AI-powered operator bot

Building an AI-powered operator bot

Our customer is a FTSE 100 multinational with a prolific service desk. Already using WhosOn behind this service desk, the customer wanted to further speed up their support offering by incorporating AI and bots. So, we built them a custom solution.

What we did

  • Built a proof-of-concept bot application as an initial pilot project
  • Created a custom integration between WhosOn live chat and IBM Watson AI
  • Deployed a live bot to the customer’s service desk that acts as an operator
  • Enhanced the bot with the full capabilities and features of the WhosOn platform
  • Allowed oversight, monitoring and chat transfers/human escalations to be handled natively

How we did it

  • Consulted the customer on bot design and best practices
  • Arranged sessions with IBM to “teach” the AI technical support
  • Used Botkit to connect the back-end processes between bot and chat
  • Updated the WhosOn client libraries and ported into Node.js using Edge.js
  • Interfaced with Botkit, learning the pipelines for how to create a new connection type

The results

With a service desk that incorporates a tech support chatbot, our customer has been able to:

  • Improve overall customer satisfaction scores for long processes
  • Take simple and repetitive actions away from the service desk, reducing the number of tickets
  • Take long, complex process-driven answers (for example internal hardware requests) away from agents, preventing them from getting tied up in a script
  • Provide quick feedback and follow up on open tickets
  • Explore further chatbot potential, with a project for a Facebook Workplace/WhosOn bot already initiated


*We typically deal with regulated industries and enterprises with more complex live chat requirements. So, we often deliver chat within a legally binding NDA. This is one such client whose identity we keep under wraps.