More research on the history of cognitive assistants in business

As I research the history of cognitive assistants, these blog posts by Franz Dill and one more by D.J. Powers are interesting, relating to executive information systems to help CEO’s in the pre-spreadsheet era of computing, as well as executive decision support systems:

Early Business Intelligence Needs

A Slice of the History of Executive Information Systems

A Brief History of Decision Support Systems

Broadly, Franz is interested in the topic of how the executive (and every decision maker, beyond the analyst/research), uses data, analytics, and intelligence systems to make better decisions.   Cognitive assistants with 3 L’s capabilities – language, learning, and levels (of confidence) – provide a next step and and new set of pathways to explore.

Franz emailed me about his work at P&G developing and deploying early advisory systems – “The consultant to all of our advisory systems in the 90s was Stanford and Teknowledge.   Using Prolog style systems, like M1.  Edward Feigenbaum and others were involved.  The most famous of our systems was the Coffee Blending advisory system,  which was used by green coffee blenders, and in use for the next decade, saved P&G millions.   Another famous system was the Copy Owl, which let Ad experts, use, reuse, modify and apply company advertising assets.  We even played with learning systems, I wrote an early neural net based induction system that matched advertising campaigns to new product initiatives.  A related system gathered cases and then used case based reasoning (CBR) to find best fits.   It was used for several years, then the task was outsourced.   Another system adapted or ‘learned’ supply chain solutions from traffic and inventory data. Early big data.   Except for the four examples of executive systems I sent you, and to a certain extent the Copy Owl, these were all used by corporate experts to augment their own specialty expertise.  Sometimes to replace their own expertise for easier problems and also often to scale their expertise more broadly.”  

Clearly some of these systems worked better than others, and their fates depended on many factors.  Franz is also interested in why some worked for a while, some did not work,  and how they could have been made to work in a specific corporate executive culture.  One comment that stuck with him was when an executive said something like “I am most interested in useful creativity, not keeping alive the expertise of past executives, even my own …. “.  So successful cognitive assistants will require more than just the expert system processes of accurate expertise capture … more effort is needing in trying to find the true dividing line between creativity and handling routine tasks efficiently.   Cognitive assistants have to be collaborative partners that handle routine tasks, but also engage in natural and creative exploration of new ideas and possibilities.  

Franz also commented that “What ultimately worked was extreme focus.  Narrow understanding of what was needed to be done, simple or complex.   And making sure the user (exec or analyst) was willing to follow the lead.  And their organization also would follow the recommendation.  And that the data involved was trusted for the given purpose by all …  All of this worked better with focus.    In three years we did perhaps thirty projects,  only three in the executive space. Many other projects in marketing, manufacturing, R&D, Supply Chain, HR, Sales … With a wide variance in success.  The successes paid for the effort, but did not sustain the AI function.”

The three systems in the executive space were part of a C-Suite Advisory Development effort at P&G from 1989-1991:

Automated CEO – lead by Bob Herbold (Later COO at Microsoft)
New Initiative Advisor – lead by Tom Moore

Major Capital Appropriation Screener  – lead by Bob Hunt

Across the industry it has been estimated the annual spending for artificial intelligence projects in industry was reaching $1B with nearly ten thousand employees, both the major firms and startups were experimenting with better AI tools and techniques, decision-support systems, advisory systems, performance-support systems, and a variety of other intelligent systems, but none with the language, learning, and levels (of confidence) capabilities of the cognitive assistants that are beginning to appear today.

Smart service systems will depend increasingly on the people inside and outside them equipped with cognitive assistants.   The ability to serve external and internal customers in the business world will likely depend on cognitive assistants more and more, as our tools come to know us.

See “A Brief History of Cognitive Assistants:” for more in this ongoing discussion:

Leave a Reply