Posted by: Adam Deane | 22/08/2011

BPM and Artificial Intelligence

BPM FuturamaIt turns out that there is a possibility that we are all living in a computer simulation.

We could all be a simulation, created by some advanced civilization, in a computer they have built in their universe.

Might our existence not be down to chance? Might there be a “creator” with a grand design?… Could our creator could be a future fourth-grader working on a science project…

Futuristic theories and trends might be sometimes delusional, sometimes inspiring, sometimes nonsense and sometimes just wishful thinking… but always interesting.

One of my futuristic theories on BPM is of embedding AI – Artificial Intelligence

I was reading a blog by Gartner’s Jim Sinur on the topic of Next Generation BPM.

Processes must do better than depend on process participants and managers to catch when critical, important and interesting events are occurring during the execution of process instances across a process. Processes should notify and potentially fix issues automatically

It might not be what Jim meant, but I read into it “adding Artificial Intelligence to BPM”
Processes should be able to fix issues, route themselves, based on learned knowledge

And before the ACM guys jump up, let me clarify – I do not mean enabling users to dynamically change the routing. I mean letting the software decide dynamically to change routing, based on previous process instance behaviour

The people that can make this happen are the Business Rule Management vendors, but they have disappeared, gone underground, been bought out by other vendors, or just hibernating. The simulation vendors are still stuck in the past.
It would be great if they both could come together to provide Artificial Intelligence ability in BPM software.

Wouldn’t it be great if we had the reasoning, planning, learning, perception and the ability to change routing and manipulate processes based on learned knowledge?


Responses

  1. Hi Adam, most of the ‘ACM guys’ refuse that such a ‘machine learning capability’ (which is the right term and not AI) should be part of an ACM system. I do propose it as part of an ‘Adaptive Process’ platform and we call it a ‘User-trained Agent’. I assume that Jim Sinur made this reference to such a capability after our last briefing where I explained to him what we offer.

    It is pattern matching technology that I applied a patent for in 2006 and recently was awarded it. http://www.faqs.org/patents/app/20110178965

    Our technology has since gone through three generations of updates and performance improvements and can recommend user actions for any (external or complex business) event happening within a business process. While there is the ability to actually automatically execute, it is my experience that this is unlikely to happen because it FREAKS people out that the software would do something that they didn’t tell it to do, no matter how transparent that is!

    I do not agree that this can be solved by Business Rule Engines because they can not handle arbitrary data types within a case. A BRE needs a rigidly defined data model and a well-defined data passing and the need to run a COMPLEX RULE SET EACH TIME somehting happens within a business process is prohibitive.

    More in this post: http://isismjpucher.wordpress.com/2010/11/15/can-bpmn-and-rules-handle-complex-events-no/

    My pattern matching technology runs in real-time right within the process engine and recognises any repeating data patterns without any need to predefine data models. It reads the data models from the repository itself based on what it finds in the state-space definition. The reason that the learning and the analysis is real-time is because it doesn’t just analyze the data but it analyzes data relationships. This provides strongly distinctive patterns. It also performs feature disambiguation and relevancy checking because when data are always the same or different they are irrelevant as cause.

    But the user interaction is after all the most important part. First because it provides the training from intuitive user decision making who do not need to explain WHY they did something. Second, because it provides the verification if recommendations are good or not. If users reject a recommendation it is as important for training as if they accept it.

    We have the reasoning, planning, learning, perception and the ability to change routing and manipulate processes based on learned knowledge!
    So AI or better ‘machine learning’ is not the future. It is actually here and has been here for a few years, but the market and the anaylsts simply aren’t ready for advanced solution because they aren’t ‘standard’ or ‘best practice’.

  2. When it comes to AI in BPM I like to take a more macro view. What if processes could learn not just from previous process instances of itself, but other processes within the organization and why not factors outside of the organization that can influence the way a process operates? What if a process engine could monitor twitter or the stock market and change suppliers used in its ordering process in real time based on sentiment analytics of a particular supplier?
    This is where technology build in systems like Watson could potentially make a big splash when it comes to the next frontier of BPM https://www-304.ibm.com/connections/blogs/aim/entry/april_7_2011_4_42_pm10?lang=en_us

  3. @BlueworksLive: could you elaborate how ‘Watson’ would do anything like that? Any why would Twitter or the stockmarket have anything to do with chosing suppliers? The only thing relevant in a supplier is reliability, quality and price and neither Twitter and certainly the stockmarket have nothing to do with this at all! Secondly, what truly mass market commodity is there (save maybe some raw materials) where you could change suppliers more or less automatically? Plus, most suply contracts are long term and some momentary sentiment would be not practical to influence a buying decision. I feel that this is no more than a ‘Smart Planet’ promise with little to no substance …

  4. @Max It was just an example if anything, let alone a promise. I agree with both of you that AI will indeed play a part in BPM in some form or another. I was just trying to expand the “machine learning” from a few process instances to the bigger context in which the process operates (organizational or industry specific or other macro trends). You have demonstrated that AI is already here, so what next?

  5. Either BPM has been in a constant state of suspended animation or I should retag my LinkedIn profile with the ‘BPM Visionary’ moniker…..

    http://theopriestley.com/2010/04/22/the-future-of-bpm-intelligent-business-processes-and-the-self-aware-enterprise/

    http://theopriestley.com/2010/03/22/could-distributed-grid-computing-create-a-brms-cep-and-intelligent-process-powerhouse/

    • Welcome back Theo. We missed you.
      The BPM community has been very quiet since you moved to the Netherlands.

      Looking forward to your provoking insights again.

      Cheers,
      Adam


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Categories

%d bloggers like this: