人工智能ai以算法为基础
执行摘要 (Executive Summary)
Most companies fail to scale up their AI pilot projects. Most AI initiatives have their budgets cut because they don’t deliver results quickly enough.
大多数公司未能扩大其AI试点项目。 大多数AI计划都削减了预算，因为它们没有足够快地交付结果。
The problem is not technology or talent. The culprit is often a corporate culture and organizational structure designed for a pre-AI age. To deploy AI at scale, executives must build a culture where business and technical teams can collaborate seamlessly.
问题不在于技术或人才。 罪魁祸首通常是为AI前时代设计的企业文化和组织结构。 为了大规模部署AI，高管必须建立一种文化，业务和技术团队可以无缝协作。
What follows is a 10-part framework to help executives prepare their company for AI implementation at scale. This is not a sequential list for companies to progress along. Most companies will be proficient at one or more parts of this framework.
接下来是一个由十部分组成的框架，可帮助高管为大规模实施AI的实施做好准备。 这不是公司前进的顺序列表。 大多数公司将精通此框架的一个或多个部分。
Instead, this executive guide describes the necessary conditions for a culture of successful AI implementation. Business leaders can use this framework to diagnose their company’s proficiency in each of these areas.
相反，本执行指南描述了成功实施AI的文化的必要条件。 企业领导者可以使用此框架来诊断其公司在这些领域中的熟练程度。
And then, they can build.
然后，他们可以建造。
人工智能转型框架 (The AI Transformation Framework)
Almost every CEO says they are ‘doing AI’, just like your friend tells you that he is ‘hitting the gym’.
几乎每个CEO都说他们在“做AI”，就像您的朋友告诉您他在“锻炼健身房”一样。
Both of these people probably won’t get the results they hoped for. Many companies won’t go beyond AI pilot projects and your friend will tell you that he’s too busy to exercise.
这两个人可能都无法获得他们期望的结果。 许多公司不会超出AI试点项目的范围，您的朋友会告诉您他太忙了，无法锻炼。
The Gartner CIO survey in 2019 claimed that 37% of 3,000 surveyed companies have deployed AI. It is a safe bet that many of the remaining 63% have hit a snag or two in their AI journey.
Gartner CIO在2019年的调查称 ，在3,000家被调查公司中，有37％已部署了AI。 可以肯定的是，其余63％的许多人在AI之旅中遇到了一两次障碍。
Technology and talent are not enough. Firms must break down cultural barriers and rethink organizational structures to deploy AI across departments and geographies.
技术和人才不足。 公司必须打破文化障碍，重新考虑组织结构，以便在部门和地区之间部署AI。
There is no single blueprint for enterprise AI deployment. How firms approach AI implementation depends on their size, talent pool, and AI maturity.
没有用于企业AI部署的单一蓝图。 公司如何实施AI实施取决于其规模，人才库和AI成熟度。
Executives can, however, use this 10-part framework to guide them on their AI transformation journey:
高管可以使用这个包含10个部分的框架来指导他们进行AI转型：
- Aim to deploy AI at scale 旨在大规模部署AI
- Build AI awareness throughout the firm 在整个公司内建立AI意识
- Commit to a ‘AI transformation vision’ at the C-Suite level 致力于C-Suite级别的“ AI转型愿景”
- Plan a portfolio of AI projects 规划AI项目组合
- Build an in-house AI team and partner with AI vendors 建立内部AI团队并与AI供应商合作
- Distribute AI talent across the firm and assign responsibilities 在公司范围内分配AI人才并分配职责
- Embrace data-driven decision making throughout the firm 拥抱整个公司的数据驱动型决策
- Break down data silos 打破数据孤岛
- Bridge the gap between business & technical teams 弥合业务与技术团队之间的鸿沟
- Budget for integration and change management 整合和变更管理预算
一，旨在大规模部署AI (I. Aim to Deploy AI at Scale)
AI enables businesses to classify data, find patterns, predict outcomes and make repeated decisions at scale.
人工智能使企业能够对数据进行分类，查找模式，预测结果并进行大规模重复决策。
Scale matters. It is relatively easy for a bank to deploy a machine learning tool for customer segmentation to help with cross-selling and up-selling. It is more challenging —and far more rewarding — to deploy a suite of AI solutions to optimize the entire customer journey from on-boarding to ongoing relationship management.
规模很重要。 对于银行而言，部署机器学习工具进行客户细分相对容易，以帮助交叉销售和向上销售。 部署一套AI解决方案来优化从入职到进行中的客户关系管理的整个客户旅程具有更大的挑战性和更大的收获。
How can firms build and deploy a portfolio of scalable AI solutions? It comes down to organizational structure and culture. Firms must promote collaboration across business and tech teams so that AI solutions can accommodate evolving business needs. Organizational structures must also be fluid enough to allow AI talent to be deployed where it is needed most.
公司如何构建和部署可扩展的AI解决方案组合？ 它取决于组织结构和文化。 公司必须促进业务和技术团队之间的协作，以便AI解决方案可以适应不断变化的业务需求。 组织结构还必须足够灵活，以允许将AI人才部署到最需要的地方。
二。 在整个公司内建立AI意识 (II. Build AI Awareness Throughout the Firm)
Firms must increase AI awareness throughout the organization. Everyone from the C-Suite to individual contributors must be aware of how AI can solve business problems and how they can work with AI tools.
