The Problem

The rapid rise and accessibility of AI and ChatGPT has opened up so many opportunities and benefits for businesses. However, actually understanding and integrating AI is complex for many business owners. We had developed the technical aspects of creating, training and managing AI Agents to interact with customers, but did not have a suitable interface for customers to use it autonomously.  

How Might We

Give customers an easy to use platform that allows them to create, train and manage AI agents autonomously, with minimal manual help from the AI Agent SaaS team.

Customers are aware of AI and its powerful capabilities, but struggle to know how to integrate and apply it to their business.

How might we bridge this gap?

Quantitative & Qualitative Research

Understanding Stakeholder Needs
To initiate the project, a rapid round of stakeholder analysis was conducted. Interviews with sales teams, managers, and executives were conducted to understand pain points, challenges, and expectations. Key insights gathered included the need for real-time data analytics, lead prioritisation, and personalised customer insights.

Market Analysis
A brief market analysis was conducted to identify competitors and industry trends. This analysis revealed gaps in existing solutions, emphasising the opportunity for innovation in the design of our AI sales tool.

User Personas
Based on the research findings, detailed user personas were created to represent the diverse needs and roles within the sales ecosystem. This helped in tailoring the tool to specific user requirements, ensuring a more user-centric design.

Design

The AI Portal

Ideation and Conceptualisation
Collaborative workshop sessions were held to generate ideas for the AI portal. This would be the product that customers would interact with to create, train and manage their AI Agents. The focus was on creating a seamless user experience while integrating advanced AI capabilities. Designs and prototypes were developed quickly to visualise the concept and gather initial feedback from stakeholders.

Clear Categorisation
‍I prioritised the information architecture for the portal, splitting the core functions in to 'AI Agents', 'Training' and 'Insights'. These three pillars covered the primary value of the product in a simple and clear way.

Visualising Vast Data

Recognising the Challenge
One of the key challenges in designing the AI sales tool was the need to process and present vast amounts of data in a way that was not only comprehensive but also easily digestible for users. The success of the tool hinged on its ability to empower users with actionable insights derived from complex datasets. To tackle this challenge, a deep dive into understanding the users' mental models and preferences for data interpretation was essential.

Engaging with Potential Customers
To bridge the gap between technical sophistication and user comprehension, a series of interviews and workshops were conducted with potential customers. The goal was not just to gather requirements but to understand how users naturally conceptualised and mapped data in their minds. Conversations delved into their existing workflows, preferred visualisation styles, and pain points in dealing with data overload.

Uncovering Mental Models
Through these interactions, distinct mental models began to emerge. Sales professionals often visualised their data in a linear, time-based fashion, tracking leads from initiation to closure. Managers, on the other hand, tended to conceptualize data in hierarchical structures, emphasizing team performance and goal attainment. Recognizing and respecting these mental models became a cornerstone in the design process.

Balancing Detail and Simplicity
A crucial aspect of the design process was finding the delicate balance between providing sufficient detail and maintaining simplicity. The challenge was to avoid overwhelming users with too much information while ensuring that critical insights were not lost. This required continuous refinement and testing to optimise the level of granularity in the visualisations.

Transparancy = Trust

‍A primary concern from customers was not having any visibility of how their AI Agents were actually interacting with customers. I wanted to include a clear way to access the conversation data and make it simple to 'flag' or 're-train' any concerns the customer had about the interactions.

Delivering Value With Insights

Understanding that raw data alone does not drive decision-making, the design process placed a significant emphasis on the integration of an "Insights" section within the AI sales tool. The objective was to empower users with actionable intelligence derived from the AI agents' analysis of customer behaviour. This section aimed to go beyond presenting data and instead provide users with a deeper understanding of their customers' actions, focusing on both popular questions and reasons for not converting.

Conversations with Stakeholders
To ensure the Insights section met the diverse needs of users, extensive discussions were held with sales teams, managers, and customer support personnel. Through these conversations, common pain points and recurring challenges were identified, laying the groundwork for the types of insights that would be most valuable. Feedback from these stakeholders played a pivotal role in shaping the content and features of the Insights section.

Insights and Actions Via Chat
‍As well as clearly displayed data, it made sense to utilise the conversational AI technology we had built to allow customers to chat directly to their own 'AI Analyst'. This gave the customer the opportunity to simply ask what they were looking for, or for the AI Analyst to make smart suggestions automatically.  

AI Agent Performance Monitoring
Central to the Insights section was the monitoring of AI agent performance. Machine learning algorithms were employed to analyse customer interactions and discern patterns. By tracking the customer journey, the AI agents could pinpoint common reasons for both successful conversions and customer drop-off. This real-time analysis provided users with a dynamic view of their sales processes, allowing for proactive decision-making.

Key Insights Highlighted
Within the Insights section, a carefully curated set of key insights was presented. This included:

Popular Questions:
Identifying common queries and concerns raised by customers.
‍
Reasons for Not Converting:
Highlighting patterns associated with customer bounce, allowing for quick intervention and targeted improvements.

Seamless Onboarding

Initially, behind the scenes, a lot of manual processes were taking place by the support team. This included uploading knowledge documents from the customer and training the AI to behave in a particular way after asking the customers a series of questions, either via email or on a call. The business goal was to scale the product and remove the manual processes from the team. This required the product to be intuitive enough for customers to start using themselves.

Creating a smooth onboarding flow would be the best opportunity to set the customers up for success and help them understand how to get the most from the product. I was keen to avoid using too much technical language, which could be intimidating for customers and provide a friendly, human-to-human type of communication.

Simplifying the Complex

Breaking down the original lengthly process into smaller chunks reduced the perception of complexity and intimidation. It also provided an opportunity to guide and communicate with the customer in an easy-to-understand manner. For example, explaining why we were requesting certain information and how it would benefit them. A structured onboarding flow helped to shape a mental model of the product for the customer, making future use more intuitive for them. It also provided a smooth transition from the 'Setup' moment to the 'AHA' moment.

Testing & Rollout

Beta Testing with Customer Collaboration
As a pivotal phase in the development, we engaged with a select group of beta customers to validate and refine the AI portal. These beta customers, representing diverse industries, played a crucial role in providing real-world feedback. Regular collaboration sessions were conducted to gather insights, understand user experiences, and identify areas for improvement.

Global Release and Ongoing Collaboration
With insights gained from beta testing, the AI sales tool was globally released. The collaborative spirit established during beta testing was extended into the post-release phase. Continuous collaboration with users worldwide has became integral to our approach. Regular feedback loops and customer success teams were established to facilitate an ongoing dialogue, allowing us to adapt the product to the evolving needs of a global user base.

Business Impact