As far as it concerns adaptability, I surely think AI can make a huge difference here. Think about self learning systems for instance. Also regarding "taking emotion out of the machine", AI will have impact. Where often human emotion drives the design or appliance of systems (say driving a car...), AI could help in a positive way... But there are still challenges... Does my car know that the person waving at me is someone who needs help, has bad intentions? Or just ignore the person, as the object is not in the driving path of my car...
AI and specifically speech recognition will have a huge impact on BPM and on how users interact with processes.Speech IMO will emerge as the dominant user interface for interacting with the web and especially for the IoT. All of the major software/IT organizations now have investments inspeech recognition (Siri, Cortana, Alexa). This is an opportunity for BPM. We’ve discussed here recently about the issues BPMs have always had with the UI, being always just behind the curve. I can see voice becoming a vendor agnostic UI for process. The emergence of chatbots indicates it's already happening.
Integral as data prepared to be used as input or output to/from the AI source. A full audit trail required to ensure reliability of outcomes frankly no different from information created by people? Also essential that when a build of AI capability is undertaken the principles of BPM should be followed to ensure integrity of any AI output.
Yes, and it will come from interoperability between BPM and IoT devices with embedded process control points in BPM run time environments looking for the presence/absence of data in data inflows.
The quality of this data will be important, so pre/post processing will become much more important.
The comment regarding BPM UIs always being "just behind the curve", for me, is the result of a "circle the wagons" mode of operation..
I have had no problems with generic Case management UIs that work across multiple industry areas.
As for plan side UIs for mapping out processes, well, that is another matter.
For the next two to three years, a lot of bleeding edge suffering and undue expense for early adopters. This won't happen to any great degree of efficacy anywhere near as quickly as some would like, never does.
Good AI is based on machine-executable models. Thus a proper use of BPM by creating good executable models and collecting big amount of data enables AI.
Like IoT, AI will be one major disruptor around BPM technology. From a use case perspective, AI can be used to generate intelligence and awareness around an existing process and properly identify the conditions and situations that require starting or notifying a business process. The other scenario is in assisting the process participant with a suggested best outcome (for example based on past executions and similar context). Additionally, AI could help prevent bad decisions, by warning the end user of the impact of taking a particular action at a given moment in time in the process.
IMHO, I think AI can help make process better and sooner than later we will see AI integrated with business processes.
BPM comments by multiple correspondents capture the AI/BPM opportunity well, for example @Eduardo ("awareness") and @Charles ("integration").
Let's ask though if "practical AI exists", other than as a desperate play by executives.
AI is as I'm sure everyone on BPM.com would agree is very, very difficult. Usually what we are talking about is some very sophisticated rules engines (e.g. a neural network), but there is little in the way of "real autonomous intelligence" that could substitute for a human being. For simple tasks of course where human beings aren't being used to full potential -- sure -- but then by definition there probably aren't a lot of savings.
So that leads us to the question: where can we find use cases where there are big savings? Or major new opportunities? And which justify AI as a technology choice? Sure there are use cases, but the cost to teaching the AI system is also enormous. (Tacit knowledge is the elephant in the room.) And thus the business case is likely a challenge (especially with high unemployment).
Gratifyingly, AI is best suited to highly structured environments -- and thus we would expect that BPM should be an area where AI opportunities (e.g. "awareness", "integration") could be found. And paradoxically, the application of AI to BPM has the potential to allow for more flexible, less brittle BPM processes. A little awareness goes a long way.
I had my first real job (20+ years ago, while still a student) in building an AI system for real-time financial time-series forecasting and tick-by-tick investing. Back then AI was a little more than a black box - an interesting academic reasearch field. The project didn't pan out commercially as it could - turns out big institutional investors in Western Europe wouldn't put their money in black boxes :-)
So I was always fascinated by the field and was also quite acutely aware of its limitations. Some of those fundamental limitations have been overcome in the past 5-6 years, but I think a deep impact on enterprise business is still many many years away.
Yes, it's cool to order your Uber from a bot. But I wonder when will occur the purchase of an aircraft fleet from that same bot.
AI is particularly suited in objective function optimization challenges - business is far from being a field where the objective function is even remotely defined. What is the ultimate objective function of a business? How do you express it in a way that motivates an AI to optimize it?
Interestingly, I think AI is far more suited towards solving mining challenges, therefore can be of much more immediate help in case management scenarios, especially when there's a lot of persisted data to be mined in order to come up with next steps. Medicine (as a natural science) is one clear example, and the most comfortable. Law could be the next one (but since this is an artificial science, therefore with far greater variability, conclusive results may gravitate for a long while around local optima).
So, yes, there's a lot of potential for the technology, but as said in previous posts, I take issue with people blowing the horn too loud, too soon, and throw the promise of this technology into mockery.