公司必须提高整个组织的AI意识。 从C-Suite到个人贡献者，每个人都必须意识到AI如何解决业务问题以及它们如何与AI工具一起工作。
AI education can be internal or external. Companies with high AI maturity will set up internal AI academies and on-the-job training programs. Other companies can use external trainers and consultants for classroom sessions and workshops.
AI教育可以是内部的或外部的。 人工智能成熟度较高的公司将建立内部人工智能学院和在职培训计划。 其他公司可以使用外部培训师和顾问来进行课堂会议和讲习班。
高级管理人员 (Senior Executives)
C-level executives and senior managers will already have a deep understanding of their company’s business needs, goals and challenges. Therefore, they must build AI awareness so that they can:
高层管理人员和高级经理已经对他们公司的业务需求，目标和挑战有了深刻的了解。 因此，他们必须建立AI意识，以便他们能够：
- Gain a high-level understanding of how AI technologies work (e.g. machine learning, machine vision, natural language processing) 深入了解AI技术的工作方式(例如，机器学习，机器视觉，自然语言处理)
- Identify high-value AI use cases within their industry and company 识别其行业和公司内的高价值AI用例
- Recognize specific AI tools & techniques their firm can use to achieve business needs 认识到公司可以用来满足业务需求的特定AI工具和技术
- Learn to prioritize AI initiatives 学习确定AI计划的优先级
- Understand barriers to adoption, the impact on people’s roles, and the cultural changes required 了解采用障碍，对人们角色的影响以及所需的文化变革
技术人员 (Technical Staff)
Firms must invest in technical training for their data scientists, AI engineers, and those who build AI tools. Depending on their roles, their training can include:
企业必须为其数据科学家，AI工程师以及构建AI工具的人员投资技术培训。 根据他们的角色，他们的培训可以包括：
- Data best practices (e.g. collection, cleaning, governance, fixing bias) 数据最佳做法(例如，收集，清理，治理，纠正偏差)
- Technical understanding of machine learning and deep learning 对机器学习和深度学习的技术了解
- Understanding of open source and third party tools for building & training AI and data models (e.g. Python, PyTorch, TensorFlow) 了解用于构建和培训AI和数据模型(例如Python，PyTorch，TensorFlow)的开源和第三方工具
- Awareness of industry standard and emerging AI techniques 对行业标准和新兴AI技术的意识
商务翻译 (Business Translators)
Also known as analytics translators, this emerging role will bridge the business and tech teams to ensure that AI products satisfy business needs. Business translators may also manage technical staff that build AI tools and take ownership of AI project implementation and adoption.
这种新兴角色也称为分析翻译，将桥接业务和技术团队，以确保AI产品满足业务需求。 业务翻译人员还可以管理构建AI工具并拥有AI项目实施和采用所有权的技术人员。
Business translators usually come from the business side (e.g. project managers, business analysts, subject matter experts, business unit managers). They will already have a sound understanding of the business and may also be skilled in project management, people management, or strategic planning.
业务翻译通常来自业务方面(例如，项目经理，业务分析师，主题专家，业务部门经理)。 他们已经对业务有很好的了解，并且可能还精通项目管理，人员管理或战略规划。
Business translators will need fundamental technical training and AI awareness in order to:
商务翻译将需要基本的技术培训和AI意识，以便：
- Communicate business needs & requirements in technical terms to data scientists and engineers building AI tools 在技术上与构建AI工具的数据科学家和工程师交流业务需求和要求
- Apply analytical approaches & AI tools to business problems 将分析方法和AI工具应用于业务问题
- Develop AI use cases in granular detail 详细开发AI用例
- Understand how deploying AI tools will change workflows 了解部署AI工具如何改变工作流程
商业用户 (Business Users)
These are the end users of AI tools in marketing, finance, sales, or other functions. These employees need training on how to use AI tools in their day-to-day jobs. They also need to overcome any fear of AI.
这些是市场营销，财务，销售或其他功能中AI工具的最终用户。 这些员工需要接受有关如何在日常工作中使用AI工具的培训。 他们还需要克服对AI的任何恐惧。
Most people are afraid that AI and automation will take away their jobs. Managers may feel offended if executives value machines over their own people’s skill and experience. Executives must convince employees that AI will help them achieve more. They must tell a compelling story about why AI is critical and how it will benefit the company and its people.
大多数人担心AI和自动化会抢走他们的工作。 如果高管们对机器的评价超过了他们本人的技能和经验，那么他们可能会感到冒犯。 高管们必须说服员工人工智能将帮助他们实现更多成就。 他们必须讲述一个引人入胜的故事，说明为什么AI至关重要，以及AI如何使公司及其员工受益。
Crucially, executives must convince employees that humans will always be the most important part of the equation. While AI can generate data-driven insights and automate processes, only humans have the common sense and functional knowledge to apply these insights. Think of AI as augmented intelligence instead of artificial intelligence.
