1. Peter Schooff
  2. BPM Discussions
  3. Wednesday, December 28 2016, 09:50 AM
  4.  Subscribe via email
With artificial intelligence one of the key technologies for 2017, where do you see AI having the biggest impact on BPM and processes?
predictive analytics
I believe AI as regards predictive analytics is still largely analogous to recognition technologies - it's only as good as you tell it to be. That means somebody still has to go in and set the parameters, thresholds and bounds for the AI-driven analytics, review and tweak to enable the AI to get smarter as it goes. Those heuristics few still engage in, make the efort; the day(s) of "set it and forget it," deferring to AI on this aren't quite here yet.
  1. Patrick Lujan
  2. 3 weeks ago
Peter Whibley Accepted Answer
Initiation of and interaction with processes using voice recognition. Voice recognition replacing the need for a traditional BPM UI.
Where? How? Give me an application example. I don't see corporate grunts in their cubicles on the floor doing this anytime soon.
  1. Patrick Lujan
  2. 3 weeks ago
Well if you only see BPM in a back office context you'll be under that misapprehension. If you see BPM also in a customer engagement context you'll find plenty of examples e.g. the UK based chatbot lawyer.
  1. Peter Whibley
  2. 3 weeks ago
David Chassels Accepted Answer
Where multiple choices are available then it is now possible to allow the task in the process to work out the optimal choice and present to the user to allow to move to next step. Once widely recognised and understood by business just how this is achieved it will add significant value to the BPM proposition?
Don't rules engines do that now? How would that be different with AI?
  1. Patrick Lujan
  2. 3 weeks ago
Rules operate at the process/case level within Cases.

Predictive analytiics come from the processing of data collected /analyzed across multiple Cases.
  1. karl walter keirstead
  2. 3 weeks ago
Yes and add seamless integration into the process and maybe an algorithim or two real intelligence is readily achieved. Must remember full audit trail of activity required and of course transparency needed just what is happening. Artificial Intelligence is equally applicable to a process as to IoT/gadget but custom build and flexibility in a process could catch on giving BPM a new dimension?
  1. David Chassels
  2. 3 weeks ago
Could'a swore he said "Where multiple choices are available then it is now possible to allow the task in the process to work out the optimal choice and present to the user to allow to move to next step." That's at the case or process instance level, not any aggregate level.
  1. Patrick Lujan
  2. 3 weeks ago
Interesting point yes the instance is the decision making point but can be relying on many past actions to predict the best course of action which could be automated or presented to users to aid their decision making....?
  1. David Chassels
  2. 3 weeks ago
@David. re : "predict(ing) the best course of action'

I recall the following from my days at General Electric - I never used GERT but it would not, today, be all that difficult following data mining to overlay probabilities on the outbound side of decision branching boxes.

See page 309 Construction Project Scheduling and Control, By Saleh A. Mubarak (2010)

From Wikipedia, the free encyclopedia
Graphical Evaluation and Review Technique, commonly known as GERT, is a network analysis technique used in project management that allows probabilistic treatment both network logic and estimation of activity duration. The technique was first described in 1966 by Dr. Alan B. Pritsker of Purdue University and WW Happ.[1][2]

Compared to other techniques, GERT is only rarely used in complex systems. Nevertheless, the GERT approach addresses the majority of the limitations associated with PERT/CPM technique. GERT allows loops between tasks. The fundamental drawback associated with the GERT technique is the complex programme (Monte Carlo simulation) required to model the GERT system. Development in GERT includes Q-GERTS - allowing the user to consider queuing within the system.
Pritsker, A. A. B. (April 1966). "GERT: Graphical Evaluation and Review Technique" (PDF). RM-4973-NASA. National Aeronautics and Space Administration under Contract No. NASr-21. Retrieved 2006-12-05.
Modeling and Analysis Using Q-GERT Networks A. Alan B. Pritsker, 2nd Edition, Wiley, 1979 ISBN 0-470-26648-1
  1. karl walter keirstead
  2. 3 weeks ago
Very easy for real time predictive analysis to get off track.

