Skip to Content

Before BI there was AI: will Enterprise SOA bring it back?

Before BI there was AI. Not just alphabetically. In the history of IT also. And the thing about AI that was so interesting is that its underlying theory contained no preconceptions about how much data might or might not be required to make any given intelligent decision or take any given intelligent action. The classic example comes from the Campbell Soup Company’s experience with AI. As the story goes, the company had an old guy on its staff whose job was to maintain the huge vats that cooked the soup. If one of these vats failed, a lot of product was ruined, so his job was really important. So when this guy was about to retire, the company realized they had to do something to make sure his job kept on getting done as well as he had done it. And it turns out that in order to emulate his performance, an AI system would have to monitor very few cues from the vats themselves – how they sounded, how they smelled, etc. I personally suspect that this is true of business in general, and in fact, will give this suspicion a name here and now: Halitsky’s Hypothesis 99% or more of all important business decisions or actions in a company are actually made or taken based on 1% or less of the data that the company is actually paying SAP or Oracle to store, update, and make available. And if this hypothesis is correct, then a reasonable model for one kind of BPX (Business Process Expertise) would be as follows: a) the back-end continually crunches all 100% of the company’s data to extract the 1% or less that really matters for decision-making and action-taking; b) the Enterprise SOA component communicates this 1% to the folks who need it to make decisions and take actions. When Enterprise SOA is used in this way, its known performance limitations are irrelevant, since it’s actually handling very little data. Well, something to think about anyway …

You must be Logged on to comment or reply to a post.
    • Yes, I agree that data mining and BI and analytics share aspects of AI.

      But an important difference remains.   AI’s goal was simply to model the success of folks who were successful at what they did.   It came with no preconceptions about why folks are successful at what they do – there was no “built-in” model concerning the way rational businesses should behave or do behave.

      Suppose there’s a guy who consistently predicts desirable stock-levels with uncanny accuracy, but he does so using a very peculiar combination of very peculiar signals from the data – a combination of signals that doesn’t really “jibe” with received wisdom about how businesses should predict desirable stock-levels.

      I suspect that:

      1) the old AI practitioners would say – OK – if that’s what he does, then that’s what we have to do: listen for that really weird combination of signals from the data. 

      2) the new practitioners of BPX, on the other hand, would probably say, “Nah – his success is just a fluke.  What we’re going to do is give folks really nice-looking pie-charts, bar-charts, and scatter-graphs centered on all the standard traditional indicators.”

      Of course I’m playing a bit of a “devil’s advocate” here, but I think there’s an element of truth in what I’m saying.

      • ->AI’s goal was simply to model the success of folks who were successful at what they did. It came with no preconceptions about why folks are successful at what they do.

        As per my understanding Data Mining also serves the same purpose, it also not looks for why folks are successful. Could you elobrate it more?

        Though I agree with you, as AI is very broader term, and Data Mining/SAP Analytics just a small part of it. But it is also not in much demand by SAP Customers as of now, and I am sure SAP must be having some future plans for it.


        • When complex weapon systems are “provisioned”, decisions are made as to which parts can be removed and/or repaired and/or replaced where. (The “where” is usually a small number of alternative locations: “front-line” maintenance crew, rear maintenance center, regional depot, or vendor.  Or something like that.)

          But anyone who has spent even a moment talking to military “loggies” knows the nightmares that continually occur because of incorrect decisions about what can be done to what where.

          For example, someone needs to replace a really cheap part and he is authorized to repair it at his loctaion in a “front” maintenance unit.

          The problem is – he can’t physically get to the part he wants to replace because his access is blocked by another part that is coded as a part that he is not allowed to remove at his location.

          So the whole piece of equipment gets shipped upstream to the “rear” maintenance unit where the “blocking” piece can be removed.  And of course, there is a dollar cost attached to this, as well as the cost of the extra down time.

          And therefore, one would think that a really hot Enterprise SOA system would be one where the low-level repairman at the front-unit could send a request to the system for a reclassification of the parts so that the situation doesn’t occur in the future.  And the system would send back a notification indicating that the recoding had occurred and it was OK to proceed.

          A simple work-flow, no?

          The problem is: “blocking relationships” among parts  in 3-dimensions are very difficult to detect – the academics are still trying to figure out whether it’s even possible.

          So “data mining” and analytics won’t do the trick – what’s needed is an old-fashioned AI approach which says that if I trust a certain mechanic and he says that I’m doing something stupid which is hurting “the business”, then I’m going to listen to him and fix the problem.