Even its most rabid promoters seem to agree, which is like, good news. Who knows.
This is a bit of a rushed answer. If I have time I'll revisit and expand/tidy up!
First of all we need to unpack "AI". We did a webinar recently that did some of this quite nicely, I think (see replay below), but very briefly there are a number of relevant pieces that any tech strategist needs to think about in this context, and they operate across two distinct layers: interactions and insights.
AI applications at the interaction layer are largely about creating learning systems that enable automations to fit around the natural expectations/envronments of people - rather than expecting people to modify their behaviours to fit around the needs of automated systems.
At the insights layer a lot of this is to do with advanced pattern detection, recommendation and advisory capabilities.
I see AI applications in the modern workplace as delivering value in two main types of scenario: firstly, in increasing the impact of scarce expert resources in the context of high-expertise tasks (think cancer diagnosis). Secondly, in automating more aspects of procedural high-volume tasks.
How will these things impact BPM?
If we're looking at BPM as a set of technologies to enable work co-ordination at scale then initially, the impact will be at the edges - it will shape new user interaction options around task management and execution, for example. Utimately there's room for AI applications to directly augment process, task selection and assignment.
Sorry this is a bit rambling. Check out the webinar replay.
The questions and many answers show a lack of understanding what technology can do today or might do in the future. Even the coolest self-driving cars are today dangerous pieces of automated plastic unless all human and natural interaction with the road system is removed. It can't distinguish between a rabbit and a baby on the road and can't make a human judgement of such situations. Drones can fly highly autonomously but they are small so errors just end it a little heap of junk. AI isn't happening yet or soon.
As you might know my stance towards BPM is that it is utterly useless in improving what a business does on a human level. It can however automate and dumb down business interactions so they no longer require much intelligence or knowledge ... which has been so expensively analyzed through methodology and practice. Therefore it is odd to then consider that BPM will be improved, amended or replaced by AI, which everyone should really be calling machine learning or ML. To do anything sensible by itself AI would have to emulate human emotions to simulate human intelligence and as that is a chemical experience engine it most likely never will in pure software. Also self-awareness will not happen as we know nothing about the functionality of the medulla -- sitting between the spinal cord and the brain -- which enables that.
But .... ML is a perfect replacement for the orthodox, and rather shortsighted concept of BPM because it will not require any kind of analysis, or monitoring or improvement as all that could happen through machine learning. But learning how and from whom and with what accuracy? We have taken that step to use ML for automated process discovery and Next Best Action recommendations about 5 years ago (the patent is a few years older) and found that no one had any interest in using it, mostly for odd reasons. This included general fear of the technology and possible errors, and aloof rejection of the idea that software could actually do that. Well, it does work in our platform as the famous User-Trained Agent, but those who do BPM do not get it or do not want it ... much like Adaptive Processes. Those who do not like BPM also do not like any ML functionality connected to it and prefer hard-coded applications that after rigorous testing produce a frozen and dead version of business knowledge that becomes instant legacy dead weight to the business.
We do not use AI to emulate human reasoning (which is purely emotional and the reason it works so well) but simply observe human actions and interactions in a well-defined environment of our platform and once the ML software sees repeated patterns of actions and data it will start to recommend these actions, no longer requiring all the BPM mumbo jumbo. But still, there is little interest given the hords of BPM experts who need a job.
A key problem of human interaction -- may it be in written form or speech -- is ambiguity. Humans solve it through context which computers find really hard. Modern speech recognition only works so well as it uses a dictionary and grammar library to turn gibberish into correct words and sentence structures, which we did 15 years ago for OCR recognition. For a business interaction more is required, as much as I agree that speech is the computer interface of the future. A grammatically correct sentence can still not make any business sense at all and we do not gain great benefits if our inputs are single word answers to questions.
Which is why we focus on building ontologies that help to clearly define the terminology of a knowledge domain. I wish we would be able to ML that part but that will still take a little while. But once user input can be made matching to a domain knowledge model, ambiguity in design and Use Case Interactions is reduced and simple text or speech becomes well-working input to an application. ML can learn what inputs it recognizes correctly in a given context of a business architectured capability map and interface the user to the right transactions, guided by user-defined boundary rules and regulative constraints. And yes, that is all part of the patent.
It is not yet simple enough, but that approach will be a practical use of ML for improving the transactional collaboration that every business performs to fulfill its goals and serve its purpose ... also defined through same domain terminology. It does make BPM utterly obsolete. Try to link that with a typical approach to BPM and I truly fail to see why one would bother with the overhead. Unless ... the definition of BPM is simply once more expanded to include not only the heave-ho IoT but also all forms of AI. People surely seem to try ...
So fun to beat around the bush on this subject :-) Some quite outlandish scenarios... I'm surprised the famous tweeting fridge hasn't made a cameo yet.
