Do you really want this guy running around performing process instances ?

Ok. Agents. Love the idea, however it's just sub-task automation by another name; calling an instance of a defined process that's been triggered by a specific event. There's still an event trigger, a delegation, a request for an action to be performed and completed, a resultant action that closes the task/ process.

If it's external to the BPMS itself (as the definition suggests, it's a software package) then where's the difference in the existing stack we have today of a BPMS calling another system via API to complete some other task ?

If you're really looking at next-gen then the 'agent' is listening all the time, it's not waiting for the delegation of a task to take place. That's perhaps where the difference needs to lie. It's a background task itself, like Foursquare constantly refreshing your location within the app, not waiting for you to call it into play to perform the location request.

So how big a role ? No bigger than today to be honest as it stands within Jim's current definition.

The lack of the expected telemetry from MH370 is the reason the plane has yet to be located. Events on board removed essential tracking systems from ATC. Result: mystery, suspicion, conspiracy, but most of all, anguish.

Whether it is commercial aviation, automobile safety, individual health or inventory levels we are seeing increasing deployment of technology that delivers telemetry into the fabric of our business systems and personal lives. We are dependent on it. These devices already interact with BPMS solutions to trigger events and to report status.

From sensors on a distant planet to step counters in our pockets to asset discovery systems in the data center: the agents are here; their role is central.

My reaction is the same as Theo's, articulated thusly: "Is this new?!" Heck, I can't remember how many years/decades ago it is since I saw my first "watched folder." So, yeah: important? Absolutely! but bpmNext? More like bpmNow, continuing from bpmThen.

I consider that agents are “operationally independent” and “programmable”. Theo gave good examples of old workflow robots which can do one task at the moment (as “execute and forget”). As a next step, agents can be “programmed” by executable processes – thus they become more useful, e.g. in IoT scenarios with some coordination of work.

Thanks,
AS

On one hand I agree that there is "nothing new under the sun" and that agents are "sort of like" sub-routines or sub-processes. Except that at some point "quantity turns into quality" and we have a substantially new capability. So, what is new and different then?

1) Governance and liability are different when agents may be controlled by different legal and operational entities. This is possible with existing software architectures, but with agents we expect massively distributed and scaled quantities of agents, where there may be a large number of originating legal entities. Thus governance and liability issues could be orders of magnitude more demanding.

2) And from a technical perspective, we see agents as "possessing" demonstrably greater autonomy, possibly with embedded rich domain models enabling complex behaviours. It is true that all these capabilities can be built with traditional software; but the degree of capabilities which are implied by software agents makes thinking about them as a new category of software useful.

It's probably not really helpful to think of a "watched folder" or a "Google Alert" as agents. But a machine-specific software agent for an HVAC system, which includes interfaces to building management systems, and which can be updated as needed, and which can be "set" for certain target behaviours, but which otherwise acts "autonomously", regardless of the state of our overall building system, it's useful to think of this software as "an agent".

My example is relates to the Internet of Things. As the IoT is reputed to be set for massive scaling, agent and BPM software together will will be essential -- otherwise the governance issues will make scaling impossible.

Frankly it is essential for the delivery of the end to end process application to reflect the real world of work and the required support users expect with "automation" much as already articulated. As such it will be very”big” as BPM supporting software drives next generation enterprise level software (with almost 100% zero coding?)

I am getting used to the term "agent" but it should be part and parcel of how your BPM Platform works as an automatic "orchestrator" of requirements as dictated by such as time, rules, sensors, the past actions or just as planned in the sequences of co-ordination of human or system activity in the process design. I would add that this capability will form the basis of building “intelligent processes”.

I'll put my question out there, that I had after bpmNEXT... so let's start by assuming the hypothesis. Agents are the future of BPM. So, as a BPM practitioner, what is your advice, to me? What would we prescribe to the practitioner?

For example, we could talk about climate change. But what's the advice to those practicing BPM? (that's a more extreme example of not connecting the dots).