          And of course, I’m going to use an Enterprise SOA system to make sure that the problem gets fixed as soon as he reports it.

          • an automated system could also analyze all “upstream” movements of equipment to determine what was done and why, and thereby isolate problems “ex post facto”.  So in that sense, the AI and BI approaches are not entirely incompatible, as we agree.
  • David, gotta say how totally off mark you are on this. To compare AI in the realm of Enterprise SOA?

    Enterprise SOA provides the means for a real human to interpret data be 1% of the whole or 10% of the whole (which I think is more true) whereas the purpose of AI was to eliminate the human factor and have a self running system?

    You programmed an AI with various parameters and conditions, the system then responded with actions based on those parameters and conditions but more importantly learned to adapt those parameters and conditions based on past experience. They are evolving systems what you lay out here is not even close.

    You can’t simply make a blanket statement about will Enterprise SOA bring back AI – AI first off is still here but they are not connected in the sense you are detailing here.

    As a blogger I would expect you to take in consideration that the topics you choose and how you present them can add either confusion or clarity and in this case “clarity” is no where to be seen… rough!

    • So let me try to restate what I was trying to say and why I wanted to say it.

      For some time there have been whispers about whether or not the protocols underlying Enterprise SOA are ready for prime time or not – from the point of view of performance when high-volume high-thruput is required.

      Recently these whispers surfaced here at SDN in Dan’s recent and most impressive post about some comaparative benchmarks he ran involving Enterprise SOA vs other protocls.

      And Dan’s post got me thinking along the following lines.

      Suppose it is true that the protocols underlying Enterprise SOA really aren’t quite “ready for prime time” and that folks are going have to wait for their optimization before trying to use them for data-intensive applications.

      Well if that’s the case, then the question arises as to what can be done with Enterprise SOA in the meantime – while we’re waiting for it to become optimized. 

      And one reasonable answer to this question would seem, at least to me, to try and use Enterprise SOA in mission-critical but not data-intensive applications.

      And this is where AI occurred to me as one possible back-end foundation for systems that use Enterprise SOA in non-data-intensive ways.

      But before continuing, let me say at the outset that you and I are thinking about two different kinds of AI. You’re thinking about AI in the very specific and more modern sense of a “neural network/fuzzy logic” type system that can be programmed to “learn” to become a non-human SME.

      I was thinking about AI in the older and more generic sense as any system which is capable of emulating the human capability to detect important signals in a set of data.

      So I was not trying at all to compare AI with Enterprise SOA in the sense of saying that they do similar things.  

      Rather, I was suggesting that when folks look around for good ways to use Enterprise SOA, they should consider it as a way of providing people with information that is extracted from data by AI systems.  And the reason I suggested this is because AI systems (in my sense, not yours), tend not yield data-intensive outputs (although their inputs are often data-intensive.)

      The other reason I suggested AI systems as a foundation for Enterprise SOA systems is because I personally find a very distressing tendency in a lot of the commentary about Enterprise SOA here at SDN and elsewhere (articles, books, etc.)

      This tendency manifests itself in the apparent belief of some BPXers that the only “correct” way to use Enterprise SOA is to restructure a company’s business processes.

      And while this may be one possible use of Enterprise SOA, or even the best use of Enterprise SOA, it is certainly the most dangerous way to use Enterprise SOA until it is certified as “ready for prime time” from a performance perspective.

      But if BPXers were to adopt the mindset of the original business AI folks (not the “neural network/fuzzy logic” folks that you’re thinking of), then they would see that there a lot of ways to bring Enterprise SOA into business operations in meaningful ways that don’t have anything to do with restructuring existing processes.

      Rather, Enterprise SOA can play a key role in creating new business processes that result from the judicious re-application of old-fashioned AI approaches to business processes.

      Hope this clarifies the intent of my original post, as well my reasons for posting it.

      • David…I am hoping you are/were playing Devil’s Advocate here because I have to totally disagree. I’m with Craig on this one. I was really confused by your blog and did not see the connection you were making with AI (your variant or the “fuzzy logic” type) and SOA. Now, even after reading your response and clarification to Craig, I am even more at a loss to understand the original intent of your blog. I just don’t see the SOA and AI points you are trying to make. A little more help? I guess one starting point for me would be the clarification of this…to simplify it for me, are you trying to find a pathway via SOA to make use of collected data to make automated intelligent forecasted/future decisions (via AI) or using SOA (specific intelligent services, actually) to take collected data and past experience to make specific real-time intelligent decisions (such as given the military “parts” example you gave…that when a system sees the request for ‘part A’ maintenance, it could immediately notify or move the order to first address the ‘blocking’ part). Anyways….if you feel like explaining your blog more, I’d love to hear what you were thinking.
        • Right there with you Christopher – David either be very clear with what you write or don’t write all you end up doing is confusing the whole issue and the topic of Enterprise SOA at it’s current state in the Enterprise world and that of SaaS is already confusing enough for people – take responsibility for your blogs here!