Consider a flowgraph for healthcare that branches to a small healthcare facility or to a larger facility.

We observe that patients close to the small facility favor this option. All of a sudden, the stats change and these same patients now drive to the larger facility. Decision support will take time to ratchet down, meanwhile the stats will be giving out bad advice.

It could be the larger facility reduced their previously exorbitant parking fees, or, it could that access to the small facility is temporarily encumbered because of construction activity.

Someone has to do the research and adjust the decision support.
  1. karl walter keirstead
  2. 3 weeks ago
Emiel Kelly Accepted Answer
I think it will have the biggest impact in the context of specific processes. Think about applications like recognizing diseases in healthcare. Or when someone shouts at his pc a helpdesk chat is automatically started. Or predicting crimes that did not happen yet

And for years we have already seen applications like "based on what other people with no opinion bought, we we recommend this product" or document classification and data extraction in ECM tools.

For BPM (especially tool things) in general we will probably see, like others said, applying all the learned patterns for predictive analysis or a "next best step recommendation" Also this is only applicable for some kind of processes.

For process analysis we've seen AI for many years in process mining tools already.

All nice, but as said above, I think the biggest value will be in the context of (the execution of) specific processes where AI does, or supports, specific tasks.

Thanks all for 2016 and I predict (actually guess) we will see/read each other again in 2017

Sharing my adventures in Process World via Procesje.nl
Hi, Emiel

We developed a diagnostic algorithm in a product called RapidTox in 1995 for toxicology - input a few symptoms/signs and you get a prioritized list of candidates.

I tried to get permission from Merck to put in all diseases/symptoms/signs in their Merck Manual, they never responded.

"Not feeling well" would light up everything like a Christmas tree, whereas "Mees' lInes" gave a high score to arsenic/thallium etc. poisoning [Mees’ lines typically appear after a person experiences poisoning from arsenic, thallium and other heavy metals. They can also appear in patients who have had chemotherapy or are suffering from renal failure]

Rapid convergence occured when you picked a high ranking poison and then confirmed/ set aside symptoms/signs that were characteristic of that poison.

Very impressive to go into a PCC, ask the lead toxicologist to open Robert Driesbach's Handbook of Poisoning, read you a few symptoms and you then tell them what page they are on.

Re: "Next best step"

Next best step is difficult but "next most popular step" is not.

If you assume that the knowledge workers making decisions at decision boxes have some intuitive feeling that the branching they pick leads to better outcomes you are providing the users with a modest predictive analytics capability by displaying i.e. "most of the time, they went that way"

I don't consider this to be AI.

  1. karl walter keirstead
  2. 3 weeks ago
Jim Sinur Accepted Answer
Blog Writer
I think there will be several phases of AI interacting with process. These phases are not necessarily serial in nature and can overlap.

The first phase will be making the resources smarter like ingesting unstructured and changing data to assist knowledge workers become faster at making decisions in the confines of a process.This would include more NLP based digital assistants or bots. This would include adding smarts to an IoT device or a set of collaborating devices. These processes are more flow directed with pre-planned levers of agility.

The second phase would be where goals where stated and all resources dynamically bid to complete the work of a process, process snippet or a component. These processes dynamically assemble and disburse using an robotic programming automation where processes self assemble, complete work and record each dynamic flow for potential reuse.
  1. http://jimsinur.blogspot.com/2016/08/cognitive-and-process-better-together.html
  2. http://jimsinur.blogspot.com/2016/07/iot-will-rely-heavily-on-cognitive.html
  3. http://jimsinur.blogspot.com/2016/11/the-top-seven-uses-of-cognitive-ai-today.html
FWIW, I concur with Jim's read. I also believe the term AI is too broadly understood or defined for most, so the question Peter asks can't help but generate an overly broad spectrum of answers. AI as an adaptive mechanism - that learns on its own after it is programmed with its initial parameters and learning algorithm - can't help but continue the encroachment on what the knowledge worker does. The pace of progress is quick and in the long-term far reaching - for example, as Nathaniel and I have wondered, how much longer will trucks with drivers be a thing given the state of driver-less technology (a form of AI) or the use of delivery drones (also a form of AI)? We are continuing to quicken the order-to-delivery cycle while continuing to shorten the transaction processing cycle, with consequences we continue to experience. I don't think that 2017 will be much different in that regard than 2016, but one should predict with caution and humility.
  1. Lloyd Dugan
  2. 3 weeks ago
I, for one, welcome our new pre-cog masters.
  1. E Scott Menter
  2. 3 weeks ago
Bogdan Nafornita Accepted Answer
Yes, probably predictive analytics is one of the more immediate applications.

However, I was reading yesterday an academic paper trying to do predictive analytics using deep learning over NLP (associating verbs with tasks and processes with sentences) in order to answer the three key questions of process predictive analytics: "what's the next step?" "when will it end?" "how long is the whole process going to take?"

All along reading, I could not escape the feeling that, for the BPM space, AI is a solution in search of a problem.
Managing Founder, profluo.com
"All along reading, I could not escape the feeling that, for the BPM space, AI is a solution in search of a problem."
  1. Patrick Lujan
  2. 3 weeks ago
E Scott Menter Accepted Answer
Blog Writer
Risk mitigation, already an important BPM use case, is a great AI target. Credit decisions, program trading, disaster recovery, logistics: all areas in which massive numbers of transactions and sudden changes in circumstance have to be continuously analyzed. Risk management may represent the highest-stakes form of the predictive analysis others have referenced above. (If you don't believe me, take a look at the video link below.)

Wishing you all a very happy and healthy new year. Thank you especially to Peter and bpm.com for another year of providing this platform, giving me something new and fun to look forward to each and every week.
  1. https://youtu.be/mbCLmF4a79o
http://www.bplogix.com/images/icon-x-medium.png Scott
John Morris Accepted Answer
Terrific answers. How about an economic perspective too, i. e. "beyond the use case horizon"?

BPM is about automating repetitive work patterns. With better technology, especially automated decisioning technologies, i. e. "AI", ever more complex patterns become affordable.

And while we consider that BPM + AI enable more complex work patterns, consider also that profit margins are typically higher on market segmentation. (Profit margins are also a function of risk (tip of hat to Mr. Menter for his focus on risk)).

So there's a hot intersection of technology and business -- technology reduces the cost of more complex processes; more complex processes are a gating factor for more market segmentation; more segmentation drives higher margins; therefore more consumption of AI and BPM. (Assumes good sales and marketing and a reasonably efficient BPM technology market ... a tall order!)

So, what markets demand increasingly specialized services, are process-oriented, and are suitable for AI? I vote for field service.

Coincidentally here's an article on the AI and field service topic from earlier this week - business process is also covered in the article.

How Artificial Intelligence Will Bring New Advances to the Field Service Industry
BJ Biernatowski Accepted Answer
IMHO the impact of AI on BPM will be relatively quick and if used right it will yield high returns for its adopters.

