【中文版在后面】

To sharpen their competitive edge and foster long-term success, organizations must regard data as a high-value business asset and a basis for developing AI tools.

The scorching pace of AI development promises to transform not just companies, but entire industries.  Customers are also rapidly taking to AI and, as a result, their expectations around speed, quality and personalization are changing.  AI must feature prominently on every company's business strategy.  But how should companies get to grips with it and use it to accelerate growth or enhance their service offerings?  First, they need to invest substantially in reliable data, the foundation for developing and trying out multiple AI models.  Then they must prioritize the models which provide quick wins or promise to propel them ahead of their competitors for the long term.

Getting ahead in the age of AI

AI has grown from being a mere support function, sitting at the fringes of the organization, to technology that sits at the nerve center – a crucial component of critical functions and decision-making.  This marks a dramatic shift and companies have to reorient to an AI-first model.  “We believe AI-first companies will be the trailblazers of the future,” “They use AI to transform the wealth of data they have into smart insights which then transform their business models and value propositions to customers.”

But not all companies possess this wealth of data.  Some organizations are still lagging behind due to a lack of data, or stricter frameworks around data handling and usage in some industries.  Data is a strategic business asset and organizations need to dedicate adequate resources to obtaining it – or they risk competitors getting ahead of them with new, innovative products and services.

Five steps to establishing an AI-first company

1. Conduct a data inventory – existing data, deficits, sources and suppliers

2. Assess the organization's AI-readiness – evaluation of data, skills and technology

3. Develop an AI-first strategy – objectives and a technological roadmap

4. Identify business priorities for AI leverage – quick wins and long-term gains

5. Establish and execute a proof of concept – design, scope and testing

Investing in data equates to investing in AI. But what more should a company do? The following steps will help a company progress with AI development and AI-supported solutions:

1. Conduct a data inventory

An organization must have an overview of the data it currently generates and processes, identify gaps and prioritize critical data elements for their organization by considering the potential business impact of bridging these gaps.

As the data will form the backbone for new, powerful AI products, it must be evaluated with a view to processing it effectively. The sources, quality and consistency should be scrutinized, as well as how it is labeled and categorized. If key data is lacking, a plan for creating, leveraging or acquiring it must be made. C-level managers must decide on strategic partnerships with data collection service providers to gain access to relevant data assets.

Regarding management of the data, We explained, “Already multiple challenges have been encountered among organizations in diverse sectors, from trying to adopt a one-size-fits-all approach with data management practices across multiple business units (BUs), to a lack of standards or defined data roles throughout the organization to champion a data agenda.” A lack of robust mechanisms for the evaluation and governance of data regarding privacy and ethics has also proved problematic and must be prioritized. Organizations must adhere to data protection policies, security and ethics principles, including monitoring and governance of all related processes, decision-making and output.

2.   Assess the organization's AI-readiness

Embracing the prospect of more advanced AI solutions is likely to be easier for companies which experienced a digital transformation in recent years, as well as those with a deeply ingrained culture of innovation.   But even they need to consider employee attitudes: the development of AI tools could completely reshape team roles and functions.   For successful AI-integration, employees must be willing to work with new technologies.   Leaders must be tuned in to any reluctance to work with AI tools.   They also need to consider the skill sets of their workers.   For example, can they write generative AI prompts in English or will they require training for this?   As with all major business changes, assigning an owner for the AI agenda can help steer its implementation, overcome perceived limitations and address specific requirements and concerns.

How a company is set up impacts its ability to drive and facilitate the use of AI.   We said, “When designing the organization structure, the proximity of AI resources to data teams is an important consideration.   The functional closeness of AI teams to data-centric teams has the potential to form synergies or create silos.”

Decisions must also be made regarding short-term versus long-term AI leverage.   Short-term thinking would revolve around the optimal organizational structure, governance guardrails and resource profiles needed to enable successful development of AI use cases.   Leaders might also consider how AI could be developed at scale, to allow for standardization and knowledge transfer across processes, technology and tools throughout the entire organization.

3.   Develop an AI-first strategy

“An AI strategy serves as the guiding beacon for a company's AI initiatives by ensuring fewer ‘reactive and hype-oriented initiatives’ and more ‘strategic decision-making and AI-capability building’,” We said.

But how does one go about adopting an AI-first approach?   A company behind a popular app might look to AI to enhance or personalize the user experience or plan to use large language models for content creation.   A service provider might want to implement AI tools to take on laborious, repetitive tasks or analyze vast data sets.   Leaders must take a holistic look at their own company and consider AI-driven measures to support their entire workforce in their regular business operations or where AI could completely free up time and space for teams to invest in more complex, strategic or creative endeavors.

