like the UK, Australia generated most of its electricity using coal. In both cases there were islands of hydro-electric (Scotland and Tasmania) because of all the rain.
Anyway, the UK has pretty much stopped using coal for this purpose, and there is a push in Australia to do the same. Mark showed a map with the current 35 renewable energy and storage projects underway in Queensland. The other states in Australia have similar plans.
You cannot do something like that properly unless you accurately report on the progress, so there is government mandated CO2 reporting which needs to be done.
To coin a phrase “There’s an App for That“ – in this case the SAP Sustainability Control Tower (SCT) for ESG reporting (Environmental / Social / Government). Powerlink have chosen to supply the data to this product using another SAP product namely Datasphere. As you can probably tell Powerlink is a bit of a poster child for SAP, using oodles of different SAP products.
I have had the pleasure, if you can call it that, of going to SAP HQ in Walldorf on two or three occasions with the other Mentors for a “Datasphere Day“ so I am familiar with that product. The basic idea is instead of replicating the data from SAP and the other source systems into another database, like you do with BW, the data stays put in the source system and is read and transformed and aggregated and what not on demand whenever a query comes in.
Jumping onto a totally unrelated point, what both the last speaker and this one did, was to have some of the pictures in their presentations generated by an AI to see how they came out. The answer is the results are very strange indeed. Mark said it took him all day to get the AI to create a proper picture of a sandcastle on a beach with the words SAP underneath it. The context was a “sandbox system“. He said in his experience AI system liked to create pictures of people with seven fingers and he could not work out why it keeps doing that.
In the last presentation the AI was asked to create a picture of a CEO having lunch being served by robots. Because AIS are biased (due to all the racism and sexism on the internet which trains them) naturally the CEO was middle-aged, male, and white, looking like Blake Carrington from Dynasty.
Going back to the SCT, as you know, cloud products like S/4HANA, SuccessFactors, Ariba, get an upgrade every so often whether you want it or not, usually every three months.
With the SCT you get such an upgrade every two weeks at this point, with new features, new things you can report on and whatever else SAP can think of. I imagine that high rate either means SAP have mastered the agile “sprint” methodology, or this is not yet a mature product, or both.
Powerlink is moving to RISE with SAP as that is the only way “to get all the cool stuff “. Fair enough, my organisation has also signed up to RISE with SAP for our S/4HANA migration, so I am looking forward to seeing “lots of cool stuff“ in the next few years. I hope that cool stuff is real.
Silly Panel
Silly panel. Silly, silly, panel. Conferences feel they have to do at least one but as my CIO says “When the chairs come on the stage, I walk out the door“.
This one was a bit better than usual due to compere interviewing the panellists in a witty way.
Anyway, a load of the so called „AI/ML“ cases I heard discussed all through the conference seem more like automation to me rather than a machine thinking like a human, working out solutions by itself, or learning from experience and thus constantly improving itself. Maybe a new level of automation, automating things we could not automate before, but automation, nonetheless.
Another random point is that the recurring theme throughout this conference in regard to AI was “2024 is the year of the reality check“. That is, people are trying to work out what AI can do in reality rather than being a magic silver bullet that can solve every single problem humanity has in about ten seconds flat. It is the latter sort of comment – and I hear so many of them all over the place including from SAP marketing – that makes me reach for my bucket.
Liam Mischewski from SAP – Unleash the Power of SAP Business AI
Liam had the pleasure of me introducing him, which I did in song, in the style of Kermit the Frog.
As mentioned, the AI foundation sits on BTP. Joule is the name of the “SAP Co-Pilot“.
AI can “hallucinate“ and thus come up with all sorts of strange answers and make all sorts of strange decisions. This point was being made to warn everyone to watch out for that and try and prevent it – better data for example.
My point of view – nothing to do with any speaker at the event – is that from what I can see AI systems now can sound like a human, right down to making things up, deliberately lying, telling people what they “think“ they want to hear, seeing things that are not there, and making barking at the moon, foaming at the mouth, free energy crazy decisions. And everyone seems to think we are at at a stage we can trust a thing that asks like that to run our business for us.
