The fortunes of Microsoft’s CoPilot (and Google’s DuetAI, whenever it is launched) will depend on how well they function with other elements of any Enterprise Information System, going beyond being just a better organiser. It is likely that the terrain of the battle will be Enterprise search, which opens up multiple possibilities.
There is groundless and grounded speculation. What follows is the latter.
CoPilot’s ambition is nothing less than fusing information from the web with enterprise information, as Microsoft says it will work with Office 365 while unifying context and intelligence of the web. It has to go beyond ChatGPT (the Outside, guarding itself from fake data) and work within an enterprise information system (the Inside).
An intense battle is ahead between CoPilot and Google’s DuetAI, not yet available to the public, which will work with Google Workspace (Google Docs, Mail, Slides, Meet, Chat and many more). CoPilot has been available to enterprise customers since 1 November 2023, at $30 a user per month with a minimum of 300 licenses, underlining the importance of scale and legacy.
There is a third possibility – IBM. Will it remain silent, given its famed decades-old work in AI? I am willing to wager that IBM will not let go off this territory, especially since its exclusive focus is on enterprises. Perhaps a fourth? Amazon, but I will address this a little later.
How CoPilot could work
Microsoft has stated that customers need to ensure the integrity, privacy & confidentiality of their data, with appropriate access controls and security environment; else, security breaches will exponentially increase.
The success of CoPilot is premised on effectiveness of integrating information from without and within. Most enterprise information systems are characterized by a proliferation of software, each of which generates its own data, which means that there is already some degree of organized intelligence, with or without the use of some analytics software. In addition, there will be multiple presentations to different internal and external users with different objectives, reports, quarterly and annual data compilations or any other frequency, bid documents (successes and failures) and so on. Hence, taxonomy of information will play a crucial role especially since the web follows a certain taxonomy and CoPilot seeks to blend the two through context.
If a feedback mechanism were followed, these repositories will represent valuable sources of information. For example, some customer feedback to a presentation which is incorporated in it or notes on a bid document on what went right or wrong and so on. The ease with which any of this suite of information can be retrieved depends on the efficiency of the knowledge management system followed or something similar.
CoPilot ought to go beyond ChatGPT. Can it isolate and aggregate, say, all customer feedback relating to product performance or quality of service or whatever, based on some criteria suggested by users or on its own? Will it be able to summarise failures in bidding for project with any classification? Will it be able to cross-reference certain information? This is the litmus test for CoPilot and not some organizer that might just threaten the employment of secretarial staff.
Enterprise search in a new avatar – integrate, integrate, integrate
Disregarding tall claims about how this will free employees’ time for innovation, the targeted goal is a ‘better picture’ emerging from an integrated effort overcoming information silos and multiple data sources, but as Marvin Minsky observes, “The secret of what anything means to us depends on how we have connected it to all the other things that we know”. (The Society of Mind, Page 64)
This makes CoPilot (and DuetAI) at once ambitious and daunting, because it challenges users to identify and accept multiple perspectives excavated from multiple sources, within and without, and put them together to form a picture of the identified problem. Equally important, this exercise, which will be continuous, could well throw up newer ‘gaps’ to understand, potentially leading to newer ideas for newer businesses. In the broadest terms, this is enterprise search combined with web search facilitated by some AI engine and held together by a rigorous process of integrating, which is also open to welcoming surprises.
The contested future
It is extremely unlikely that this battle will remain a duopoly, especially with IBM’s known work in the field of AI and enterprise search, where it was among the first to launch a product called ‘Omni’. Google did also attempt to enter the enterprise search market, a just a little less than 20 years ago, but didn’t succeed.
Will Amazon remain on the sidelines, given its proven ability to organize relevant information concurrent with the changes made by users as they ‘browse’? While not perfect, the algorithm keeps reorganizing the information in sync with the changes made by users, with each ‘new page’ containing information relevant to the current search. After all, it has built a huge cloud computing business out of what it was using for its own operations. The challenge for Amazon is to go beyond its walled environment, which is possible given its sophisticated mathematical formalism.
Meanwhile, there is a significant current market for enterprise search software, some of which claim to be AI-oriented. They cannot but respond or find ways to work with CoPilot.
This is a potentially radical change, unsettling in many ways, with an intense contest ahead that will turn acrimonious, but the axis of this change are users, who need to think afresh and upgrade their skills to make use of CoPilot. Else, it will crash.
Takeaways
CoPilot seeks to combine the Inside & Outside
Has to work within an environment of proliferation of software
Some form of organised intelligence is already present in enterprises
The terrain of battle is Enterprise search
IBM and even Amazon could enter the field