A strategic AI-first approach would consider initiatives which

• offer the most business value as top priority – leveraging AI to do more with less and transform value creation.

• impact the entire organization, not just one business unit, and ensure synergies between organizational functions.

• are convincing and feasible enough to secure buy-in and sponsorship

4. Identify business priorities for AI leverage

To understand where AI could support business further, companies could revisit their understanding of customer needs and critically evaluate the services and products they offer customers. Are the services still in demand? Do the problems they typically solve for customers still exist? An industry-wide look at how AI features and solutions already serve customers, and observing current customer-AI interaction, offers insights into opportunities for new value creation in the future.  

Knowing where to zero in is an essential element for the success of an AI strategy. A breakdown of an organization into functions, such as HR, Finance, etc., or by industry value chain helps determine key business activities. AI use cases can then be identifed and mapped to each activity as relevant. Long-term decisions on exploring AI capabilities would consider an organization's people, skills, processes and the technological roadmap required for implementation.

The two main clusters of AI use cases are “time-savers” and “quality improvers” with the potential for impact in these areas:

• Personalization of the customer experience
• Improved customer acquisition, retention and sales
• Creation of new products and revenue streams
• Improved system efficiency and automation
• Increased productivity of knowledge workers
• Rapid data-driven decision-making
• Reduction in product and research innovation costs
• Improved risk prediction and management

As the prioritization of AI use cases is the most critical task in devising an AI strategy, experts at Roland Berger have developed a guideline for individual businesses to establish their own priorities for AI initiatives. Criteria include all inputs, from the business context to AI focus areas, then a look at prioritization factors which can be clustered under the overarching key benefits “business value” and “ease of implementation.” Once priority use cases have been assessed, the value behind each use case would be defined by identifying value drivers, defining the revenue and cost metrics impacted, estimating the benefts of the use case, calculating implementation costs and conducting risk and scenario analyses.

Companies could consider adopting a two-speed approach to AI:

Quick wins Quick wins to maintain market share through speed or personalization improvements.
Examples:

• A chatbot to answer pre-established FAQs for faster response times 24/7

• The integration of predictive maintenance in production or safety systems to avoid downtime

• Product recommendations for customers based on previous purchases

Long-term game-changers to revolutionize a company's use of AI.
Examples:

• A sophisticated chatbot for solving all issues within an application, thus replacing specialized IT-support workers

• Enterprise search capabilities which learn from consolidated past queries and responses for faster knowledge retrieval and subsequent service offerings

5. Establish and execute a proof of concept

Once the top priority AI use case has been identified, all relevant parameters can be determined for a proof of concept (PoC). These would include the scope, development requirements, key hypotheses to validate or test the PoC and how results will be evaluated with key stakeholders. Decision gate analysis – a “go” or “no-go” decision – for a minimal viable product (MVP) must be established along with all other MVP requirements prior to the development kick-off.  

Ready-made AI models are also becoming increasingly affordable. This makes purchasing them an appealing option compared to creating a customized solution from scratch. We explains, “To build or to buy – and which model or mix of models to buy – is best considered strategically and depends on a company’s own capabilities and synergies, time-to-market, the cost or benefit business case, among countless other factors.” 

Further factors to aid decision-making point to the human aspects of AI use, such as ethical considerations within the PoC design, sufficient involvement of humans in the AI loop and the level of personalization possible. In addition, if an AI model is to be employed in a highly accurate environment, such as cancer treatment, then model responses and outcomes must be critically evaluated to ensure they are precise, with no risk of AI hallucinations – the term used for algorithmic output based on processing incorrect, incomplete or biased data. 

To get ahead in a fast-paced AI-driven world, businesses must pay attention to every AI development that could impact their industry. They must actively invest in the generation or acquisition of relevant data, as well as ethical frameworks for data management and analysis. Only organizations that are fully committed to an AI-first approach will be the ultimate winners, with diverse opportunities to explore new avenues of business. 

