Using Personas and Chat for Content Analysis
We are natural storytellers. Since we interpret the world through the lens of narratives, personas enable us to craft compelling stories about our users. We can tell stories that make their identity, needs, challenges, and goals tangible.
It’s these stores that become the foundation for empathetic and intentional design. These stories can be leveraged to shape artificial intelligence (AI) assistants, like AI assisted chatbots, into improved customer service touchpoints.
As part of a WebFirst team hackathon, our CX team recently used personas to customize an AI chatbot assistant for a federal agency website. Here’s a glimpse into how we reshaped our chatbot assistants to adapt to unique audience groups on one of the largest federal agency websites.
Part 1: Personas are like magic. They make your users appear.
Our team saw an opportunity to reduce burden on massive government agency websites. Beyond using an AI-powered chatbot to make finding content more efficient (and reducing burden on our users); we saw the potential to tailor the chatbot’s customer servicing protocols to unique user groups.
Grounded in research
We used the site feedback survey on our federal client’s agency site to dig into the data from survey respondents. From the survey, we know why each respondent visited the site, if they found what they were looking for, and what “group” of users they self-identified with.
Qualtrics Site Survey
The data revealed five primary audiences. From the data, we also knew what things each audience group were looking for on the site. One example is the “Agency Employee/Contractor” audience. Their most common goals were to:
“Review policy language concerning trainings and conferences.”
“Complete Cybersecurity Training to maintain access to the CCW VRDC.”
“Learn more about recently updated (HHS) workplace flexibilities policy.”
“Read the Updated (HHS) Workplace Flexibilities Policy.”
Bringing users to life through audience archetypes
We converted our research insights into five distinct, relatable fictional characters. Each persona is a simplified representation of the key user groups; and each highlighted the needs, goals, and pain points of a fictional character.
Part Two. A tailored AI Chatbot optimizes customer servicing experiences.
Hackathon.
Our WebFirst development team developed our AI chatbot pilot using Drupal’s AI Module and Open AI GPT for the Learned Language Modeling (LLM). The Drupal AI module allows for multiple ways to use AI in the content management system and create an “assistant.” For our chatbot, our developers used Chat GPT and configured the module to enable retrieval-augmented generation (RAG) to deliver contextually aware information unique to our target website.
This development approach supported our goal of configuring the chatbot to respond uniquely to our site visitors, depending upon each of our personas. We provided parameters for the chat agent, including custom instructions for how it should behave.
HCI-Driven Prompt Engineering
To test the efficacy of the AI chatbot in responding to unique audiences, like the “Parent/Caregiver” persona, we leveraged the persona itself as instructions for the AI assistant. We prompt engineered the chat agent to respond as an assistant specially programmed to speak and behave empathetically to a parent or caregiver’s concerns.
Outcomes
By programming the chat agents using the personas, we could see distinct differences between the chatbot’s tone, language use, and voice. Across all the persona-informed chatbots, we saw significantly more personalized communications than those delivered via the unprompted chatbot.
Public communication and information access are the fundamental purposes of most federal agency sites. And the enormity of information created and kept through federal research activities and communication artifacts is hard to fathom, much less navigate efficiently. The joint AI chatbot challenge aimed to improve how users find information on these agency sites. Ultimately, we successfully showed that the service of providing information can be improved in ease of access through increased results accuracy and via delivery formats.
It’s our team’s intention and hope to bring forward the value of using personas, like these for HHS, to deliver better client experiences across federal websites. Have you considered facilitating the flow of information and supporting this customer servicing feature on your site?
Please contact us to learn more about our AI chatbot research and its potential for improving employee and customer services.