Project: Message now platform Date: September 2022
Client: BT Wholesale Company: EPAM
Skills used: Leading workshops/ Wireframing/ Prototyping/ CUI design/ Team leadership/ 3 Amigo’s/ Requirement gathering/ Stakeholder management/ Sprint planning/ Content strategy
The problem: After speaking with BT Account Managers and the Service Help Desk, we found a lot of their time was spent solving issues users could easily self-serve—like resetting passwords or finding the right documentation. This took time away from fixing bigger problems, like outages, and often put users into frustrating, multi-call journeys.
My role: To create a chatbot that could sit across the site, surface common questions based on the page the user is on, and guide them to the right support teams when needed.
Intro: Collaborating with a Business Analyst, Content Strategist, and UI Designer, we explored how a chatbot could ease pressure on service teams by answering common, self-serve queries.
Sprint planning: Since the team was new to chatbot development, I started by breaking the work into epics and sub-tasks. Outlining what was needed from UX, UI, BA, content strategy, development, and QA. This helped the Scrum Master and PO shape the quarter’s workload and populate Jira, giving the business visibility into our squad’s roadmap.
It also helped set expectations that the chatbot launch was just the starting point, highlighting tasks that would need to become BAU, especially around ongoing content updates based on user behaviour and feedback.
Initial workshops: The BAs had mapped initial requirements and technical flows. Building on that, I ran a FigJam workshop with the internal team to identify key areas, surface questions, and map out conversation components. This helped us define core interaction types—like intro messages, navigation prompts, FAQs, intent flows, and handoff to a live agent.
Scripting: We then used the conversation flow from the workshop to run a quick scripting exercise—helping us explore tone of voice, user inputs, and how the chatbot might guide interactions in practice.
Flow workshop: Now that the conversation parts and language were outlined, the Content Strategist, UI Designer, and I collaborated to create a more formalized flow. This helped us see how the conversation pieces fit together to create a smooth user journey and identify where flows might differ. Like offering general FAQs on the homepage but more specific support during an order process, and deciding when to route users to a live agent.
Delivery flows: Next, I formalized the flows into a clear deliverable for the business and to guide the copywriter in refining the text.
This was then used in end-of-sprint demos and Amigo sessions with developers, helping us address questions, gather feedback, and collaborate closely with the dev team.
Content: Next, we ran workshops with four support teams to identify key user issues and challenges. We grouped these into reseller self-serve, service team tasks, and issues for other teams. This helped define chatbot content and highlighted that ongoing content updates would be needed as the site evolves.
Content. Synthesis workshop findings: Next, I organized all the content ideas into a spreadsheet to prioritize what needed creating. We reviewed this with support team leads to confirm details and gather more input. The document also became a useful resource for other teams, like the learning squad managing the training portal.
Content. Creating content for the chat bot: The next step was creating a spreadsheet to populate content for the devs to integrate into the chatbot. We ran several sessions to develop 20 intent variants, helping the AI learn and match the right responses.
Review build After the build was complete, the final step was reviewing it and making small adjustments to improve any flows that weren’t working as smoothly as expected.