【中文版】

人工智能优先的制胜法则

为了提高竞争优势并促进长期成功，组织必须将数据视为高价值的业务资产和开发人工智能工具的基础。

人工智能的迅猛发展不仅会改变公司，还会改变整个行业。客户也在迅速接受人工智能，因此，他们对速度、质量和个性化的期望正在发生变化。人工智能必须在每家公司的商业战略中占据突出地位。但是，企业应该如何把握它，并利用它来加速增长或提高服务水平呢?首先，他们需要大量投资于可靠的数据，这是开发和尝试多种人工智能模型的基础。然后，他们必须优先考虑那些能让他们快速获胜的模式，或者那些能让他们长期领先于竞争对手的模式。

在人工智能时代取得领先

人工智能已经从仅仅是一个位于组织边缘的支持功能，发展成为位于神经中枢的技术——关键功能和决策的关键组成部分。这标志着一个巨大的转变，公司必须重新定位到人工智能优先的模式。“我们相信人工智能优先的公司将成为未来的开拓者，”“他们利用人工智能将他们拥有的丰富数据转化为聪明的见解，然后将他们的商业模式和价值主张转化为客户。”

但并非所有公司都拥有如此丰富的数据。由于缺乏数据，或者在某些行业中围绕数据处理和使用的更严格的框架，一些组织仍然落后。数据是一种战略性的商业资产，组织需要投入足够的资源来获取数据，否则他们就有可能面临竞争对手以新的、创新的产品和服务超过他们的风险。

建立人工智能优先公司的五个步骤

1. 进行数据盘点-现有数据，赤字，来源和供应商

2. 评估组织的人工智能准备情况——对数据、技能和技术进行评估

3. 制定人工智能优先战略——目标和技术路线图

4. 确定人工智能杠杆的业务优先级-快速获胜和长期收益

5. 建立并执行概念验证——设计、范围和测试

投资数据等同于投资人工智能。但公司还应该做些什么呢?以下步骤将帮助公司在人工智能开发和人工智能支持的解决方案方面取得进展:

1. 进行数据盘点

组织必须对其当前生成和处理的数据有一个概述，通过考虑弥合这些差距的潜在业务影响，确定差距并为组织确定关键数据元素的优先级。

由于数据将成为新的、强大的人工智能产品的支柱，因此必须对其进行评估，以便有效地处理它。来源，质量和一致性应仔细审查，以及如何标记和分类。如果缺少关键数据，则必须制定创建、利用或获取关键数据的计划。c级管理人员必须决定与数据收集服务提供商的战略合作伙伴关系，以获得对相关数据资产的访问权限。

关于数据管理，我们解释说:“在不同部门的组织中已经遇到了多种挑战，从试图在多个业务单元(BUs)中采用一刀切的数据管理实践方法，到在整个组织中缺乏标准或定义数据角色来支持数据议程。”缺乏关于隐私和道德的数据评估和治理的强大机制也被证明是有问题的，必须优先考虑。组织必须遵守数据保护政策、安全和道德原则，包括对所有相关流程、决策和输出的监控和治理。

2.   评估组织的人工智能准备情况

对于近年来经历过数字化转型的公司，以及那些拥有根深蒂固创新文化的公司来说，拥抱更先进的人工智能解决方案的前景可能更容易。但即使是他们也需要考虑员工的态度:人工智能工具的发展可能会彻底重塑团队的角色和职能。为了成功地整合人工智能，员工必须愿意使用新技术。领导者必须意识到不愿使用人工智能工具。他们还需要考虑员工的技能组合。例如，他们是否可以用英语编写生成式人工智能提示，或者他们是否需要为此进行培训?与所有主要的业务变化一样，为人工智能议程分配一个所有者可以帮助指导其实施，克服感知到的限制，并解决特定的需求和关注点。

一家公司的成立方式会影响其推动和促进人工智能使用的能力。我们说:“在设计组织结构时，人工智能资源与数据团队的接近性是一个重要的考虑因素。人工智能团队与以数据为中心的团队在功能上的密切关系，有可能形成协同效应或产生孤岛。”

还必须就人工智能的短期和长期杠杆做出决策。短期思维将围绕着实现人工智能用例成功开发所需的最佳组织结构、治理护栏和资源概况。领导者可能还会考虑如何大规模开发人工智能，以便在整个组织中实现跨流程、技术和工具的标准化和知识转移。

3.   制定人工智能优先战略

我们表示:“人工智能战略可以作为公司人工智能计划的指路明灯，确保减少‘反应性和炒作性计划’，增加‘战略决策和人工智能能力建设’。”

但是如何采用人工智能优先的方法呢?热门应用背后的公司可能会寻求人工智能来增强或个性化用户体验，或者计划使用大型语言模型来创建内容。服务提供商可能希望实现人工智能工具来承担繁重的重复性任务或分析大量数据集。领导者必须全面审视自己的公司，并考虑人工智能驱动的措施，以支持他们的全体员工进行日常业务运营，或者人工智能可以完全腾出时间和空间，让团队投资于更复杂的、战略性的或创造性的努力。