One researcher I know says that having an AI model be able to understand pictures and sound should help reduce hallucinations as it can cross-check the various inputs – text, vision, sound – for consistency.
The Joule co-pilot will be embedded across all SAP solutions (Ariba, Concur and so on as well as S/4HANA) but part of it can be bought as a standalone. The sickening brand name is “JustAsk“ because (a) someone in marketing got a billion dollars for coming up with that name and (b) you ask it all sorts of questions and it comes back with pretty graphs and charts and tables of data.
Adam Reece – Australian Tax Office (ATO) – E-Invoicing
This one was not about AI but rather B2B. Australia as a country has always seemed behind the EDI/B2B eight ball, compared to, say, European countries. B2B is one of my favourite subjects but it only just seems to be gaining a little bit of traction (in Australia) this very year, 2024.
What this is all about is the PEPPOL PIG network which is active in 45+ countries thus far but is managed by a different agency in each country. In Australia it is managed by the ATO.
Back in 2001 when I was setting up point to point connections between my company and suppliers, we initially used the business connector, and then XI as soon as it was available. As soon as we got XI working, we were told to upgrade to PI and as soon as we got that working, we were told to upgrade to PO and now we have got working we are told we are hopelessly old-fashioned and should upgrade to the SIS. I can pretty much guess what is going to happen as and when we do.
Back in the year 2000 you would configure both systems (sending and receiving) to send IDOCS for purchase order, order response, despatch advice, invoice, payment advice and so on. I was always dealing with two companies which both ran SAP, two SAP systems talking to each other so the process was not that difficult.
Fast forward many years and we have Ariba doing much the same sort of thing, also via XI/PI/PO/SIS, the nature of the messages ae just the same, but the systems do not have to be SAP systems.
PEPPOL PIG is the same thing again, again totally technology agnostic. It is a secure network for sending exactly the same type of messages as Ariba does, via a set of defined APIs. Many different software systems from many different vendors are already enabled to connect via PEPPOL PIG including Ariba and S/4HANA itself (I think).
We are just talking message transfer from one system to another. We are not talking about fancy things like Optical Character Recognition (which has been a big thing for my company for many years) or a fancy workflow in the receiving system to identify and rectify problems with incoming supplier invoices (I wrote such a workflow once, took over a year to get it right).
The ATO is not concerned at the moment about companies sending purchase orders or despatch advice message via P.PIG. They are solely focussing on E-Invoices, the equivalent of the INVOC IDOC.
Now ATO is promoting E-Invoicing in Australia – their mission is to have 100% of invoices in Australia sent electronically - but oddly enough it is not about tax collection. The ATO cannot even see the invoice data being transmitted. The government could have picked any department to run with, you might have thought the department of commerce would be a more logical choice.
The benefits of B2B are so obvious to be they do not even bear repeating here. One point raised was in Australia in 2022 there was $222 million stolen in payment redirection scams and B2B can increase security here (or make it a million times worse if done incorrectly).
If a government wants people to do something they can of course pass a law making it illegal not to do whatever it is. In the case of B2B you would have thought that simple commercial sense would have had everyone using B2B wherever they possibly could, many years ago. But that has not happened.
So, as a first step the Australian Government has made e-invoices compulsory for all government departments. There was a rumour floating around that soon e-invoicing would be compulsory for all commercial businesses as well, but it looks like that is not going to happen – the death blow was the recent change of government. The new government just don‘t care anywhere near as much as the last one did. They have other fish to fry.
However as I understand it, from 2024 (maybe earlier), the law in Australia is that if someone (organisation you are buying from or selling to) asks you to send them an e-invoice via the PEPPOL PIG, or they are going to start sending you such e-invoices, the law says you have to agree to this, which may mean an emergency IT project to enable connectivity, meaning you have to drop or delay some currently planned IT work.
As I said to a finance business analyst today when he outlined all this to me, it is like having the Sword of Damocles hanging over your head (Sha-la-la-la, it ain‘t no crime).
Andrew Dwyer – PanAust (Copper Mining Company) – BTP Build Work Zone
This product has both BTP and the word BUILD in the name, so it must be good. Some years back it would have the word HANA somewhere in the name as well, and in the year 2000 SAP would have put “E” on the front. I will shorten the name to BTPBWZ which sounds a bit like the noise a bee makes.