至关重要的是，高管们必须说服员工，人永远是最重要的部分。 虽然AI可以生成数据驱动的见解并实现流程自动化，但只有人类具有常识和功能知识才能应用这些见解。 将AI视为增强智能而非人工智能 。
Realistically, there will be some job loss due to automation. Jobs that involve routine and repetitive tasks are at most risk. However, headlines screaming about how one-third of jobs will be lost to automation don’t tell the whole story. AI automates tasks and not entire jobs. The real story is about AI augmentation of employees, not large-scale AI substitution.
实际上，由于自动化，会有一些工作上的损失。 涉及例行和重复性任务的工作风险最大。 但是，有关如何将三分之一的工作机会流失到自动化的头条新闻并没有说明全部。 AI自动执行任务，而不是整个工作。 真实的故事是关于员工的AI 扩充 ，而不是大规模的AI替代。
Employees should welcome AI augmentation. Once mundane tasks are automated, employees can devote more time to impactful and fulfilling work. Employers want this too. Deloitte’s AI in the Enterprise survey of 1,900 AI adopters revealed that a leading benefit of AI was that it freed up employees to be more creative.
员工应该欢迎AI增强。 一旦完成了平凡的任务，员工就可以将更多的时间投入到富有影响力的工作上。 雇主也希望如此。 德勤在1,900名采用AI的企业中进行的AI调查显示，AI的主要优势在于它可以释放员工的创造力 。
Employees are more likely to embrace AI if they believe it is critical to growth (or survival). Retail executives need only point to the existential threats posed by Amazon and e-commerce. Explaining how AI can make the retailer more efficient and responsive is compelling. Underscoring the critical role that employees play and painting a picture of future success will bring employees fully on-board.
如果员工认为AI对增长(或生存)至关重要，那么他们就更可能接受AI。 零售业高管只需要指出亚马逊和电子商务带来的生存威胁。 令人信服的是，解释人工智能如何使零售商更有效，更及时地响应。 强调员工扮演的关键角色并描绘未来的成功图景，将使员工充分发挥作用。
三， 致力于在C-Suite级别实现AI转型愿景 (III. Commit to an AI Transformation Vision at the C-Suite Level)
A company’s odds of AI success increase when its executives can explain their AI transformation vision. A transformation vision is not about individual use cases. It is about winning big in the market.
当公司高管可以解释其AI转型愿景时，公司获得AI成功的几率会增加。 转换愿景与单个用例无关。 这是关于在市场上赢得巨大的胜利。
Specifically, the C-Suite should have detailed answers to four questions:
特别是，C-Suite应该对四个问题有详细的答案：
- Which business challenges can AI help us with? 人工智能可以帮助我们解决哪些业务挑战？
- How will AI differentiate us from the competition in 3-5 years? 人工智能将如何在3-5年内使我们从竞争中脱颖而出？
- How will we use AI to grow and capture market share? 我们将如何使用AI来扩大和占领市场份额？
- What steps do we have to take today in terms of data availability, talent, and culture of innovation? 就数据可用性，人才和创新文化而言，我们今天必须采取什么步骤？
Consider an HVAC company that installs heating, ventilation and air conditioning equipment in office buildings. They see an opportunity to combine sensors with machine learning to manage temperatures throughout the building based on human activity. This enables the HVAC company’s clients to minimize energy usage. In this case, energy savings from AI adoption will differentiate the company and help it gain market share.
考虑一家在办公室建筑中安装供暖，通风和空调设备的HVAC公司。 他们看到了将传感器与机器学习相结合的机会，可以根据人类活动来管理整个建筑物的温度。 这使HVAC公司的客户能够最大程度地减少能耗。 在这种情况下， 采用AI节省的能源将使公司脱颖而出，并帮助其获得市场份额。
This is the start of the HVAC company’s AI transformation vision. Next, they should ask themselves how they are going to win by leveraging their data, talent, and culture.
这是HVAC公司AI转型愿景的开始。 接下来，他们应该问自己如何利用自己的数据，才能和文化来取胜。
An AI transformation vision should inform the company’s AI strategy and project portfolio. It should also help the firm prioritize AI projects.
人工智能转型愿景应为公司的人工智能战略和项目组合提供信息。 它还应帮助公司确定AI项目的优先级。
通过试点项目长期思考并寻求短期知识 (Think Long Term and Seek Short Term Knowledge with Pilot Projects)
A long term AI transformation vision will help executives commit to a multi-phase AI journey. Executives will realize that the real benefits will take time. Without a long-term vision, executives might pull the plug if they don’t see results quickly.
长期的AI转型愿景将帮助高管人员致力于多阶段的AI旅程。 高管们将意识到，真正的收益将需要时间。 没有长远的眼光，高管们如果看不到结果很快就会停下来。
Even successful AI projects may take time to generate ROI. Pilot projects may not yield a financial ROI at all. Still, if the pilot teaches the firm about how data infrastructures should be overhauled to enable large-scale AI adoption, this is a valuable quick win.
即使成功的AI项目也可能需要一些时间才能产生ROI。 试点项目可能根本不会产生财务投资回报。 尽管如此，如果飞行员向公司讲授如何彻底改革数据基础架构以实现大规模AI的采用，这将是一次宝贵的快速胜利。
Pilot projects don’t cost much and can give the firm valuable insights to build scalable AI solutions. They can reveal what type of data should be collected at higher volumes or detail and identify where the current data gaps are. This knowledge will help the firm develop core capabilities related to data collection and governance, for instance.
试点项目成本不高，可以为公司提供宝贵的见解，以构建可扩展的AI解决方案。 他们可以揭示应该以更高的数量或更多的细节收集哪些类型的数据，并确定当前的数据差距在哪里。 例如，这些知识将帮助公司发展与数据收集和治理相关的核心能力。
Executives that commit to a long term transformation vision are more likely to incentivize and encourage a culture of innovation where financial ROI isn’t the only measure of success. This mindset opens the door to greater financial ROI in the long term.
致力于长期转型愿景的高管更有可能激励和鼓励创新文化，而财务投资回报并不是成功的唯一衡量标准。 从长远来看，这种思维方式为更大的财务投资回报率打开了大门。
IV。 规划AI项目组合 (IV. Plan a Portfolio of AI projects)
Successful AI transformation is driven by a portfolio of projects with different time horizons.
成功的AI转型是由具有不同时间范围的项目组合驱动的。
Measurable benefits from larger, more ambitious AI initiatives may take years while incurring significant costs in the short term. Investing solely in large multi-year projects creates budgetary pressure and executive impatience.
更大，更雄心勃勃的AI计划带来的可衡量的收益可能需要数年时间，而短期内却会产生大量成本。 仅投资大型的多年项目会产生预算压力和行政上的不耐烦。
Firms should plan a portfolio of AI projects with different time horizons. This enables the firm to enjoy a steady stream of benefits from short term projects, which persuades executives to continue their support. A well-structured AI project portfolio consists of:
公司应该规划具有不同时间范围的AI项目组合。 这使公司可以从短期项目中获得稳定的收益，这说服了高管继续提供支持。 结构良好的AI项目组合包括：
- Small pilot projects that teach valuable lessons about how to scale AI 小型试点项目，教有关如何扩展AI的宝贵经验
- Short term projects with measurable returns in 6–12 months 短期项目，可在6-12个月内获得可衡量的收益
- Medium term projects that tackle increasingly valuable use cases and generate ROI over 12–24 months 中期项目可解决越来越有价值的用例并在12-24个月内产生ROI
- Long term projects that apply scalable AI throughout the enterprise 在整个企业中应用可扩展AI的长期项目
Suppose a bank has an AI transformation vision to ‘gain greater market share by streamline the entire customer journey with AI.’ Projects should then make it easier for customers to sign up, help the bank provide a convenient and tailored customer service, which in turn convinces more people to sign up.
假设一家银行具有AI转型的愿景，即“通过使用AI简化整个客户旅程来获得更大的市场份额”。 然后，项目应该使客户更容易注册，帮助银行提供方便且量身定制的客户服务，从而说服更多的人注册。
Pilot projects are mainly for learning and concept validation. Their value comes from showing the firm where they are now, and what they need in terms of data, talent, and infrastructure in order to successfully deploy AI.
试点项目主要用于学习和概念验证。 他们的价值来自向公司展示他们现在的位置，以及为成功部署AI而需要的数据，人才和基础设施。
Short term projects focus on generating ‘quick wins’ from single use cases. The bank might start with a project that automates Know Your Customer (KYC) processes during client on-boarding. AI-enabled automation tools are readily available and KYC processes are standardized, making this a shorter-term project that helps the bank cut costs and increase efficiency.
短期项目侧重于从单个用例中产生“快速获胜”。 银行可能会从一个在客户入职期间自动执行“了解您的客户”(KYC)流程的项目开始。 支持AI的自动化工具随时可用，并且KYC流程已标准化，这使其成为一个短期项目，可帮助银行削减成本并提高效率。
Medium term projects can focus on higher value use cases that need more time to generate returns. After automating KYC, our bank might undertake a project to build a customer segmentation tool using unsupervised machine learning. This tool would place customers into clusters based on behavior and characteristics, enabling the bank to cross-sell more effectively and increase revenue.
中期项目可以专注于需要更多时间来产生回报的高价值用例。 KYC自动化后，我们的银行可能会进行一个项目，以使用无监督的机器学习来构建客户细分工具。 该工具将根据行为和特征将客户置于集群中，从而使银行能够更有效地交叉销售并增加收入。
Long term projects provide the real value to the firm and its customers. These can be standalone projects or an initiative that combines smaller project implementations into a coherent solution. Our bank that wants to streamline the entire customer journey might create an app or web platform that handles customer on-boarding, provides tailored product recommendations, and provides customer service.
长期项目为公司及其客户提供了真正的价值。 这些可以是独立的项目，也可以是将较小的项目实现组合为一致的解决方案的计划。 我们希望简化整个客户流程的银行可能会创建一个应用程序或Web平台来处理客户入职，提供量身定制的产品推荐并提供客户服务。
A well-structured portfolio provides ROI in phases. In addition to the knowledge and insights from earlier phases, projects that generate ROI in phases can fund (and validate) future phases.
结构良好的投资组合可分阶段提供ROI。 除了早期阶段的知识和见识之外，分阶段产生ROI的项目还可以资助(并验证)未来阶段。
V.建立内部AI团队并与AI供应商合作 (V. Build an In-house AI Team and Partner with AI Vendors)
Companies should aim to build AI internally in the long run. In the short run (and for specific use cases), buying AI tools from vendors can yield immediate returns.
从长远来看，公司应该致力于内部构建AI。 在短期内(针对特定用例)，从供应商处购买AI工具可以立即获得回报。
The case for buying AI. Partnering with AI vendors might speed up one-off AI projects, especially if the company is early on in its AI journey. An AI vendor might have the perfect tool for a use case, which saves the company time. Vendor expertise can also shorten the AI learning curve for a new internal AI team.
购买AI的案例。 与AI供应商合作可能会加速一次性的AI项目，特别是如果该公司处于AI旅程的早期阶段。 一家AI供应商可能拥有一个用例的完美工具，从而节省了公司时间。 供应商的专业知识还可以缩短新的内部AI团队的AI学习曲线。
The case for building AI. Internally built AI tools are more likely to satisfy business needs and mesh with data and workflows. Relying on vendor products for multiple AI initiatives isn’t feasible in the long run. Vendors won’t be familiar with the company’s business needs, processes, and data. Off-the-shelf vendor tools may not integrate with the company’s data and business processes. Companies also can’t give vendors access to sensitive data. Crucially, building AI tools in-house allows the company to grow its AI capabilities and scale up.
构建AI的案例。 内置AI工具更可能满足业务需求，并与数据和工作流紧密结合。 从长远来看，依靠供应商的产品进行多种AI计划是不可行的。 供应商不会熟悉公司的业务需求，流程和数据。 现成的供应商工具可能未与公司的数据和业务流程集成。 公司也不能让供应商访问敏感数据。 至关重要的是，内部构建AI工具使公司能够扩展其AI功能并扩大规模。
The hybrid approach. Partnering with an AI vendor to build customized AI tools is an option where a tailored solution is urgently needed. When internal staff can tell the vendor exactly what the tool should do, the tool is more likely to integrate with the company’s processes and data. For instance, HSBC partnered with AI vendor Ayasdi to develop an AI-powered anti-money laundering tool. While HSBC undoubtedly has internal AI teams, they leveraged vendor expertise for quicker results.
混合方法。 与迫切需要定制解决方案的选项是与AI供应商合作以构建定制的AI工具。 当内部人员可以准确地告诉供应商该工具应该做什么时，该工具更有可能与公司的流程和数据集成。 例如，汇丰银行与 AI供应商Ayasdi合作开发了一种由AI驱动的反洗钱工具。 汇丰银行无疑拥有内部AI团队，但他们利用供应商的专业知识来获得更快的结果。
In addition to vendor partnerships, a good AI transformation vision calls for a centralized AI team that helps the entire company. This team will include data scientists, data engineers, machine learning engineers and AI product managers. Depending on the company’s organizational structure, the team may report to the CTO, CIO, Chief Data Officer, or even a Chief AI Officer.
除了与供应商之间的合作伙伴关系，良好的AI转型愿景还需要一个集中的AI团队来为整个公司提供帮助 。 该团队将包括数据科学家，数据工程师，机器学习工程师和AI产品经理。 根据公司的组织结构，团队可以向CTO，CIO，首席数据官，甚至首席AI官报告。
The internal AI team’s responsibilities will include:
内部AI团队的职责将包括：
- AI strategy & problem identification 人工智能策略与问题识别
- AI standards, and processes AI标准和流程
- Planning and executing a portfolio of AI projects 规划和执行AI项目组合
- Data & governance standards 数据与治理标准
VI。 在公司范围内分配AI人才并分配职责 (VI. Distribute AI Talent Across the Firm and Assign Responsibilities)
Which organizational model is best for deploying AI at scale? Where should AI talent reside within the firm? A Harvard Business Review feature on AI-powered organizations discusses three organizational models for scaling AI:
哪种组织模型最适合大规模部署AI？ 人工智能人才应在公司内的何处居住？ 关于人工智能驱动的组织的《哈佛商业评论》功能讨论了三种用于扩展人工智能的组织模型 ：
Centralized: Concentrate AI talent in a central Core (‘hub’) such as headquarters or regional offices
集中式 ：将AI人才集中在中央核心('hub')中，例如总部或区域办事处
Decentralized: Embed AI talent across business units (‘spokes’)
去中心化 ：将AI人才嵌入各个业务部门(“分支”)
Hybrid: Distribute AI talent & responsibilities across the Core and business units
混合型 ：在核心和业务部门之间分配AI人才和职责
Tasks related to AI strategy, projects, and adoption can be owned by any of three organizational layers: the Core, various business units, or a ‘Grey Area’ that works across the Core and business units.
与AI战略，项目和采用相关的任务可以由以下三个组织层中的任何一个负责：核心，各个业务部门或跨核心和业务部门工作的“灰色区域”。
核心 (The Core)
The Core is responsible for AI & data strategy, recruiting, governance, and partnering with AI & data vendors.
核心负责AI和数据战略，招聘，治理以及与AI和数据供应商的合作 。
The core creates AI standards, processes and best practices that help scale AI across the organization. This ensures that work isn’t duplicated by business units and that AI deployments are seamless and in line with company standards.
核心创建AI标准，流程和最佳实践，以帮助在整个组织范围内扩展AI。 这确保了业务部门不会重复工作，并且确保AI部署是无缝的并且符合公司标准。
The Core should be in charge of data initiatives such as data cleaning, labeling and integration. These initiatives should be implemented gradually in concert with AI projects. There is no need to spend millions on company-wide data gathering and cleaning before business needs and AI use cases are identified — these data initiatives might be abandoned if management realizes that they don’t fit AI project needs.
核心应负责数据计划，例如数据清理，标记和集成。 这些举措应与AI项目一起逐步实施。 在确定业务需求和AI用例之前，无需花费数百万美元在公司范围内收集和清理数据-如果管理层意识到它们不符合AI项目需求，则这些数据计划可能会被放弃。
业务部门 (Business Units)
Business units are responsible for adoption-related activities since they are the ultimate users of AI systems. These tasks include business analysis, encouraging adoption, training users, redesigning workflows, and measuring benefits.
业务部门负责与采用相关的活动，因为它们是AI系统的最终用户。 这些任务包括业务分析，鼓励采用，培训用户，重新设计工作流程以及衡量收益 。
Business units must be ultimately accountable for the success of AI products. Since AI tools are designed to address business needs, leaders from business units, such as a regional manager, should be accountable for the AI product’s success.
业务部门必须对AI产品的成功负最终责任。 由于AI工具旨在满足业务需求，因此业务部门的负责人(例如区域经理)应对AI产品的成功负责。
灰色地带 (The Grey Area)
Tasks that fall into the Grey Area can be owned by the Core or by individual business units. These tasks include project management, algorithm development, product design & testing, IT infrastructure, and change management.
属于灰色区域的任务可以由核心或各个业务部门拥有。 这些任务包括 项目管理，算法开发，产品设计和测试，IT基础架构以及变更管理。
The choice of whether the Core or the Business units take ownership of these tasks depends on three factors:
核心还是业务部门负责这些任务的选择取决于三个因素：
AI maturity: the firm’s previous experience in deploying AI
人工智能的成熟度 ：该公司过去部署人工智能的经验
AI urgency: the pace and complexity of AI initiatives
人工智能的紧迫性 ：人工智能计划的步伐和复杂性
Business model: the number of departments, functions, or geographies participating in AI adoption
业务模型：参与采用AI的部门，职能或地区的数量
Firms should centralize AI talent & operations in the Core if they have low AI maturity, high urgency, and a simple business model. The opposite conditions call for decentralizing AI talent throughout the business units.
如果人工智能成熟度低，紧迫性高且业务模式简单，则公司应将AI人才和运营集中在Core中。 相反的情况要求在整个业务部门中分散AI人才。
AI maturity. A company starting its AI journey might centralize its data & analytics executives, data scientists, AI engineers, and supporting staff in the Core. This enables faster development of standardized tools, data processes, repositories, and infrastructure. Of course, these personnel can be deployed to the business units as needed.
AI成熟度。 一家公司开始其AI之旅可能会将其数据和分析主管，数据科学家，AI工程师以及支持人员集中在Core中。 这样可以更快地开发标准化工具，数据流程，存储库和基础架构。 当然，这些人员可以根据需要部署到业务部门。
AI urgency. A company that needs to quickly deploy AI projects may choose to centralize its AI talent in the Core. This allows better monitoring of industry tech trends and easier coordination when building AI products.
AI的紧迫性。 需要快速部署AI项目的公司可以选择将其AI人才集中在Core中。 这样可以更好地监视行业技术趋势，并在构建AI产品时更轻松地进行协调。
Business model. AI tools must sometimes support a large number of business units, geographies, or functions. In this case, the company’s complex business model might convince executives to consolidate AI talent in the Core and assign them to other parts of the organization as needed.
商业模式。 AI工具有时必须支持大量业务部门，地域或功能。 在这种情况下，公司复杂的业务模型可能会说服高管将AI人才整合到Core中，并根据需要将其分配给组织的其他部门。
At the end of the day, this is more of an art than a science. A company that has to urgently deploy AI solutions and has a complex business model (suggesting centralization of AI talent) might have high AI maturity (which suggests decentralizing AI talent). In this case, executives should consider the relative importance of the three factors and determine whether AI talent will be most useful in the Core or in the business units.
归根结底，这更多的是艺术而不是科学。 一家必须紧急部署AI解决方案并具有复杂业务模型(建议AI人才集中化)的公司可能具有很高的AI成熟度(这意味着要分散AI人才)。 在这种情况下，高管应考虑这三个因素的相对重要性，并确定AI人才在核心还是业务部门中最有用。
Suppose an AI project in a bank’s portfolio involves building a KYC automation tool for a certain country. If the country’s client relations team has previously deployed AI tools for client on-boarding, the team can take ownership of activities usually handled by the Core such as business case analysis and project implementation.
假设银行投资组合中的AI项目涉及为特定国家/地区构建KYC自动化工具。 如果该国的客户关系团队以前已经为客户入职部署了AI工具，则该团队可以拥有通常由Core处理的活动的所有权，例如业务案例分析和项目实施。
七。 拥抱各级数据驱动的决策 (VII. Embrace Data-driven Decision Making at all Levels)
AI should to improve daily operations by empowering people with data insights. Since operations are carried out by people, the firm must adopt a culture of data-driven decision making from the C-Suite to the trenches.
AI应该通过赋予人们数据洞察力来改善日常运营。 由于操作是由人来执行的，因此公司必须采用从C-Suite到战es的数据驱动型决策文化。
When AI is adopted correctly, employees can augment their skills and judgement with algorithmic recommendations to achieve a better outcome than either humans or machines could on their own.
正确采用AI后，员工可以通过算法建议来增强自己的技能和判断力，从而获得比人类或机器自己更好的结果。
This can only happen if employees can trust AI tools and feel empowered to make decisions. Trust is established through AI awareness (described above). Empowerment to make decisions happens when firms abandon the traditional top-down approach.
这只有在员工可以信任AI工具并感到有能力的情况下才会发生 做出决定。 通过AI意识(如上所述)建立信任。 当企业放弃传统的自上而下的方法时，就可以做出决策。
Consider a national supermarket chain. Decisions about optimizing floor space and product placement are usually made by regional managers using historical data. For a supermarket chain with hundreds of stores, this type of top-down decision making may not result in the best outcomes for individual stores. In a data-driven decision making culture, local managers using an AI tool that tracks real-time in-store customer behavior are better placed to decide how a store displays products.
考虑一个全国连锁超市。 通常由区域经理使用历史数据来做出有关优化占地面积和产品放置的决策。 对于拥有数百家商店的连锁超市，这种自上而下的决策可能无法为单个商店带来最佳结果。 在以数据为依据的决策文化中，使用AI工具跟踪商店内实时顾客行为的本地经理可以更好地决定商店如何展示产品。
八。 打破数据孤岛 (VIII. Break Down Data Silos)
AI needs lots of data from many parts of the organization. Many corporate departments store data in silos — systems that don’t interface with each other and can only be accessed by specific teams. This is a barrier to AI adoption, but one that can be overcome.
人工智能需要来自组织许多部门的大量数据。 许多公司部门将数据存储在孤岛中，这些系统彼此之间不接口，只能由特定团队访问。 这是采用AI的障碍，但可以克服。
Large insurance companies are notorious for their data silos. Insurers tend to have dozens of standalone legacy (i.e. old) systems that are not connected to each other or to newer digital and cloud platforms. This isn’t conducive to AI and digital initiatives that the industry is embarking upon.
大型保险公司因数据孤岛而臭名昭著。 保险公司倾向于拥有数十个相互不连接或与更新的数字和云平台不相连的旧式(即旧)系统。 这不利于该行业正在着手进行的AI和数字计划。
Insurers, like most data-heavy industries, are investing in either modernizing legacy systems or migrating data to digital systems, data lakes, and data warehouses. Data lakes and data warehouses both store big data. A data lake is a huge pool of raw data without structure or labels. A data warehouse stores structured, labeled data for specific purposes.
像大多数数据密集型行业一样，保险公司正在投资对旧系统进行现代化改造或将数据迁移到数字系统，数据湖和数据仓库。 数据湖和数据仓库都存储大数据。 数据湖是没有结构或标签的大量原始数据池。 数据仓库存储用于特定目的的结构化标签数据。
Breaking down data silos isn’t accomplished overnight. It is generally a bad idea to invest in expensive, large-scale data transformation before implementing AI. It is better to do both together so that data transformation is done according to the needs of your AI initiatives.
打破数据孤岛并不是一朝一夕的事。 在实施AI之前，对昂贵的大规模数据转换进行投资通常是一个坏主意。 最好同时进行这两种操作，以便根据您的AI计划的需求完成数据转换。
AI pilot projects are helpful here — they reveal where the current data gaps are. With this knowledge, firms can start breaking down data silos intelligently.
人工智能试点项目在这里很有帮助-它们揭示了当前的数据差距在哪里。 有了这些知识，公司就可以开始智能地打破数据孤岛。
九。 弥合业务和技术团队之间的鸿沟 (IX. Bridge the Gap Between Business & Technical Teams)
Business translators, described earlier, ensure that AI and data science solutions are designed with business needs in mind.
商务翻译 ， 如前所述，确保在设计AI和数据科学解决方案时考虑到业务需求。
Business and technical teams don’t always speak the same language. A regional sales manager might know exactly what they want from an AI-powered customer segmentation tool — it should divide customers into buckets that represent which products they are interested in. However, the sales manager may not be able to communicate these requirements in technical terms to the data scientists or machine learning engineers actually building the tool.
业务和技术团队并不总是使用相同的语言。 区域销售经理可能会从基于AI的客户细分工具中确切知道他们想要什么-它应将客户划分为代表他们感兴趣的产品的存储桶。但是，销售经理可能无法用技术术语传达这些要求给实际构建工具的数据科学家或机器学习工程师。
This is not a new problem in the corporate world. Companies that deploy internal IT systems or customer-facing mobile apps give ownership of these projects to IT-focused project managers and business analysts. IT project managers, for instance, understand the business objectives of the new IT system. They will also have a basic understanding of the technology and can oversee the technical staff that build the system.
在企业界这不是一个新问题。 部署内部IT系统或面向客户的移动应用程序的公司将这些项目的所有权交给以IT为中心的项目经理和业务分析师。 例如，IT项目经理了解新IT系统的业务目标。 他们还将对技术有基本的了解，并可以监督构建系统的技术人员。
When it comes to AI initiatives, these business translators can be project managers, business analysts, or even internal consultants. They will need broad awareness of AI methods & capabilities so that they understand what the technical team is doing and give them direction.
关于AI计划，这些业务翻译可以是项目经理，业务分析师，甚至是内部顾问。 他们将需要对AI方法和功能的广泛了解，以便他们了解技术团队的工作并提供指导。
Business translators can use their awareness of business and AI to identify roadblocks to AI adoption. Early in the project, they can survey end-users, study workflows, speak with key stakeholders in the business and technical realms. This puts them in a position to diagnose and fix problems such as lack of employee buy-in or unreasonable expectations from end users.
商业翻译人员可以利用对商业和AI的了解来确定采用AI的障碍。 在项目的早期，他们可以调查最终用户，研究工作流程，并与业务和技术领域的主要利益相关者对话。 这使他们能够诊断和解决问题，例如缺乏员工支持或最终用户的期望不合理。
Identifying employees with business translator capabilities is essential. This role will soon be in high demand — and not too many people possess both AI awareness and business knowledge. The Deloitte AI in the Enterprise survey in 2019 found that business talent was almost as valued as AI talent, and even more so after companies implemented over 20 AI systems.
确定具有业务翻译能力的员工至关重要。 不久将对这个角色提出很高的要求-并没有太多人同时具备AI意识和业务知识。 德勤AI在2019年的企业调查中发现，业务人才的价值几乎与AI人才一样，在公司实施了20多个AI系统之后更是如此。
十。整合和变更管理预算 (X. Budget for Integration and Change Management)
AI awareness throughout the firm coupled with employee buy-in for AI initiatives lay the groundwork for AI adoption. However, these alone are not enough to ensure smooth AI integration with business processes. Firms must budget at least as much for adoption activities as they do for development.
整个公司的AI意识以及员工对AI计划的支持为AI的采用奠定了基础。 但是，仅凭这些还不足以确保AI与业务流程的顺利集成。 企业必须至少为采纳活动预算与发展预算相同的预算。
Integrating AI tools involves workflow redesign, training, and change management. These supporting activities should start well before deploying the AI solution. It helps prepare staff to work with the AI tool and avoid unpleasant surprises. It also ensures that staff are kept aware of, involved in, and supportive of, the AI journey.
集成AI工具涉及工作流程的重新设计，培训和变更管理。 这些支持活动应在部署AI解决方案之前良好地开始。 它可以帮助员工做好使用AI工具的准备，并避免令人不愉快的意外。 它还可以确保员工时刻了解，参与和支持AI之旅。
Starting early allows business translators and end-users to identify potential adoption issues before implementation. Perhaps the AI tool requires some workflow to be redesigned in a way that create more complications than benefits. Realizing this before deployment allows the technical team to modify the AI tool.
尽早开始，业务翻译和最终用户可以在实施之前确定潜在的采用问题。 也许AI工具需要重新设计一些工作流程，从而造成更多的复杂性而不是好处。 在部署之前意识到这一点，使技术团队可以修改AI工具。
外卖 (Takeaways)
AI is not easy. The ROI will take time. A company’s AI journey is defined by its unique needs and situation — there will be uncharted territory to cross.
人工智能并不容易。 投资回报率需要时间。 公司的AI历程由其独特的需求和情况来定义-将会跨越未知的领域。
Executives can prepare their companies for this journey by promoting a culture tailored to AI transformation. The 10-part AI transformation framework described above can help executives diagnose how the organization needs to change in order to deploy AI at scale.
高管可以通过倡导一种适合AI转型的文化为他们的公司为此旅程做准备。 上面描述的由10部分组成的AI转换框架可以帮助主管人员诊断组织如何进行更改以大规模部署AI。
Scaling up AI takes time and knowing what to do is only the first step. A strong culture, AI awareness, and employee buy-in at all levels will be key.
扩大AI需要时间，而知道要做什么只是第一步。 牢固的文化，对AI的意识以及各级员工的支持将是关键。
翻译自: https://towardsdatascience.com/preparing-corporations-for-artificial-intelligence-adoption-e67603a22037
人工智能ai以算法为基础