Conversational integration part 2.
In part one of this series I talked about the broad concept of ACI (Artificial Corporate Intelligence) and how it will help in moving forward on the OCP (Organisational Cognitive Process) journey.
In this post, I want to explore the possible architecture of amalgamated corporate AGI and the interactions between different systems’ SAI.
NOTE: When I refer to “AI”, I mean a member of the AI, ML or DL group, depending on what would best suit that situation. That sort of decision is reserved for an undertaking outside of this article — I am simply using it as a catch-all at this juncture.
AGI: Artificial General Intelligence
ACI: Artificial Corporate Intelligence
SAI: System-specific AI
WAI: Web-facing AI
TLDR: I contend that the next 10 years of SME AI we will see the evolution and development of a system specific, internal, heavily biased nodal approach to “cognitive” architecture where, using the API interfaces of specific systems (ERP/CRM etc.) you will see multiple ML/DL systems present their own isolated perspective of the movement, meaning and method of system specific data, in turn presenting that to a multi-faceted corporately biased inner ACI. This ultimately will be bias-adjusted by an external “WAI” feeding into it predesignated information based on wide aggregated corporate search data requests/trends etc. and industry related traffic. This will result in a company-focused, industry-weighted corporate-perspective driven by the SAIs that span an organisation, leading to an ACI that will offer information and knowledge based on all avenues of data, suited to the specific organisation — a “fingerprint” solution.
I still love the idea of using APIs to explore information and whilst one *could* use the raw atomic data or create your own pathways to the records and draw your own conclusive views from those data, it seems that the simplest way of accessing collected information and data in the first instance would be to utilise existing application APIs.
Imagine each discrete system having it’s own gatekeeper, it’s own knowledge artisan, I call this the SAI. The SAI (System AI) would be tasked with interrogating and using the (in this example) ERP API to work with, adopt, adapt and extend existing perceived and designated data structures as well as the predefined calling endpoints and their implied meanings. This would of course bring with it not just the RAW DATA but both the intrinsic and extrinsic biases that led to the data creation (and that are inherent within the design of the API itself) but I would suggest in this case, that bias (in the corporate data sense) is POSITIVELY important due to the uniqueness of each organisation and the uses to which it puts data, as well as the trained implications further on. In this case, we should embrace the bias because it actually tells us more about the corporation, its implied meanings, its derivative leanings, its internalised response frameworks and, by considering the queries raised, the rational and intellectual capacity/facility as a whole for those who access these data.
So, the SAI would need to gain an understanding of the data within the system in regards to THAT system alone and the interactions that users, other systems have with it. Essentially the SAI would need to generate institutional knowledge deliberately biased to the system itself, in short, it would need to perceive the system that provides data via the API as the centre and single element of its own universe.
It would need to analyse the transaction taking place, what’s happening, when, what are the intrinsic knock-on effects data and record-wise when say, an invoice is posted, or a GRN is created. Perhaps it will need to understand the logic of the BOM module to really “get a feel” for the intricacies of the manufacturing timeline and relationships between AP and AR data in both the short and long term.
It would need to be trained on the data present and presented, developing its own corporate “knowledge” outside of the standard users interacting with the system but, using knowledge based on how they used the system and how data is moved around it.
The same would be said for a separate CRM system, the understanding it will need to gain from customer centric non-sales or sales-quote data, learning the timeframes, nuances and wider view of the data and how it generates it’s own understanding of that system.
These are only my preliminary thoughts — would love to hear your comments and I am no AI scientist but a mere enthusiast. Howvere, having spoken to a LOT of CEOs, CTOs and CFOs about how data enters, is used, refined, finessed, extracted, remodeled and presented it is clear that there is a lot of scope for the affordable automation and digitiation of these processes and perhaps we need to move beyond RPA or simple workflow of data from one system to another, we need to move organisations into the automated knowledge, the digital seer, environment where the next step from knowledge to insight, and beyond that into wisdom that is both accessible and achievable through the small steps of SAI.
In a sense we need to de-silo the collective corporate wisdom, safe in the knowledge (excuse the pun) than the necessity for human input resides within each organisation for without humans, there is no corporation. We are fast approaching the nexus of the shift to the fourth industrial age, where we humans are freer to be creative, where we can have the time and space to think the unthinkable, to do that which we never thought possible, to stand on the shoulders of digital giants and reach the heights that we thought were beyond us.
My next post will look at the ramifications for this new phase of corporate evolution and what the technology really means for people. Do get in touch and share your thoughts !— shaun.turner@thoughty.studio.
If you are interested in coming onto my upcoming podcast show (launching in 2022) to discuss this or any other aspect of automation, business 4.0, dgitisation, automation and/or creativity then please, get in touch!
A word on biases. Normally, biases would be a bad thing, normally in general AI e.g. Siri or Alexa or in particular, image based AI scientists try to eliminate both conscious and unconscious (or intrinsic) biases but I would contend in this particular use-case where the system a) naturally acquire biases from the design of the API (through which the AI will access the data) and b) naturally acquires system-specific organizationally contaminated biases (SSOC), this will help in the organsiational-fit for this particular SAI. The ACI would still inherit the sub-biases by dint of it’s interaction with the SAIs but, the Extrinsic AI (EAI) will need to be as far removed from biases as possible as it looks at industry/demographic/media data to feed into the ACI, which will result in a de-biasing to some degree depending on external data weight.