          Your clarifications to me end up making less sense in terms of your blog and confuse the topic more than you originally did.

  • So let’s give it a try and see what happens.

    Question (1):

    Are there customers who have gone or will shortly go to ECC5 NW2004s for various reasons that do NOT include their desire to use VC/CAF, NWDS/WDJ, X(P)I, or WDA within Enterprise SOA solutins in the near term?

    Answer (1):

    I assume you both would have to answer “yes” here.  There are lots of reasons why customers might upgrade with no near-term plans to implement any Enterprise SOA solutions involving VC/CAF, NWDS/WDJ,  X(PI), or WDA.

    Question (2):

    Then of the customers in the above category (call them Category-1-Customers or C-1-C’s), are there more than a few whose reasons for not plunging into Enterprise SOA include:

    a) the fact that they’ve heard whispers about the performance of SOA protocols in data-intensive applications (high-volume /high-throughput)

    b) the fact that on the one hand,  they have only heard Enterprise SOA pitched as a means to make revolutionary restructurings in their business processes, while on the other hand,  they’ve got a lot of senior line and staff managers who are tired of restructuring their business processes for the second or third time simply because there’s a new gospel being preached in the IT coummuinity.

    Answer (2):

    I again assume you’d both have to say “yeah – there are probably some C-1-C’s who feel the way that’s described in (2a-b) above”.

    Question 3:

    Will some C-1-C’s who have feelings like those in (2a-b) wind up going to Enterprise SOA in two to three years because:

    a) the techie have successfully optimized the underlying protocols, or found a way to
    parallelize them, so there are no more performance worries in data-intensive applications.

    b) some of their recalcitrant senior managers will have retired by then, or at least have been promoted to positions where their opinions no longer matter.

    Answer (3):

    Again, I assume you’d have to say “yes” – there might well be some C-1-C’s who WILL go to Enterprise SOA in two to three years.

    Question (4):

    So suppose you’re a loyal member of the SAP community and a true believer in SAP’s ability to enable Enterprise SOA solutions for its customers, but you happen to work for a C-1-C customer as described above in 1-3.  What can you do during the intervening two to three years to:

    a) get and keep your technical “chops” so you can hit the ground running when your company finally agrees to “move out” on Enterprise SOA?

    b) help convince the company that they really can shorten the two-three year period and put some worthwhile Enterprise SOA solutions into production in the meantime?


    I have to assume that you’d agree with the  idea that one way to achieve (4a-b) is to
    think up a significant real-life business improvement that is:

    c) not data-intensive (low-volume, low throughput);

    d) wouldn’t involve signficiant restructuring of any core business processes.


    OK – so if you’re with me so far, here’s where AI comes in, and here’s where its relationship with Enterprise SOA is so potentially promising.

    But this response is getting pretty long for just a response, so I’m going to hold-off.

    If both of you respond back here and indicate you agree pretty much with my answers to question 1-4, then I promise that I’ll post a blog post describing an AI/ESOA military logistics project that will knock your socks off .  It’s actually a simplified version of the scenario I laid out in the original post here but just as real-to-life.  It has to do with what happens when a repairman is told that he can’t order a replacement for a given part, because he has to order the NHA (next-higher-assembly) …. which just so happens to be a $25 million-dollar airframe.  I swear to you this is a true example from real life and was related to me by a USAF loggie I worked with back in 1996-98 at the US Army Missile Command in Huntsville A (now AMCOM – Aviation and Missiles

    Best regards

    • You assume too much and are still off mark and it’s Enterprise SOA not EOSA!

      Perhaps it’s better you start completely over here David and be careful of the terms you’ve gone through some major revisions to your original statement so far but not yet addressed any of our beginning points…

      Sorry no desire anymore – perhaps your next blog will be clearer and to the point and not just use lots of “big terms” to sound like it’s important – this one was a major let down.

      • “You can lead a horse to water but you can’t make him drink”.

        There’s only one word in there with more than one syllable, unless you count “can’t” as “can not”.