  • Chatbots can already enhance self-service capabilities in the world that's sprinting towards digital commerce everywhere, all the time. Instead of searching through 1000s gift ideas on Amazon.com, you will be able to ask a chatbot for recommendations identified through a conversation. Thus, AI as the interface to a business process. What about AI augmenting sales reps (would you rather have a conversation with a sharp looking and very energetic human who looks like your sister/brother or with a bot on your cell phone)? The answer may depend on the product you are planning to buy.
  • AI as the source of knowledge and recommendations replacing enterprise SMEs and Knowledge Repositories. For example, instead of having to research how the work in the org is performed today, AI engines may be able to offer recommendations and suggestions to BPMS engines about how the work should be done based on its awareness.
  • AI will most likely start replacing business rules-driven systems. It could be a case of an iPod killing your Sony Walkman, to later emerge as an iPhone. Think about the scale of impact, increased availability of real-time decision-making capabilities and accessibility.
  • Could you teach your AI engine how business is done in your best performing line of business, to later port this insight to your newly started offshore subsidiary?
  • Also how about letting AI decide how to run parts of your business i.e. have most of the decision-making authority? What would you say if your next line of business leader was a software system interfaced by a human offering empathy, emotions, and charisma? Are your LOB leaders today managing the performance of your key business processes?
  • AI may change how BPMS solutions are created. Designing solutions with cognitive abilities in mind will raise stakes for solution architects. AI powered BPM systems will be harder to design at the same time offering much higher returns on investment. The following article from HBR seems to agree with this viewpoint: https://hbr.org/2016/02/companies-are-reimagining-business-processes-with-algorithms
  • AI driven BPMS for some time will be controversial, creating many ethical questions. Some industries will be quicker to jump on this bandwagon than others.
Can't seem to get to the hbr page
  1. karl walter keirstead
  2. 3 weeks ago
I would ask this question slightly different: why BPM and AI must work together?

As we already know from one of the previous topics [ref1], only BPM is able to convert a complex problem [ref2] of “better management of the business” to a complicated problem [ref2] “better management of the business by processes”. So far, this is only the way to implement “Industry 4.0”. But, to realise this ability at the scale of an enterprise or a supply-chain it will be necessary to operate systems which comprise many artefacts - BPM and non-BPM (e.g. microservices, blockchains, risks, etc.). And this can be done only by AI tools and techniques on the base of an explicit and machine-executable model provided by BPM.
So, BPM and AI working together is a mandatory condition for Industry 4.0.

The next question may be – what are the obstacles for BPM and AI working together?

  1. http://bpm.com/bpm-today/in-the-forum/4513-does-implementing-best-practices-assure-you-re-only-as-good-or-bad-as-your-competitors
  2. https://en.wikipedia.org/wiki/Cynefin_framework
Possible Obstacles:
- datasets are too small (AI is fit for large volumes of data)
- models are already explicit (so no value add - AI is fit for discovery of non-linear, complex, implicit models that are impossible to deconstruct deterministically)
- discovered models are too hard to audit
  1. Bogdan Nafornita
  2. 3 weeks ago
- operations datasets are the biggest in an enterprise.
- some models are explicit but not everything - e.g. artifacts life-cycles which act as constrains
- everything that was discovered must be modeled explicitly
  1. Dr Alexander Samarin
  2. 3 weeks ago
- operations datasets are largest only in the largest enterprises. Widespread AI application equates in my mind to all sizes of companies, hence my comment.
- agree
- that is something nobody cracked with AI.
  1. Bogdan Nafornita
  2. 2 weeks ago
Aaron McKeehan Accepted Answer
I definitely think that BPM can really take advantage of AI by implementing it in BPM. A hybrid of both would be incredibly valuable.
Boris Zinchenko Accepted Answer
One crucial and often underestimated impact of AI on all business environments is recent development of autonomous robotics, which moved robots from narrow industrial segment to mass consumer market. With computing power of an average smartphone exceeding capacities of supercomputers just few years ago, this increasing population of artificial brains just awaits engineers to give those wheels, wings, legs and arms for autonomous existence.

We can see these now in every home (take, e.g. robotic vacuum cleaners) and on ever increasing rate in business environments, where AI rapidly comes out of limited areal of desktops and server racks into real world. Autonomous AI devices and IoT become a new and rapidly growing class of business units, which involve into all essential business processes alongside with humans.

With this new quality growth of workforce, BPM becomes essential as never before. With more or less success, you can manage a collective of humans even without any knowledge of BPM. But it is simply impossible to manage in plain words robots, which just do not know any other language, except for workflows, tasks and process chains.

Robots and AI turn BPM into the only way to manage modern business.

  • Page :
  • 1

There are no replies made for this post yet.
However, you are not allowed to reply to this post.