Situation:

Design chatbot to serve as a digital assistant, to obtain info and also complete certain tasks, on demand. Such as:

 

Personas & use cases:
Compliance Counterparty Credit Research reviews potential fines, rule violations, market manipulation, and frauds. It searches FINRA (Financial Industry Regulatory Authority) sites. And other countries' sites, like FINRA is to the United States.

Traders use the Bot to find out information about recent trades, legal and regulatory updates, contact info for people, etc.
Analysts use it to obtain info, and complete tasks.

Portfolio Managers The client, the company, has 800 combined info sources (systems, apps, platforms, models). So, sometimes, a user does not even know where to start. Especially a Portfolio Manager, who is not so entangled in the weeds, and day-to-day operations.

 

Location:

Bot lives in Symphony. Symphony is a chat client, used only for institutional finance. It's like Facebook Messenger or Slack, for financial corporations. Read more about it here.

 

Team:

Delivery Lead, Product Manager/Business Analyst, Data Scientist, Front-End Developer, Back-End developers offshore.

 

Task:

I joined the project, when Bot was already in code. But it did not have UX support, prior to me.

● My job was to make it easier for users to engage with the Bot. To understand what Bot can do and how it can help them.

● To serve information proactively, to think one step ahead.

● To think through how to engage different types of users. Trader vs. Compliance officer, for example.

● To advise on error handling.

 

Action:

● I added opening lines, and landing dashboards, which eased initial engagement

● Outlined error handling, which reduced abandonment rates

● Structured information into common reading patterns, for effortless consumption

 

Note:

Due to confidentiality considerations, and competitive edge aspect of this project, client name is withheld, all real data is blurred, and bot name is made up. For now, it's web based (not mobile), and text-based (no voice). Both mobile and voice are slated for next year.

 

Result:

● 100% success rate. With traders as users, there's no room for error. If there were wrong answers, no one would use the Bot.

● 200 users globally; analysts and traders

● post COVID, the number of users significantly increased, as users were no longer co-located

● 20+ use cases, across analytics, trade history, holdings, and Compliance
● 1-2 seconds response time
● 50+ intents (one intent could answer 50+ questions, as it was designed in a specific way)
● 12 ontologies
● 6.2M entities