If the advice is "find a new job, the robots will replace you" well, okay, then. Is that the real hypothesis? or no?

Huge. All of the data generated by IoT devices is only of use if it is connected to business processes. Successful IoT companies won’t simply supply devices, they’ll provide business and customer process solutions. These IoT process solutions may be agents of other parent processes or stand alone business processes solutions.

Programming by any name is just programming .... unless it isn't

At one level, "agent" is just another name for background processing. Every BPMS system I know of offers some form of background processing. I polled a couple dozen people at BPMNext and concluded that most people give a definition of agent that is just background processing, however there is a desire for something more warm and cuddly.

If you consider a system designed and built by one team, then agents are just background processing. However, if you consider that multiple different teams may be working in an uncoordinated manner on the system (like the real Internet is) then there is a specific role for agents to take.

1. One person/team designs a workprocessing system of some sort
2. Another person/team designs some software to manipulate that system as if it was you. It is a Personal Assistant.

This idea of an agent as a personal assistant is what is most promising. The agent is in many ways just software like anything else, but it is the role that it plays that makes it special. It interacts with other systems as if it was you.

To do this it needs to be quite flexible about that other system, which might change without notice. It needs to have the ability to map the data values into meaning at some level, and then apply rules to take actions on your behalf. It fits into a BPMS in the same way that a human does.

The simplest agent is one that accepts a task on one system, and starts a process in response. If designed by one person this looks like a sub-process call, but because this is done outside of the original system

See more at:

• Not an Agent, but a Personal Assistant
  
• Assistants Transform Data, Synchronize as Well
  
• Personal Assistants can connect a Doctor to a Specialist
  
• A Scenario for Discussing Personal Assistants
  

Oh yeah, VB scripting. Love it.

Clearly, very few in this forum have an idea what an agent is and how it is different from a 'watched folder', 'VB scripting', or 'background processing' ... cynical humor doesn't really hide the lack of understanding. I am not saying that agents will play a huge role in BPM. They will be the norm for ALL future software. Even a Nespresso coffee machine doesn't have a dial to set the 'length' of your espresso: 'YOU ACTUALLY TRAIN IT by holding down the button for the desired time!' So go ahead and make fun of it ...

But I do accept that all such terms are misused heavily in marketing such as 'intelligent, smart, and even adaptive'. And yes, the term 'agent' has been used for many odd things creating even more ambiguity. https://en.wikipedia.org/wiki/Agent

The only sensible agent description is the one used in artificial intelligence or -- as I prefer -- in machine learning. https://en.wikipedia.org/wiki/Intelligent_agent

'In artificial intelligence, an intelligent agent (IA) is an autonomous entity which observes through sensors and acts upon an environment using actuators (i.e. it is an agent) and directs its activity towards achieving goals (i.e. it is rational).[1] Intelligent agents may also learn or use knowledge to achieve their goals.'

So clearly the terminology describes a software entity that learns autonomously by observing a state space of information and/or is guided by humans while being focused on achieving well-defined goals. We have developed and advanced autonomous machine learning agent technology for ten years, especially in relationship to observing and acting on business processes. But many of the above responses show why it will take quite some time until it will be accepted by small minds in IT.

My 2006 patent on the subject: http://www.freshpatents.com/Method-for-training-a-system-to-specifically-react-on-a-specific-input-dt20080515ptan20080114710.php

Recent evaluation by Forrester Research:
In the evaluation of DCM vendors, ISIS Papyrus received a Strong Performer positioning in each of three Wave evaluations for Overall Capabilities, Design Time Capabilities and Runtime Capabilities, based on its scores in Current Offering and Strategy. Profiling Papyrus as

“a powerful development platform that can be targeted at a number of process problems,” the 2014 Wave report for DCM cited ISIS Papyrus for:

Flexibility of its platform for runtime behavior
Emphasis on business terminology and goal orientation
Patent for a user-trained agent (UTA)
Provides recommendations and training based on the latest knowledge gained in case handling

Read more at:

A post from 2008 on the subject.
A recent post on the subject.