        Seriously, I’m glad to see that SDN is a moderated site where moderators feel free to engage in personal insults.

        I hate those “wussy” web-sites where moderators feel they have an obligation to set an example for others!!

        Best regards

        • David, since you love quotes and such so much…

          “We’re all sensitive people with so much to give” (Marvin Gaye)

          Being a little sensitive, DJH?

          Are you implying that by “signing up” to be a moderator, Craig is no longer allowed to be “just” a participant with his own opinions and ideas outside wearing his “moderator hat” 24/7? In any event, I have no problem with the example Craig has set forth as a moderator….”if you disagree with a blog/post/others comments/etc, you may feel free to post your opinion also”. I think Craig might be on to some kinda crazy idea there!

          • Hi Chris –

            Au contraire, I was completely serious in my response to Craig.

            In his statement here:

            “perhaps your next blog will be clearer and to the point and not just use lots of “big terms” to sound like it’s important”

            he pretty explicitly claimed that I adopt a certain writing style to sound important.

            So at most sites, such a statement would be taken as impugning a poster’s character or motivation, and as such, might well come in for a rebuke from a moderator.

            But that’s what I love about SDN – it’s such an open site that not only can such statements be freely made without rebuke – they can be made by moderators themselves.

            I’m not sure why you think I was being sensitive or sarcastic when I said this the first time. 

            Sites that get too strict about personal insults tend to get ossified very quickly.  Sites that let posters “go with the flow” and “roll with the punches” tend to stay lively and interesting.

            Anyway, as long as you’ve decided to keep this conversation going, I feel compelled to point out that you haven’t as yet made any substantive response to my attempt at a clarification of my original post – the clarification that you yourself asked for. 

            Of the questions that I posed to you, is there one or more that you would answer differently than the answer I gave?

            If so, which question(s)?

            If not, then how can you disagree with my conclusions?

            Oh wait a second – maybe I’m beginning to see something here.  Neither you nor Craig could find any reason to disagree with my conclusions, so you had to go off topic.

            One final note – since you like trading quotes, here’s one that always strikes me as relevant when people like Craig (and Ajay in a different post) start throwing around insults about someone using “big terms”.

            At this link Here , you will find this passage:

            “Hermann Göring often remarked, “Whenever I hear the word ‘culture’, I reach for my Browning,” a misquote from Hanns Johst’s play Schlageter(1933): “Wenn ich Kultur höre … entsichere ich meinen Browning” – “Whenever I hear the word culture… I release the safety-catch of my Browning!” (Act 1, Scene 1)

            That quote is one of my personal favorites and I hope you like it too. 

            It’s such a cool thing to say when you feel like being anti-intellectual, don’t you think?

          • I chose not to continue to respond because you’ve clearly labeled me as a moderator and thus excluded me from being a “neutral” party and commenting on something I feel is “off”.

            I have plenty of reasons to disagree but you made it impossible for me to do so – one of the drawbacks of being the blog admin/editor/moderator…

            Have a nice day

          • I agree with Craig…Not because he’s a moderator he’s banned for posting comments or disagree with your blogs.

            I know that my English it’s not the best…But I can see where he has insult you…You can’t take criticism as an insult…I know Craig from a long time, and maybe sometimes he can looks like angry…But he just wants the best for the SDN, and if your blogs confuses people…The he should do something about it…

            P.S: I hope you don’t take this as an attack -:) I just love controversies and discussions -:D



          • Hi Alvaro –

            I think you’re right in everything you say.

            First – I agree with Craig (and you) that the original post might have been confusing for people who don’t want to take the time to “read between the lines”.  That’s why I took the time and trouble to do a detailed response.  If I hadn’t taken Craig’s criticism seriously, I wouldn’t have bothered doing this response. 

            Second – I agree that Craig should feel free to say whatever he wants to anyone at SDN, including personal insults.  I don’t know why you and Craig and Chris aren’t taking me seriously when I say this – I’ve said it twice now and both times, it seems that people think I’m being sarcastic when I’m not. 

            Third – here’s the difference between criticism and personal insult – and it’s an important one.

            Suppose someone says to me:

            “I think you should try to use less big words in your posts because: a) there are a lot of SDNers whose native language is not English; b) of the SDNers whose native language IS English, a lot of them are techies with undeveloped vocabularies.”

            Then that statement would be a “criticism”, not an insult.  It’s not a “criticism” I would agree with, but I would recognize it as a valid “criticism”.

            On the other hand, suppose someone says to me:

            “I think you use big words in your post to sound important.”

            then that’s a personal insult.

