AI in a professional environment 专业环境中的人工智能
One week. One product. Three integration concepts. 一周。一款产品。三个整合概念。
During a one-week workshop, guided by the team at Upgrade agency, we were tasked with researching and proposing three ways to integrate AI into the software of Cegedim, a publisher of management software for both private and public sector companies. 在为期一周的工作坊中，在 Upgrade 机构团队的指导下，我们的任务是研究并提出三种将 AI 融入 Cegedim 软件的方式——Cegedim 是一家面向私营和公共部门企业的管理软件出版商。
The product in focus: Upharma, Cegedim's pharmacy management software. Our mission: make AI an ideal colleague. 聚焦产品：Upharma，Cegedim 的药店管理软件。我们的使命：让 AI 成为理想的同事。
Problem Statement 问题陈述
"How can AI help Cegedim software users make better decisions and take more effective action?" 「AI 如何帮助 Cegedim 软件用户做出更好的决策并采取更有效的行动？」
The challenge wasn't to add AI for the sake of it, but to integrate it where it creates real value for professionals, without disrupting their existing workflows or generating distrust toward the tool. 挑战不在于为了 AI 而添加 AI——而是要将其融入真正为专业人士创造价值的场景，同时不干扰现有工作流程，也不引发对工具的不信任。
The stakes of AI in a professional context 专业情境中 AI 的关键挑战
Four constraints that shaped every design decision. 塑造每一个设计决策的四大约束条件。
Challenge 01 挑战 01
Frictionless adoption 无摩擦采用
Avoid resistance toward a tool perceived as intrusive or destabilizing in an already established workflow. 避免对被视为具有侵入性或破坏现有工作流程的工具产生抵触情绪。
Challenge 02 挑战 02
Multi-profile users 多样化用户群体
Manager, department head, seasonal employee: very different needs, skill levels, and degrees of autonomy. 管理者、部门负责人、季节性员工：需求、技能水平和自主程度差异显著。
Challenge 03 挑战 03
One-week constraint 一周约束
Deliver 3 concrete, argued, client-ready proposals in 5 days of intensive work. 在 5 天高强度工作中，交付 3 个具体、有充分论证且可提交给客户的方案。
Challenge 04 挑战 04
Complex business software 复杂的企业软件
Operate within a dense software environment (stock, billing, HR, purchase flows) with an AI that remains discreet and non-blocking. 在密集的软件环境中（库存、计费、HR、采购流程）运作，同时确保 AI 保持低调且不阻断流程。
Guiding Principle 指导原则
"The AI only takes up the space it's given. The user always remains the origin of every decision." 「AI 只占用被赋予它的空间。用户始终是每个决策的原点。」
From research to proposal in one week 一周内从研究到方案
Four steps, two personas, three deliverables. 四个步骤，两个用户画像，三份交付成果。
AI capability research AI 能力研究
Three families of applicable AI were identified: prediction (anticipating needs from data), generation (producing content and analyses), automation (reducing repetitive tasks). Interaction modes selected: button, questionnaire, notifications and reminders, intentionally non-intrusive. 识别出三类可应用的 AI 能力：预测（从数据中预见需求）、生成（产出内容和分析）、自动化（减少重复性任务）。选定的交互方式：按钮、问卷、通知与提醒——刻意保持非侵入性。
Persona construction 用户画像构建
Two user profiles guided the entire project, representing the breadth of Upharma's user base. 两个用户画像贯穿整个项目——代表 Upharma 用户群体的广度。
Florence
Manager · 50 管理者 · 50 岁
Goal 目标
Efficiently manage orders, finances, and disputes 高效管理订单、财务和纠纷
Pain point 痛点
No visibility on future stock needs; manual management of supplier disputes 对未来库存需求缺乏可见性；人工处理供应商纠纷
Jordan
Department Head · 33 部门负责人 · 33 岁
Goal 目标
Ensure optimal stock levels; track sales and promotions 确保最优库存水平；跟踪销售和促销活动
Pain point 痛点
Interface too rigid; limited access to real-time sales data 界面过于僵硬；获取实时销售数据受限
Ideation across 4 domains 四大领域的创意生成
From the personas and research, ideas were generated across four functional areas of Upharma: HR management, billing, stock, and purchase & sales flows. These explorations converged into the 3 final solution axes. 结合用户画像和研究成果，围绕 Upharma 的四大功能领域展开创意生成：HR 管理、计费、库存以及采购与销售流程。这些探索最终汇聚为 3 个解决方案轴线。
AI representation & interaction design AI 形象与交互设计
The AI was represented as a dark blue sphere: discreet enough not to interfere in a professional interface, yet visible enough to be quickly spotted. Stars symbolize artificial intelligence, making it easy to identify at a glance. AI 被呈现为深蓝色球体——低调到不干扰专业界面，同时又足够显眼以便快速识别。星形符号象征人工智能，使其一眼便可辨认。
Solutions 解决方案
3 AI integration axes 3 个 AI 整合轴线
Axis 01 轴线 01
Order Management Optimization 订单管理优化
Problem 问题
Anticipate ordering needs, avoid stock shortages, and manage supplier disputes: currently a manual, reactive process with no predictive support. 预测订购需求、避免库存短缺并管理供应商纠纷——目前是一个手动、被动的流程，没有预测性支持。
AI Solution AI 解决方案
- Analysis of past sales to anticipate future needs 分析历史销售数据以预见未来需求
- Automatic order suggestions with AI reasoning displayed: the user can always see why the AI made a suggestion 自动订单建议并显示 AI 推理过程——用户始终可以看到 AI 提出建议的原因
- Dispute management and payment reminders 纠纷管理和付款提醒
AI / User Interaction AI / 用户交互
- Notification when products are at risk of stock-out 产品面临缺货风险时发出通知
- Visual interface proposing suggested orders with justification 可视化界面提出带理由的订单建议
- For each suggestion, the AI displays a detailed explanation of its reasoning, keeping the human as the final decision-maker 对每个建议，AI 都会显示其推理的详细说明——确保人类是最终决策者
Axis 02 轴线 02
AI-Personalized Dashboard AI 个性化仪表盘
Problem 问题
Each employee has different needs (manager, department head, seasonal). Creating custom dashboard components is currently a manual process that must go through Cegedim directly. 每位员工有不同需求（管理者、部门负责人、季节性员工）。创建自定义仪表盘组件目前是一个必须直接通过 Cegedim 操作的手动流程。
AI Solution AI 解决方案
- The AI proposes the most relevant interface elements based on the user's profile AI 根据用户画像提出最相关的界面元素
- It adapts the layout and data presentation to usage habits 根据使用习惯调整布局和数据呈现
- An integrated chat allows the user to correct or refine the AI's output in real time 内置聊天功能让用户可以实时修正或优化 AI 的输出
AI / User Interaction AI / 用户交互
- Dynamic interface adaptation on login 登录时动态调整界面
- Integrated chat for real-time customization 集成聊天实现实时个性化
- The AI proposes customization axes and a layout suggestion: the user stays in control AI 提出定制化方向和布局建议：用户始终掌控全局
Axis 03 轴线 03
At-Risk Product Detection & Targeted Promotions 风险产品检测与精准促销
Problem 问题
Managing unsold items and perishable products before they become a financial loss for the pharmacy. 在未售出商品和易腐产品造成药店财务损失之前进行管理。
AI Solution AI 解决方案
- Proactive detection of at-risk products 主动检测风险产品
- Criticality scoring dashboard for real-time visibility 紧急程度评分仪表盘，实现实时可见性
- Tailored discount suggestions based on criticality level 基于紧急程度的定制折扣建议
- Generation of targeted promotional campaigns with justification 生成带理由的精准促销活动
AI / User Interaction AI / 用户交互
- Dashboard with product criticality scores 显示产品紧急程度评分的仪表盘
- Notifications alerting on at-risk products 风险产品警报通知
- AI proposes solutions with reasoning: user approves AI 提出带推理的解决方案：用户审批确认
Key Learnings 关键收获
- Designing AI for professional software in one week forced me to prioritize ruthlessly. Every decision had to be justified by a clear user need, with no time to explore every direction. 在一周内为专业软件设计AI功能，迫使我做出严格的优先级判断，每个决策都必须有明确的用户需求支撑。
- AI integration in B2B tools must be discreet first and powerful second. The biggest risk is AI that disrupts a professional's workflow and erodes trust, not AI that underperforms. B2B工具中的AI集成必须首先做到克制，其次才是强大。最大的风险是打乱专业人员的工作流程，而非AI表现不佳。
- Presenting to a real client sharpened my ability to argue design decisions. Having tight constraints made every choice feel more deliberate and defensible. 向真实客户汇报让我更擅长为设计决策进行论证，严格的约束反而让每个选择更有说服力。
