Humanizing AI Sales Training: Design Lessons from a Real Project

As AI technology advances, AI tools are increasingly applied in training scenarios. To improve the user experience of AI sales‑training tools, humanization becomes a key metric, building trust, safety, and efficiency.

Preparation Stage: User Persona & Realism

During preparation, the system prompts users into practice mode and simulates a CRM lookup of the client’s background, creating an authentic practice environment.

The author designed user personas based on research, showing key information (company name, client avatar, account status, gender, recruitment target) briefly in the welcome message.

Physical aspect: Create realistic visual cues that give users a strong sense of immersion and improve efficiency.

Soul aspect: Provide human‑like communication language so users feel they are talking to a real person, enhancing product affinity.

Practice Stage: Dialogue & Realism

AI Customer Role: Appearance + Personality

Appearance: In practice mode the AI avatar switches from the assistant to a virtual client, delivering visual realism.

Personality: Pre‑set gender and personality generate varied voice tones and speeds, delivering auditory realism.

Sales Role: Voice Input in Practice Mode

The author replaced the cumbersome record‑then‑send flow with a space‑key toggle: press once to start recording, press again to send. The first press enlarges the element with animation, providing clear visual feedback.

Low learning cost, similar to Siri interaction.

Low operation cost, no mouse clicks, shorter interaction path.

Sales Role: Virtual Phone in Exam Mode

Exam mode uses a virtual phone interface. Users call the AI, complete the conversation, and experience a realistic call flow without mouse clicks.

Calling: AI connection with a brief preparatory cue.

In‑call: Shows call duration, mimicking a real phone call.

Call end: Provides feedback on call length.

Feedback Stage: Result Visualization & Beyond Realism

After practice or exam, users receive detailed feedback. Because traditional training often lacks granular insights, the tool offers sentence‑by‑sentence feedback, highlighting loss and gain points, and allows users to click for detailed scores.

Humanized Feedback Information

Answer hints: Two types – “recall questions” provide full answers for memorization scenarios; “flexible questions” use prompts like “please try…” to encourage adaptive responses.

Abnormal prompts: For unrecognized or non‑compliant speech, the AI responds politely, reminding users to check devices or warning against prohibited language, with tone escalating after repeated offenses.

The ongoing project continues to explore AI humanization methods, emphasizing that designers must understand business logic, address pain points at each workflow node, and convey emotional value to make AI tools feel like helpful partners rather than cold machines.