sales-convin

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Convin Platform Help

Convin平台帮助

Step 1 — Gather context

步骤1 —— 收集上下文信息

references/learnings.md

exists, read it first for accumulated platform knowledge.

What do you need?
- A) Set up automated QA scoring
- B) Configure Real-Time Assist for agents
- C) Deploy AI Phone Call agent (outbound voicebot)
- D) Set up agent coaching/LMS
- E) Troubleshoot transcription accuracy
- F) Integration with my CCaaS or CRM
- G) Compare Convin to alternatives
- H) Other
Your contact center setup?
- A) Small (<50 agents)
- B) Mid-size (50-350 agents)
- C) Large (350+ agents)
Current CCaaS?
- A) Amazon Connect
- B) Avaya
- C) 8x8
- D) Bright Pattern
- E) Aircall
- F) Other / not sure

Skip-ahead rule: if the user's prompt already contains enough context, skip to Step 2.

若

references/learnings.md

文件存在，请先阅读该文件获取已积累的平台相关知识。

你的需求是什么？
- A) 设置自动化QA评分
- B) 配置座席实时辅助功能
- C) 部署AI外呼电话机器人
- D) 设置座席辅导/LMS系统
- E) 排查转录准确性问题
- F) 与我的CCaaS或CRM系统集成
- G) 对比Convin与竞品
- H) 其他需求
你的呼叫中心规模？
- A) 小型（少于50名座席）
- B) 中型（50-350名座席）
- C) 大型（350名以上座席）
当前使用的CCaaS平台？
- A) Amazon Connect
- B) Avaya
- C) 8x8
- D) Bright Pattern
- E) Aircall
- F) 其他/不确定

跳转规则：如果用户的提问已包含足够上下文信息，直接进入步骤2。

Step 2 — Route or answer directly

步骤2 —— 转至对应模块或直接解答

Problem domain	Route to
Choosing between CCaaS platforms	`/sales-ccaas-selection {user's question}`
Building a coaching program across tools	`/sales-coaching {user's question}`
Comparing note-taker / CI platforms	`/sales-note-taker {user's question}`
Reviewing a specific call for coaching	`/sales-call-review {user's question}`
Convin vs Observe.AI vs Cresta vs Balto	Answer directly — read `references/platform-guide.md` for competitive positioning

问题领域	转至路径
CCaaS平台选型	`/sales-ccaas-selection {用户问题}`
跨工具构建辅导体系	`/sales-coaching {用户问题}`
对比笔记工具/对话智能平台	`/sales-note-taker {用户问题}`
针对特定呼叫开展辅导评审	`/sales-call-review {用户问题}`
Convin与Observe.AI/Cresta/Balto对比	直接解答 —— 阅读 `references/platform-guide.md` 获取竞品定位信息

Step 3 — Convin platform reference

步骤3 —— Convin平台参考资料

Read
references/platform-guide.md
for the full platform reference — modules, integrations, competitive positioning, workflow setup.

Answer the user's question using only the relevant section. Don't dump the full reference.

**阅读

references/platform-guide.md

**获取完整平台参考信息，包括模块功能、集成能力、竞品定位、工作流设置等。

仅使用相关章节内容解答用户问题，不要直接输出全部参考资料。

Step 4 — Actionable guidance

步骤4 —— 可落地指导建议

Focus on the user's specific situation.

QA setup priority: Start with 3-5 high-impact evaluation criteria, not 20. Convin's AI scoring works best with clear, binary criteria ("Did the agent verify the customer's identity?") rather than subjective ones ("Was the agent empathetic?"). Expand criteria after calibrating AI scores against manual QA for 2-4 weeks.

Transcription accuracy: If accuracy is poor with accents or mixed languages, request a custom model tuning from Convin support — their proprietary LLM can be fine-tuned per customer. Budget 2-3 weeks for tuning.

Real-Time Assist rollout: Pilot with 10-20 agents on a single queue before rolling out broadly. Monitor for "alert fatigue" — if agents start ignoring prompts, reduce trigger frequency and focus on highest-value moments (compliance, upsell, escalation).

If you discover a gotcha, workaround, or tip not covered in

references/learnings.md

, append it there.