            Do you see the difference? 

            The first statement says nothing about why I use big words – it just says that maybe I should use a more restricted vocabulary here at SDN. 

            The second statement says something unflattering about WHY I use big words that doesn’t happen to be true.

            I hope you do see the difference, and even more importantly, I hope you believe me THIS time when I say that Craig should feel free to say whatever he wants here – criticisms and/or personal insults.  It’s part of what makes SDN a fun site to blog and post at.

            So if Craig and/or Chris decide to drop out of the conversation, it’s totally up to them.  I’m certainly not asking them to do so.

          • First, I am quite sure you were being “completely serious in my response to Craig”. You seem to be a rather serious fellow most of the time. I wasn’t trying to say you weren’t taking his responses seriously….I just kind of felt you were being a bit sensitive to criticism (and again, I was trying to lighten the mood here). Maybe it’s just my differing experience…I happen to participate/visit/lurk on quite a few other forums (and most of those non-technical /non-professional) that are far less formal. In fact, the moderators on some will even allow anything up to the point where one person will threaten to show up on the others doorstep and duke it out in the street (haha)….so little pokes back and forth between members are nothing that gets my ire up.

            So anyways….not to derail the discussion of this blog further, you point out that “you(me) haven’t as yet made any substantive response to my attempt at a clarification of my original post – the clarification that you yourself asked for”. And further, you wrote “Oh wait a second – maybe I’m beginning to see something here. Neither you nor Craig could find any reason to disagree with my conclusions, so you had to go off topic.” Well, first off, I was actually waiting ,as Craig asked for, “Perhaps it’s better you start completely over here David and be careful of the terms you’ve gone through some major revisions to your original statement so far but not yet addressed any of our beginning points”. I was hoping that you were going to revise the original blog….not just add to the comments so that I feel like I have to bounce around to get the jest of your idea. Sorry….that was my misunderstanding of your follow up. Secondly, don’t take my silence as meaning that I can not find any reason to disagree. I actually still disagree. As I said, I was waiting for the blog to be revised before responding (because by then, I might actually agree with your ideas and don’t want to jump the gun). However, now, honestly….I just don’t care one way or the other.

          • So let me get this straight, Chris.

            You ask for a clarification, don’t like the fact that I gave you the clarification in a response rather than in a revision of my original post, and as as result, you tell me that you “just don’t care one way or the other” any more.

            Isn’t that like asking a girl to take you back to her apartment and then when you get to her building, you tell her you’re no longer interested because she walked down 5th Ave and across 12th Street to 6th Ave, whereas you would have preferred walking across 13th Street to 6th Avenue and then down to 12th Street?

            Perhaps the sight-seeing might be better walking your way rather than hers, but was sight-seeing really the point of the walk in the first place?


          • Clarification should be done on the blog, because not all people is going to read thousand of comments just to understand what your trying to say here…

            I have read all the comments…as I usually do -:) But…If people is lazy enough to not post comments…They’re surely not going to read them -;)



  • Interesting blog, but very difficult for development and deployment.

    This requires integration of Sensory, Robotics, Pattern Recognition, areas of AI, plus a central rule based engine that can be used for mass design and deployment.

    Even though, I totally agree with you this is where the real $$ can be found.

    • But it’s precisely the difficulty or impossibility of obtaining correct solutions “a priori” (beforehand) that creates the possibility for Enterprise SOA to participate in “a posteriori” (after the fact) solutions that will at least allow corrections to be applied via efficient workflows after problems have been encountered and reported.

      I will be blogging on this same topic in the near future: if you’re interested, look for a blog titled “Mass-Market versus Niche” Enterprise SOA”.

      Thanks very much again for taking the time to respond.


  • I agree on the value of AI to BI, but disagree on centralization.  I’ve been developing “smart agents” which individually specialize in “sniffing” but confederate in data sharing in order to create useful information.

    Since 1984,I’ve seen the BI/AI platforms merging, and we’re finally getting the infrastructure to do so in a practical way.

    My opinion is that we’re still “bleeding edge” though.

    • … is very very interesting – something I hadn’t heard of before.

      And actually, if you think about it, a natural fit with Enterprise SOA with respect to loosely-coupled independent interaction.

      Can you refer us to any link(s) at which we might learn more about what you and others are doing in this area?  There’s something that strikes me as intutively “right” about the notion … maybe even paralleling what the gray matter itself is doing.  (I was never a believer in localization – I think it’s a convenient crock … )

      Best regards