战略性的人工智能优先方法将考虑以下举措

1. 提供最大的商业价值作为首要任务——利用人工智能以更少的投入做更多的事情，并转变价值创造。

2. 影响整个组织，而不仅仅是一个业务单元，并确保组织功能之间的协同作用。

3. 是否有足够的说服力和可行性来获得支持和赞助

4. 确定AI杠杆的业务优先级

为了了解人工智能在哪些方面可以进一步支持业务，公司可以重新审视他们对客户需求的理解，并批判性地评估他们为客户提供的服务和产品。这些服务还有需求吗?他们通常为客户解决的问题是否仍然存在?在全行业范围内观察人工智能功能和解决方案如何为客户提供服务，并观察当前客户与人工智能的互动，为未来创造新价值的机会提供洞见。

知道在哪里归零是人工智能策略成功的基本要素。将组织分解为职能，如人力资源、财务等，或按行业价值链划分，有助于确定关键的业务活动。然后可以识别AI用例，并将其映射到每个相关的活动。探索人工智能能力的长期决策将考虑组织的人员、技能、流程和实现所需的技术路线图。

人工智能用例的两个主要集群是“节省时间”和“质量改善者”，它们在这些领域具有潜在的影响:

• 客户体验的个性化

• 提高客户获取、保留和销售

• 创造新产品和收入来源

• 提高了系统效率和自动化程度

• 提高知识型员工的生产力

• 快速数据驱动决策

• 降低产品和研究创新成本

• 改进风险预测和管理

由于人工智能用例的优先级是制定人工智能战略的最关键任务，罗兰贝格的专家们为个别企业制定了一项指导方针，以确定自己的人工智能计划优先级。标准包括从业务环境到人工智能重点领域的所有输入，然后查看优先级因素，这些因素可以聚集在“业务价值”和“易于实现”的首要关键优势下。一旦评估了优先级用例，每个用例背后的价值将通过识别价值驱动因素、定义受影响的收入和成本度量、估计用例的好处、计算实现成本以及进行风险和场景分析来定义。

 节省时间的用例

• 提高了系统效率和自动化程度

• 提高知识型员工的生产力

• 快速数据驱动决策

• 质量改进用例

  
  

客户体验的个性化

• 提高客户获取，保留和销售

• 创造新产品和收入来源

• 降低产品和研发创新成本

• 改进了风险预测和管理

公司可以考虑采用双速人工智能方法:

通过快速或个性化的改进来保持市场份额。

例子:

• 一个聊天机器人回答预先建立的常见问题，以更快的响应时间24/7

• 在生产或安全系统中集成预测性维护，以避免停机

• 根据以前的购买为客户提供产品推荐

例子:

• 一个复杂的聊天机器人，用于解决应用程序中的所有问题，从而取代专门的it支持人员

• 企业搜索功能，从合并的过去查询和响应中学习，以便更快地检索知识和提供后续服务

5. 建立并执行概念验证

一旦确定了最优先的人工智能用例，就可以确定概念验证(PoC)的所有相关参数。这些将包括范围、开发需求、验证或测试PoC的关键假设，以及如何与关键利益相关者一起评估结果。在开发开始之前，必须建立最小可行产品(MVP)的决策门分析——“通过”或“不通过”的决定——以及所有其他MVP需求。

现成的人工智能模型也变得越来越便宜。与从头开始创建定制解决方案相比，这使得购买它们成为一个有吸引力的选择。我们解释说:“建造或购买——以及购买哪种模型或模型组合——最好从战略上考虑，这取决于公司自身的能力和协同效应、上市时间、成本或收益的商业案例，以及无数其他因素。”

辅助决策的进一步因素指向人工智能使用的人类方面，例如PoC设计中的道德考虑，人工智能循环中人类的充分参与以及可能的个性化水平。此外，如果一个人工智能模型被用于一个高度精确的环境，比如癌症治疗，那么模型的反应和结果必须经过严格评估，以确保它们是精确的，没有人工智能幻觉的风险——人工智能幻觉是指基于处理不正确、不完整或有偏见的数据的算法输出。

为了在快节奏的人工智能驱动的世界中取得领先，企业必须关注可能影响其行业的每一个人工智能发展。他们必须积极投资于相关数据的生成或获取，以及数据管理和分析的道德框架。只有完全致力于人工智能优先方法的组织将成为最终的赢家，拥有探索新业务途径的各种机会。