This is not your usual chatbot! The Enterprise Business Agent (EBA) is a ‘patent-pending’ approach to machine reasoning – trained to understand special areas of interest, such as Marketing.
The main issue with chatbots is their lack of support of the required depth and breadth of complexity in the workplace. What is a marketing campaign? Compare open rates across certain campaigns within the same audience? You can go pretty far pretty fast.
The EBA engine is a different approach and designed by a team which has a combined 40 years experience in active AI development and especially machine reasoning. How far it can go? Try it yourself!
EBA began in 2016 as a POC across Watson Customer Engagement. The goal was to create an assistant for the business practitioner in specific areas of interest – such as marketing. We have since released it in our Watson Campaign Automation capability and seen quite an interesting number of improvements. Allow me to give you some examples:
- Significant reduction in daily/weekly/monthly reporting efforts
- Increase of conversion by being able to plot metrics on the fly
- Reduction of work effort to manage the solution and that directly linked to
- Increase of campaign output
The platform behind the EBA is massive at scale. The Watson Customer Engagement platform processes a massive amount of data:
- 1+ Petabyte of data ingested
- 83 Billion sends
- 56 Billion customer profiles
- 8 Billion events processed
- 51 Million documents processed
- 6 Million documents transferred
- …
The scale of the supporting platform shows how far we have come on the SaaS platform side.
I have previously written a blog about the ‘Watson Marketing Assistant’ here:
To this day, it is the blog people read the most. The EBA is the capability underpinning the Watson Marketing Assistant. The good news is: it does tell jokes and can solve riddles.
The big change however is that you can give the EBA a try yourself now. Want to add it to Slack? Download it as an app, or even launch it in your browser? You can do so here: EBA.AI
Now the sense of humour of the assistant can be argued, but this is just one example to show how far the EBA can eventually go:
So what are some of the more interesting features?
- Flexible data visualization, including plotting arbitrary vectors
- Generic tone analysis
- Multi-channel and cross-channel mode
- Multi-human actor mode
- News
- Next-best action suggestions based on current session state
- Non-conversational data mode (process state/transition observation)
- Time-series evaluation
- Track and notify business user for any data state change
- User and organizational profiling (goals, language)
- Weather Data
So the above are some ‘out of the box’ features. What makes the EBA different though is the approach to an open developer environment. You can literally create and develop your own ‘agents’. Below you can see a list of the shared agents and the capability to build and create your own agents.
So when applying some of the above capability, the marketer can easily leverage the EBA engine in the Watson Marketing Assistant to identify all kind of useful insights from the data available. Have a look at some of them here:
I have had the joyful experience to see my esteemed colleague Michael Bordash also demonstrate to me, how far this capability can go. In the short period of a weekend, Michael developed an agent capable of understand a mix of Supply Chain and Marketing data points and cross-reference it with weather data. The use case? Based on severe weather conditions in the US the assistant was alarming the user. I was able to check affected shipments, review current stock in my stores and then see the impact on marketing campaigns. Net result: I stopped some of the campaigns on products unavailable and stoped further chaos from happening…
Imagine how far this capability will go, when we give Michael and his team more than a weekend.
Michael also presented this use case in San Francisco this year – just remember – he was able to build this in record time…
Would be interested to see where you think this is able to go. In the meantime, I will continue to review our ongoing releases on this capability – as this is not a chatbot, but an intern today – who may in some time become a colleague.