But frankly, if you take a look at the successful applications of AI today, they are almost exclusively centered around pattern recognition, which for business just translates into faster, more convenient ways of providing human input (text, data, images, videos) to systems, or just faster ways to query otherwise humanly impossible combinations...
I also read that AI will generate intelligence and ML will happen without any kind of analysis, corection, supervision...
Hard to believe that - there'd rather be a lot of humans working for the Machine (injecting work, knowledge, rules, training, coaching, correcting, debugging) before the Machine would eventually work for them.
Right now, it boils down to not having a clear answer on how to keep costs down. As multiple posts have already called out, this is bleeding edge stuff. That's why you see the big players in the industry sinking tons of R&D into machine learning, natural language processing, and voice recognition.
As a developer, I look at ways we can pick up parts of this tech stack to help out the implementors of process. Personally, I'm always interested in better ways to get a customer started on their process. Being able to design an AI solution that can read existing data and discover possible process flows would be a big improvement.
Part of the issue on getting this going is normalizing the data that would be put into such a system. The real customer value would be to have the ability to just dump everything they have into the system and the system would be smart enough to normalize it. This isn't impossible, it's just really expensive to get going.
I think about "Back to the Future" when Doc Brown shows back up at the end and just starts throwing garbage into the Delorion. He didn't care what it was, it was fuel for the time machine. Can't we take that same approach with data and AI? Wouldn't a system that you could throw all your data into and be given a process qualify as a time machine? It took you forward in time to a point where you had the process. 88 mph Marty!
I think AI is a misnomer for this conversation. It leads us down the rathole of discussing this in terms of replacing human endeavor with that of a machine. The point of all this artificial intelligence, or machine learning, or frankly, just automation - is to *assist* and *augment* the human endeavor.
Example: Self Driving cars.
If the endeavor is to drive, self-driving cars are inappropriate. But all the assist technology - automatic braking, even ABS, cruise control, stability / traction controls, etc. are all in the name of augmenting the human driver. All good stuff.
If the endeavor is to transport oneself from point A to point B, self-driving cars make sense. They'll augment that transportation experience, potentially allowing you to read a book like you would on the metro, and yet start and stop at completely custom lcoations.
But often we talk about AI in such a way that there's no person to be transported from A to B. The car will know where to go on its own, based on pattern matching, and will do everything itself. Except, without a human to transport, what is the purpose of the car?
What purpose BPM without the people? What purpose business, without people? Human endeavor is what lends all of these efforts purpose and why we should always be thinking about augmentation more than replacement, in my humble opinion.
It seems like there's a lot of misunderstanding about what we actually mean by AI and ML, so let's define that first in the context of process management.
AI means various things things - including natural language processing - which is really what the focus is.
ML means either numerical (Bayesian) style or adaptive/neural net style learning from past outcomes.
What's completely transformative today which was never the case in the past is that 3 BILLION people already chat apps daily. Yes, read that again.
This means that people are now not just chatting on chat apps, they actually doing WORK on chat apps. This means that that UI is now gone - people expect to do process steps in a chat app they already use.
Here's the key thing - because the current BPM industry is so focused on automation and modelling using flowcharts - this is inaccessible to the average user. Hence, we are trying a simpler approach of a checklist-like UI - it's proving to be wildly successful.
By doing this, the next piece of rationale dictates that people are now EXECUTING and doing steps within our UI and eventually - our chat bot. This solves the biggest missing pieces of BPM today for businesses:
That's where ML comes in. Bayesian style learning can easily record an outcome e.g. a rating or the fact that a process or step was late. Given enough outcomes, it can predict process failure BEFORE it happens - since we are actually recording steps people are doing. The data is gold dust and does not exist today.
Then we have the NLP portion of AI. The aforementioned chat bot will need to understand discrete actions a user takes. Normally these are really simple e.g. just answer yes/no - but the nuances come in when there's more choices. Given that BPM is not about asking open ended questions (like Google search) - BPM and workflow is by far the most easy to merge into a chat bot.
Ultimately, what you can do in a chat bot can also be done with voice - e.g. Amazon Echo. That's just a layer on top of chat.
Hope that adds light to all this darkness. We're actually building the above.
Honestly, I feel that AI is another buzzword. Not in the sense that it's pointless or useless but in the sense that it's just a new word to indicate the same thing. My point is you can read this thread and substitute the word AI for "technology" and it will still work.
Tech is here to make our lives easier, take away the tasks we don't want to do, and let us focus on the growth of our bsuiness.
In this sense AI is not an Oracle like (pun intended) predictor of the future of the industry, With or w/out AI, BPM tech is about automating processes. The future will allow SMBs to do so with the same efficiency that enterprises do it and across devices.