聚焦用户的具体场景提供建议。

QA设置优先级：先从3-5个高影响力的评估标准入手，而非20个。Convin的AI评分在清晰的二元标准（如“座席是否验证了客户身份？”）下表现最佳，而非主观标准（如“座席是否表现出同理心？”）。在AI评分与人工QA校准2-4周后，再逐步扩展评估标准。

转录准确性：如果带口音或混合语言的语音转录准确性较差，请向Convin支持团队申请自定义模型调优——他们的专有LLM可针对客户需求进行微调。调优周期约为2-3周。

实时辅助功能推广：先在单个队列中选取10-20名座席进行试点，再全面推广。注意监控“提示疲劳”问题——如果座席开始忽略提示，需降低触发频率，聚焦高价值场景（合规要求、 Upsell、升级处理）。

如果发现

references/learnings.md

未覆盖的问题、临时解决方案或技巧，请将其补充至该文件中。

Gotchas

注意事项

Best-effort from research — review these, especially items about integrations and feature availability that may be outdated.

No public API. All integrations go through Convin's pre-built connectors (30+ across CCaaS, CRM, dialers, HR, BI). Custom integrations have a ~3-day turnaround from Convin's team, but you can't build your own.
Transcription struggles with speaker separation. Multiple G2/Capterra reviewers report the AI sometimes can't distinguish agent from customer voices, especially on poor-quality recordings. Verify diarization accuracy in your pilot before scaling QA.
Dashboard delays. Calls don't always surface immediately — expect occasional delays of hours. Don't build real-time alerting workflows that depend on instant call availability.
AI scoring lacks explanations. When the AI marks a criterion as "No," it doesn't always explain why. Supervisors may need to listen to the call to understand the score, which partially defeats the purpose of automated QA.
Steep initial learning curve. Plan 2-4 weeks of admin training. The platform has many modules and configuration isn't intuitive out of the box.
Custom pricing only. No self-serve plans — you must go through sales. A free tier exists but with limited features.
Self-improving: If you discover something not covered here, append it to
```
references/learnings.md
```
with today's date.

基于研究的最佳实践——请重点关注集成和功能可用性相关内容，部分信息可能已过时。

无公开API：所有集成需通过Convin的预构建连接器完成（覆盖30+ CCaaS、CRM、拨号器、HR、BI平台）。自定义集成需Convin团队处理，周转时间约3天，用户无法自行构建。
说话人分离存在困难：多个G2/Capterra评测者反馈，AI有时无法区分座席与客户的声音，尤其是在录音质量较差的情况下。在大规模推广QA前，请先在试点中验证说话人分离的准确性。
仪表盘数据延迟：呼叫数据并非总是即时显示——偶尔会出现数小时的延迟。请勿构建依赖呼叫数据即时可用的实时告警工作流。
AI评分缺乏解释：当AI标记某一标准为“未达标”时，并非总能说明原因。主管可能需要收听呼叫录音才能理解评分，这在一定程度上削弱了自动化QA的意义。
初始学习曲线较陡：需规划2-4周的管理员培训时间。该平台包含多个模块，初始配置不够直观。
仅支持自定义定价：无自助服务套餐——必须通过销售团队购买。提供免费版本，但功能受限。
自我更新：如果发现本文未覆盖的内容，请将其补充至
```
references/learnings.md
```
并标注日期。

Before recommending a specific platform skill

Related skills

Examples

示例

Example 1: Setting up automated QA

示例1：设置自动化QA

User says: "I want Convin to score 100% of our support calls automatically" Skill does:

Reads platform guide for QA module details
Recommends starting with 3-5 binary criteria (identity verification, greeting script, issue resolution confirmation)
Explains calibration: run AI scoring alongside manual QA for 2-4 weeks, compare scores, adjust criteria
Warns about the AI explanation gap — supervisors need a plan for reviewing disagreements Result: Phased QA automation rollout plan with calibration timeline

用户提问：“我希望Convin自动为我们所有的支持呼叫打分” 技能执行流程：

阅读平台指南中QA模块的详细内容
建议从3-5个二元标准（身份验证、问候脚本、问题解决确认）入手
解释校准流程：AI评分与人工QA并行运行2-4周，对比分数并调整标准
提醒AI评分缺乏解释的问题——主管需制定评分分歧的评审方案结果：包含校准时间线的分阶段QA自动化推广计划

Example 2: Comparing Convin to Observe.AI

示例2：对比Convin与Observe.AI

User says: "We're evaluating Convin vs Observe.AI for our 200-agent contact center" Skill does:

Reads platform guide competitive section
Compares: Convin is India-based with strong APAC support and potentially lower pricing; Observe.AI has deeper virtual agent capabilities and US enterprise presence
Notes: Both offer 100% automated QA and real-time assist. Convin adds AI Phone Call agent (outbound voicebot); Observe.AI has VoiceAI/ChatAI agents
Recommends evaluating based on CCaaS compatibility, transcription accuracy for your languages, and pricing Result: Side-by-side comparison with decision criteria specific to their setup

用户提问：“我们正在为200人规模的呼叫中心评估Convin和Observe.AI” 技能执行流程：

阅读平台指南中的竞品章节
对比分析：Convin总部位于印度，亚太地区支持能力强，定价可能更低；Observe.AI虚拟机器人功能更深入，在美国企业市场布局广泛
说明：两者均提供100%自动化QA和实时辅助功能。Convin额外提供AI外呼电话机器人；Observe.AI拥有VoiceAI/ChatAI机器人
建议根据CCaaS兼容性、目标语言的转录准确性及定价进行评估结果：针对用户场景的竞品对比及决策标准

Example 3: Troubleshooting transcription accuracy

示例3：排查转录准确性问题

User says: "Convin transcription is missing what our agents say, especially with Indian accents" Skill does:

Identifies known speaker diarization issue from Gotchas
Recommends requesting custom model tuning from Convin support (2-3 week process)
Suggests checking audio quality at source — poor VoIP quality amplifies transcription issues
Notes that Convin's proprietary LLM supports 70+ languages and can be fine-tuned per customer Result: Actionable fix path with expected timeline

用户提问：“Convin的转录无法识别我们座席的说话内容，尤其是印度口音” 技能执行流程：

识别注意事项中提到的说话人分离问题
建议向Convin支持团队申请自定义模型调优（周期2-3周）
建议检查源头音频质量——VoIP质量差会加剧转录问题
说明Convin的专有LLM支持70+语言，可针对客户需求微调结果：包含预期时间线的可落地修复方案

Troubleshooting

故障排查

Transcription not capturing agent voice

转录未捕获座席语音

Symptom: Transcript shows customer speech but agent words are missing or garbled Cause: Speaker diarization struggles with certain audio configurations (mono recordings, low bitrate, echo) Solution: Ensure your CCaaS exports stereo recordings with separate agent/customer channels. If mono-only, request custom model tuning from Convin support. Check that microphone levels aren't drastically different between agent and customer.

症状：转录内容仅显示客户语音，座席语音缺失或模糊原因：说话人分离功能在特定音频配置（单声道录音、低比特率、回声）下表现不佳 解决方案：确保你的CCaaS导出带有座席/客户独立声道的立体声录音。若仅支持单声道，请向Convin支持团队申请自定义模型调优。检查座席与客户的麦克风音量差异是否过大。

QA audits hanging or not showing results

QA审核停滞或无结果显示

Symptom: Audit tab shows loading spinner or blank results Cause: Processing delays during high-volume periods, or scoring rules still in draft state Solution: Verify scoring rules are published (not draft). Check if the issue is systemic (all calls) or specific to certain queues. For persistent hanging, escalate to Convin support — they can check backend processing queues.

症状：审核页面显示加载动画或空白结果原因：高流量时段的处理延迟，或评分规则仍处于草稿状态 解决方案：确认评分规则已发布（非草稿状态）。检查问题是系统性的（所有呼叫）还是特定队列的。若问题持续存在，请联系Convin支持团队——他们可检查后端处理队列。

AI scoring disagrees with manual QA on subjective criteria

AI评分与人工QA在主观标准上存在分歧

Symptom: AI marks "empathy" or "tone" criteria differently than human evaluators Cause: Subjective criteria are inherently harder for AI to score consistently Solution: Replace subjective criteria with observable behaviors: instead of "Was the agent empathetic?", use "Did the agent acknowledge the customer's frustration?" or "Did the agent use the customer's name?". Binary, observable criteria produce 80%+ AI-human agreement; subjective ones produce 50-60%.

症状：AI对“同理心”或“语气”等标准的评分与人工评估者不同原因：主观标准本身难以让AI持续准确评分 解决方案：将主观标准替换为可观察的行为：例如将“座席是否表现出同理心？”改为“座席是否认可客户的不满？”或“座席是否使用了客户的姓名？”。二元、可观察的标准可实现80%以上的AI与人工评分一致性；主观标准的一致性仅为50-60%。