This is not about AI either, it is about the user interface.
BTPBWZ is the cloud version of the Fiori Launchpad and is very different, he gave a big list of what was worse and what was better. The point being moving from one to the other is not much fun.
This is in essence a website builder where you design screens full of charts and tiles that take you to assorted transactions. And as you know a Fiori tile can be both – can have a KPI written on it, and then you click on the link to investigate and take action to correct that KPI if you do not like the look of it. Just like the Fiori launchpad the idea is to make everything look like a single system to the end user even if you S/4 and Ariba and Success Factors and non-SAP systems under the hood.
In both cases (Fiori Launchpad and BTPBWZ) I cannot help but be reminded of the presentations I used to hear in 2002 or whenever it came out, describing what the “Enterprise Portal“ was all about. It’s exactly the same concept as far as I can see.
Once again a touted benefit of this new thing was a “universal inbox“ where you can approve disparate things from assorted systems all in one place. In the last 20 years I have probably heard the universal inbox presented as a radical new idea which will change the world, about fifty billion times. Is there more than one new “universal inbox“ invented every year? Quite probably. Maybe not quite as frequently as the ABAP to Excel blogs, but close.
The BTPBWZ was designed for S/4 but works for ECC 6.0 with some limitations.
Whenever people start talking about security my eyes glaze over, I know it is important, but I am glad I have never had to look after it for a living. I did not understand 95% of the points Andrew was making about security in the BTPBWZ but there was a lot, it looked horrifically complicated to my layman’s eyes.
Felicity Rudd – Powerlink – Digital Learning
This was not about AI. This was a success story about an SAP tool to enable digital learning (training).
They say that for every dollar you spend on training you save two dollars (obviously there must be a break-even point, but since most companies spend bog all on training it is difficult to work out where that point is).
The tool is called “Enable Now“ and it is used to manage (create/change/display) learning content. Inside it you have something called the “SAP Companion“ who was a friend of Zefram Cochrane in 1967 and now is an integrated contextual help.
Ever since the year 2000 at my company we have used that “repair“ where you can add an extra option to the drop down “help“ option so from any transaction you can call up whatever training material you have based on the current transaction code. Over the years what gets called up has changed again and again, sometimes documents, sometimes calls to third party training software.
This is the same except it is a standard SAP product, and the end user can choose between a static document or a transaction simulation. I am sure we bought something similar once where you did a transaction recording where a little wizard walked around the screen pointing his wand at whatever field you wanted to highlight.
Anyway, as mentioned, this was a big hit at Powerlink, people liked being able to call up detailed training material whenever they felt like it, or when they were puzzled by something, not that this would ever happen when executing a standard SAP transaction or when you get one of the “self-explanatory” error messages.
John Martinez – Energy Queensland – Procurement
This is not about AI. Energy Queensland is another state-owned power company in Queensland. I wondered why there were two, they probably do subtly different things.
In the Agenda the description of this session only said it was about how the company had managed to transform procurement using SAP. In that abstract the word “Fieldglass“ was not even mentioned once, but that is what the presentation was all about.
This was another success story. I did not really know much about Fieldglass – I knew it was sort of half HR and half procurement as it is all about hiring and paying contractors.
Anyway, this power company hires a vast amount of contractors every year and the presenter could not praise Fieldglass enough in regard to how much time it had saved him, making some processes ten times faster.
In essence he said the whole thing was incredibly user friendly and the integration into S/4HANA worked very well indeed e.g., creating purchase orders and whatever.
SAP must love the Australian Government, they get so much money from them, which is why I suppose there is a dedicated Public Service SAP conference in Canberra every year.
AI – Lecturers from the University of Queensland Technology
This was in two halves. The first was the research they had done on what assorted companies had been doing with AI recently. The second was more philosophical in regard to what affect AI might have on humans when and if it gets mainstream.
To start with the first strand – the researchers did a huge amount of research into does AI actually deliver value (as opposed to it just being a load of useless hype)? They showed 52 dots on the screen, each one was a different AI project done by a different organisation. Was the conclusion that such things deliver value or do not deliver value? No answer was the stern reply. That reminds me of when I was at university, it seemed that doing the research was ten billion times more important than drawing any sort of conclusion from that research. That is like when you take a math exam. You can get all the answers wrong but still get a really high score if you show all your workings in detail.
