Evaluate Weigh the pros and cons of technologies, products and projects you are considering.

Without transparent AI, the bots will run wild

LAS VEGAS -- While many vendors champion the capabilities of bots and machine learning, some are calling for a layer of AI transparency to understand how the tools arrive at their decisions instead of feeding information in and getting information out the other side of what can basically be a black box.

Transparent AI is a focus at Pegasystems, said Rob Walker, VP of decision management and analytics, because the company's customers must comply with a myriad of regulations internationally. The problem is, when a bot makes decisions that could be construed by humans as racist, misogynistic or maybe just counter to the company's mission statement, humans need to figure out how the bot got there.

Moreover, Walker said he believes that AI is in its infancy, and transparent AI will become more and more important as artificial intelligence evolves. Starting now with a clearer understanding of AI's decision-making processes can set us up to use it better -- and control it -- down the road.

Because a bot can't take the stand in a class-action lawsuit, Walker said, the humans maintaining its governance and maintenance are ultimately accountable for the bot's behavior. Transparent AI -- which in the Pega universe manifests as a feature the company calls the T-Switch -- will be an important factor in the success or failure of AI-powered software automation in the coming years.

And what will the bot future look like five years from now? Click on the above video to find out what the cautiously optimistic Walker, an AI scientist who believes bots can and will eventually be able to run businesses -- with supervision, of course -- and free humans to invent better processes and better deal with high-touch customer relationships, sees.

View All Videos

Join the conversation

2 comments

Send me notifications when other members comment.

Please create a username to comment.

Is AI transparency important to your CRM -- sales, service, commerce and marketing -- efforts? Why or why not?
Cancel
It seems that, if each example of AI's decision-making is described as instantiation of a CASE (with content, goals, rules & activities) instead of a straight through process (with pre-conditions for state change), compliance automation is feasible.  A virtual assistant/bot, guided by an ontology and by a AI/ML-detected activity pattern deemed as successful (including all CASE attributes, content, rules, KPI's etc), can advise what are the "best next actions" (i.e., most compliant).  
This does not fully automate compliance, but nudges workers to do the "compliant thing," mitigating risks considerably.
Cancel

-ADS BY GOOGLE

SearchBusinessAnalytics

SearchDataManagement

SearchSAP

SearchOracle

SearchAWS

SearchContentManagement

SearchSalesforce

Close