Mining for Gold?

With the crytpo-winter of discontent being made at least damp spring by this ton of dork-y ChatGPT chat, it would be easy to forget the other dripping hype-cicles arrayed along the eaves of your metaverse house.  Not so the Economist, whose headline ‘The real next big thing in business automation’ suggests that process mining rather than chatbots will be the key to further progress.  We think they are overlooking a bigger picture.

We have followed the work of Wil van der Aalst since his book on Workflow Management.  He has provided us with a number of principles that have influenced our thinking on business process design: that there might be some ‘best practice’ patterns to guide the re-engineering initiative and that with the right software you could round-trip from visual BPMN process flow diagrams to executable process instances.  That software used to be quite specialist, expensive with high set-up overheads but the low code revolution has placed such capabilities into the hands of even small businesses.

His next initiative was Process Mining – the idea of rather than starting with a process diagram and an idea of how you think the process works, using a output log data of the process to mine for insight on how it does work, and use that to suggest improvements.  Again this was the preserve of a number of specialist tools like the Celonis mentioned in the article but has also been added to the Microsoft Power Platform in the form of Process Advisor and therefore to a wider range of users. 

Here though we diverge slightly from the Economist.  Firstly this article is not going to mention Adam Smith, Walter Bagehot or make any reference to Potemkin villages, setting it apart from most Economist writing (Too late! -Ed).  Secondly, the true transformational impact in all of these technologies could be in their combination. 

You could start from wherever you are in the lifecycle.…either on the cusp of an interesting idea or the midst of a process spaghetti.  For the sake of this narrative lets start with your whiteboard hypotheses and use ‘text to image’ and/or ‘image to vector’ to convert it to geometrically precise visualisations that you can augment with metadata. You could then model interactions with system dynamics software to explore any non-linear emergent behaviours and, duly adjusted, you could seamlessly instantiate your low code executable modules to get your prototype working.   You could create synthetic data to stress the model, use web automation software to carry out more realistic testing.  Our current weapon of choice is Playwright but there are many options and most of the RPA tools have capabilities in this area. You might even find a use for the ‘digital twin of customer’ thinking that sadly only we and Gartner seem to be excited about.  Until this point, no real humans have been injured in your experiments, but since you have some working software you can start to test the UX.  However you do it, the output of all of these approaches is log data.  This is where data mining tools can first help, perhaps giving you insights that your hypothesising didn’t consider.

You could also start ‘mid-flight’ as the ‘fixing the plane while flying it’ analogy suggests.  You might already have an ‘as is’ hybrid model which is a mish-mash of automated and human behaviours.  As well as data mining, we are also excited about our causal inference experiments which can turn such log data into visual and statistical hypothesis of causality.  Either way you now have a functioning model of your business that mirrors your real business.  This in itself is probably a dramatic improvement in your current capability but how do you turn that capability into value?  DevOps thinking covers much of what we have presented – from hypothesis testing through to automating both business and infrastructure tasks.  You could expand to a ‘DataOps’ posture and seek to integrate your measurement processes proactively into your continuous improvement thinking.

This is where some of these elements might start to be combined.  The potential of generative AI neural nets to replace current process exception handling with a more nuanced set of behaviours ranging from the deterministic to the more probabilistic non-deterministic offer a real opportunity.  We have long subscribed to Robert Simon’s ‘Levers of Control’ as a conceptual framework but we may now be on the brink of being able to implement it as an actual control mechanism.  Perhaps the boundary between diagnostic and interactive management doesn’t have to be so clear cut…how much more ‘in flight’ adjustment could be brought under purview of the bots.

The overall vision is the ‘self-healing’ business process.  Many other specialist disciplines have made advances in this area.  Even the aviation domain of the ‘mid-flight’ analogy might be better used for the concept of ‘auto pilot’.  Think of how sophisticated an auto pilot must be to respond to a myriad of data inputs and adjust outputs for a safe and efficient journey.  If automated testing gives rise to ‘self-healing’ testing and ‘infrastructure as a service’ scripts give rise to ‘self-healing’ infrastructure then is it really such a leap to aspire to ‘self-healing’ business process.  This is where another one of our research themes may play a part: knowledge engineering.  Representing the knowledge and cognitive processes that previously were the exclusive domain of humans will be key to re-imagining the human/computer interface for the benefit of both. 

In short, the combination of new technologies and the fulfilment of others seem to us to promise that ‘glorious summer’. It is as exciting a time as we can remember in our working lives. Since this post started with a butchering of Shakespeare, let it end by asking him what Julius Caesar might counsel at this juncture:

“There is a tide in the affairs of men, which taken at the flood, leads on to fortune; omitted, all the voyage of their life is bound in shallows and in miseries.”

If you have been affected by any of the themes raised in this blog then please don’t hesitate to get in touch.