I get the feeling most organisations wanted anonymity here. One notable exception was Microsoft. In the post COVID era that company had had a lot of offices half empty so they created an AI model to predict what people would turn up in person to what office when and adjust the heating accordingly (I think that was what it was, sounds odd to me, but there you go). Amway, that worked well, and they were able to re-use the model to predict what people would use what parking spaces, and eventually were able to sell this to other companies - which is the acid test, when people are willing to pay money for something. That means whatever it people are willing to pay for is either a valuable useful thing, or everyone is BONKERS, one or the other must be true.
One of the running AI analogies was to imagine you were in a theme park and there were several queues for some exciting ride. You find all the other queues are moving faster than your one and then you think everyone else is going to get there before I do. And of course the instant you move into another queue the one you were originally in starts to move ten times faster.
As a concrete example (this is a pun), in many industries there are three or four big companies dominating the industry and it is illegal to co-operate (so called “cartel“ behaviour) behind closed doors but if one of those companies wants to shout out publicly what it is doing (on the internet or at a public conference) then it is not illegal for the competitors to listen, because anyone can listen.
So, when the researchers said that one of their case studies was Mexican building material supplier CEMEX and that company was doing all sorts of AI things, I thought that is a competitor of my company, I wonder what they are doing, and I noticed people from two other competitor companies in the room as well. Naturally we wanted to see if the competition was a hundred steps ahead of us in this area, and if so, what they were using AI for.
Sadly (for us competitors) CEMEX must have told the presenters not to provide any details as to exactly what they were doing. All we were told was that CEMEX has 5000 AI data models and they have scaled them up (whatever that means) and have scaled then out (whatever that means) and now have a “data democracy” (whatever that means).
The next point is that a lot of effort has to go in to “bias remediation”. That is, as I alluded to earlier, if you train an AI model on a vast amount of data from the internet, where much (possibly most) of that “information” is written by obvious lunatics, and then afterwards ask that model to make business critical decisions, who well do you think that is going to work?
The second half of the presentation has about “skills erosion” – that is, the more you rely on an automated decision-making system, the less you think for yourself. Eventually you turn into a vegetable, which is probably what AI systems want, as that makes it easier for them to take over the world.
There are thousands of examples of this. The compere said he had no idea what the phone numbers of his five children were. And why would you? You just press a button on your phone. A physical analogy was before word processing programs came along, I used to write hand-written notes all day long at university with no problem. Now, after thirty years of typing everything, if I actually start making handwritten notes after a while my hand starts hurting. The term would be “use it or lose it”. I think the only reason I can still do mental arithmetic is because I play Scrabble every evening. At work I just put numbers into Excel if I want to add them up.
The final thing I remember was the compere noting that nowadays you get into trouble if you are caught looking at the map on your phone whilst driving. But many years ago, people used to drive with maps on their lap, and kept looking at them whilst driving. That was just as dangerous obviously, but it proves some things never change.
One thing that has changed is the “smartphone zombie” thing. In some parts of Melbourne now they have the “walk / do not walk” signs at pedestrian crossings on the floor, so people can see the lights out of the corner of their eye whilst looking at their phone. In the entire State of Victoria they are replacing every single level crossing – either a bridge above, or a walkway below – because in the past when walking up to a level crossing people would look left and right to see if a train was coming, but now they just look at their phones and walk straight out in front of the speeding train, and they cannot hear the train coming, as they have those huge headphones on. At this rate there will not be any humans left for AI to replace.
Conclusion
I had been very worried that this conference was just going to be a load of marketing nonsense, given its relentless focus on AI, but in fact I was very pleasantly surprised. There are two more SAUG conferences in Australia this year, in Sydney in August and the TECHED thing in November, and I hope to speak at both. But it will not be about AI – probably!
Cheersy Cheers
Paul
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
User | Count |
---|---|
10 | |
8 | |
5 | |
4 | |
3 | |
3 | |
3 | |
3 | |
2 | |
2 |