MY ROLE
Product Manager Intern
(Including UX research and design work)
CONTEXT
2024.3 (3 Weeks)
Intern Project
Two-person Team
(Product Manager + Me)
TOOLS USED
Figma
COMPANY & BUSINESS PROFILE
New Oriental, an educational company, decided to expand into the cultural tourism sector in 2023 as part of its proactive transformation efforts. The company has established a dedicated cultural tourism division with a registered capital of 1 billion yuan, focusing on tourism operations, project planning, and cultural exchange activities.
This company now primarily relies on live-streaming e-commerce to attract consumers to purchase its travel services.
The business process was as follows:
DESIGN CHALLENGE
How might we best utilize Generative AI within the CRM system to enhance efficiency while ensuring accuracy in capturing and managing customer communication records?
SOLUTION
This feature could help pan-knowledge learners easily organize and thus quickly find knowledge content they collected or created.
The demo presented here is based on the scenario where the AI summary has been generated but has not yet been confirmed. At this point, the salesperson is verifying the accuracy of the AI-generated summary.
This feature focuses on leveraging AI to improve the efficiency and effectiveness of outbound sales calls.
The AI will automatically summarize key information from the call, such as customer needs and pain points, and generate insights to guide future interactions.
The AI-generated summaries will be presented to the consultant for review and verification, allowing for corrections and adjustments as needed.
WHY COMMUNICATION CONTENT MATTERS
Communication content is a critical asset for businesses offering travel services through live-streaming e-commerce due to the nature of the product, the dynamics of the transaction model, and the high lifetime value (LTV) of customers.
Product Attributes of Travel Services
Travel services are highly personalized and intangible, requiring tailored communication to address customer preferences, such as destinations and itineraries. Clear and detailed communication bridges the gap between the intangible nature of travel products and customer expectations, fostering trust and ensuring informed decisions.
Transaction Features of Live Streaming E-Commerce
Live-streaming e-commerce thrives on real-time interaction. For high-consideration products like travel services, effective communication during live sessions helps address customer inquiries, showcase features, and build confidence. Reusing past communication content can further personalize responses, improving engagement and boosting conversion rates in this fast-paced, interactive sales model.
Customers’ High Lifetime Value (LTV)
Travel service customers tend to have high LTV due to repeat purchase potential. Maintaining detailed communication records enables personalized follow-ups, tailored recommendations, and long-term relationship building. This not only drives customer satisfaction and loyalty but also unlocks future cross-selling opportunities.
Research
Currently, salespeople have to manually input communication content into their existing Customer Relationship Management (CRM) system, which helps the company keep customer contact details up to date, track every customer interaction, and finally. This way, they could quickly follow up on customer status, address inquiries, and guide leads through the decision-making journey based on the "Communication Record".
CHALLENGES & OPPORTUNITIES
However, there are 2 huge challenges from both salespeople and the company’s perspectives.
Frequent turnover among sales personnel resulted in challenging handover processes.
When sales staff leave or change positions, unclear or inadequately executed handover procedures make it difficult for new colleagues to quickly get up to speed. This hinders their ability to understand detailed client information and historical interactions, thereby affecting service quality and sales efficiency. It also leads to the loss of valuable knowledge and experience.
Incomplete and untimely updates of communication records reduced efficiency and satisfaction.
Sales staff failing to document client communications in a timely and comprehensive manner impedes effective information flow. This can result in follow-up errors and redundant work, ultimately damaging the customer experience and reducing satisfaction.
Amid these challenges, the rise of generative AI offers promising opportunities to simplify and automate the repetitive and time-consuming task of creating customer communication records, paving the way for improved efficiency and accuracy.
QUANTITATIVE RESEARCH
After realizing such a huge opportunity for AI application in the CRM system, I started to conduct user interviews to understand more detailed user scenarios for later design direction.
5 individuals participated in the interviews, including 1 with less than one year of experience, 2 with 2–3 years of experience, and 1 with over three years of experience. Here are the insights and corresponding solutions. Besides, guided by our team's programmer and focusing on the key communication elements that salespeople prioritize, I wrote prompts specifically designed for summarization.
SALESPEOPLE’S CURRENT WORKFLOW
To gain deeper insights into how these challenges impact daily operations, it’s crucial to closely analyze the current workflow of salespeople. Their current workflow consists of 3 stages:
DESIGN GOAL
How might we seamlessly integrate the AI summarization feature while enabling salespeople to provide feedback and make edits, ensuring a smooth workflow that delivers high ROI?
BRAINSTORM
Based on the existing sales workflows and CRM pages, I have undertaken the design and development of three key functionalities.
Design
THE UNEXPECTED CHALLENGE I ENCOUNTERED…
Initially, I assumed the design process would be quick since it appeared to involve 2 new pages. However, I quickly learned an important lesson.
During the meeting with the technical team, I realized I had underestimated the complexity of the backend processes required to implement the key summarization feature.
The summary cannot be generated by AI in one step!
The actual implementation involves a series of intricate steps:
This realization highlighted the need for multiple statuses for each component.
The table below outlines the stages of the summarization process, structured according to the MECE (Mutually Exclusive, Collectively Exhaustive) principle. Each stage is clearly defined, ensuring all necessary steps are addressed without redundancy or overlap.
Therefore, let’s take the “record” component for example, it seemed that I need to design 5 statuses for it.
Initially, I considered providing detailed explanations for why an AI summary might not be completed. However, after completing user testing with 3 users, it became clear that such detailed prompts only added confusion and led to information overload.
To address this issue, I simplified the status updates into a single, unified message for the first three stages of the summarization process. This ensures that users are not overwhelmed with unnecessary details while still being informed about the progress.
The demo presented here is based on the scenario where the AI summary has been generated but has not yet been confirmed. At this point, the salesperson is verifying the accuracy of the AI-generated summary.
Outcome
Future Work
Enhancing AI Summary Accuracy and Identifying Future Product Iteration Directions
Monitoring AI summary precision based on user data: To ensure the effectiveness of our AI-generated summaries, we will implement a robust monitoring system that evaluates the precision of these summaries using user data.
Deep Dive into CRM Product Stakeholder Workflows and explore other opportunities for generative AI
To fully understand the needs and workflows of all stakeholders within the CRM system, we will conduct a comprehensive analysis of their processes. Potential areas include:
Smart Lead Scoring: Use AI to score leads based on historical data and behavior, helping sales teams prioritize high-potential prospects.
Real-Time Sales Recommendations: Provide real-time recommendations for sales pitches and product offerings during customer interactions.
Predictive Analytics: Implement predictive analytics to forecast customer needs and behaviors, enabling proactive customer service and marketing strategies.
Reflection
DIFFICULTIES I MET & THINGS I LEARNED
Don't try to write the perfect design in one go. Instead, embrace an iterative approach, breaking down the process into smaller, manageable steps. This allows for flexibility, incorporating feedback and adjustments along the way.
Establish clear deadlines for brainstorming sessions to foster a sense of urgency and maintain focus. Without time constraints, brainstorming can become unfocused and unproductive.
Always document the reasoning behind design decisions. Thorough documentation ensures transparency, facilitates communication and prevents wasted effort.
When approaching a problem, be mindful of any pre-existing assumptions or preconditions that might be influencing thinking. (Yes, the Sequence diagram, I know…) Dig deeper to critically examine these assumptions and ensure they are valid. Every design choice stems from certain preconditions. Failing to recognize these can lead to flawed solutions. Sometimes using tools like recording, transcription, and AI summarization can help a lot in this process.