Applied Artificial Intelligence

1. AI: A Practical Tool for Business Leaders

Applied Artificial Intelligence is a practical guide for business leaders who are passionate about leveraging machine intelligence to enhance the productivity of their organizations and the quality of life in their communities.

Bridging the Gap. This book serves as a bridge between the technical complexities of AI and the practical needs of business leaders. It avoids overwhelming technical jargon while providing enough detail to make informed decisions about AI initiatives. The focus is on leading successful AI projects, building diverse teams, and designing solutions that benefit both the organization and society.

Beyond the Hype. The book emphasizes separating AI hype from reality. It encourages business leaders to understand the fundamental concepts of AI, including its capabilities and limitations, before investing in AI projects. This understanding helps in prioritizing the right opportunities and avoiding costly mistakes.

Actionable Insights. The guide provides actionable insights for business leaders to drive innovation by combining data, technology, design, and people to solve real-world problems at an enterprise scale. It's a playbook for those who love to drive innovation and solve real problems.

2. Understanding the Machine Intelligence Continuum

While we’ve defined the continuum to contain seven levels, keep in mind that the distinction between levels is not a hard line and that many overlaps exist.

Defining AI Capabilities. The Machine Intelligence Continuum (MIC) is a framework that categorizes different types of machine intelligence based on the complexity of their capabilities. This continuum helps business executives understand the functional differences between various AI approaches, ranging from simple rule-based systems to advanced evolving systems.

Seven Levels of Intelligence. The MIC includes seven levels:

• Systems That Act (rule-based automata)
• Systems That Predict (statistical analysis)
• Systems That Learn (machine learning)
• Systems That Create (generative design)
• Systems That Relate (emotional intelligence)
• Systems That Master (abstract concepts)
• Systems That Evolve (superhuman intelligence).

Practical Application. Understanding the MIC enables business leaders to assess the current state of AI technologies and set realistic expectations for AI projects. It helps in identifying the right level of machine intelligence needed to solve specific business problems, avoiding the trap of over-engineering solutions.

3. AI's Societal Promises: Beyond Profit

The promises of AI extend beyond the challenges of Silicon Valley and Wall Street.

Social Impact Focus. AI's potential extends far beyond corporate profits, offering solutions to pressing social and humanitarian issues. Examples include using AI to combat social injustice, address health crises, and improve the quality of life for communities worldwide.

Microfinance Example. FarmGuide uses deep learning to analyze satellite imagery and predict crop yields for individual farms in India. This information helps provide farmers with fairer interest rates, reducing the risk of predatory loans and farmer suicides.

Social Justice Example. UNICEF's U-Report uses a social reporting bot to enable young people in developing countries to report social injustice via SMS. This platform has been instrumental in exposing issues like the "Sex 4 Grades" epidemic in Liberia, leading to government intervention and support for victims.

4. Navigating the Challenges of AI: Bias and Malice

Data and technology are human inventions, ideally designed to reflect and advance human values.

Addressing Algorithmic Bias. AI systems can unintentionally amplify discrimination against underrepresented groups due to biased data or algorithms designed by undiversified teams. Examples include Amazon's same-day delivery being unavailable in predominantly black neighborhoods and women being shown fewer ads for high-paying jobs.

Combating Malicious AI. As AI becomes more powerful, the risk of malicious use increases. This includes using AI to multiply the effects of malicious campaigns, produce fake news, circumvent security systems, and deploy autonomous weapons.

Importance of Diversity. The lack of diversity in AI development teams can lead to overlooking the needs and values of underrepresented groups. It's crucial to build diverse teams and foster inclusive thinking to mitigate these risks.

5. Designing Ethical and Safe AI Systems

AI systems can’t simply be programmed to complete their core tasks. They must be designed to do so without unintentionally harming human society.

Ethics and Governance. Designing safe and ethical AI requires more than just hypothetical fail-safe mechanisms. It involves developing sophisticated policies and standards that address the myriad ways in which AI systems can go awry.

Education as Remedy. Democratizing access to quality AI education and empowering collaborations between practitioners and multidisciplinary experts are essential. Initiatives like fast.ai are making deep learning accessible to all, enabling individuals to use AI for social good.

Collaborative Design Principles. Three principles of collaborative design:

• Build User-Friendly Products to Collect Better Data for AI
• Prioritize Domain Expertise and Business Value Over Algorithms
• Empower Human Designers With Machine Intelligence

6. Building an AI-Ready Culture: The Foundation for Success

In every single tech firm that currently leads in AI, the CEO has come out strongly in favor of prioritizing AI company-wide.

Assessing Organizational Readiness. Before embarking on AI initiatives, it's crucial to honestly assess your organization's readiness. This includes evaluating your technology infrastructure, data culture, and leadership commitment.

Choosing the Right Champions. Identifying and empowering the right executive to champion AI initiatives is critical. The ideal champion should be a C-Suite executive with technical knowledge, business acumen, and the ability to secure organizational buy-in.

Building an AI SWAT Team. Assemble a multi-disciplinary team composed of stakeholders from different departments and hierarchical levels. This team will identify, prioritize, execute, and evangelize high-ROI opportunities for automation across the company.

7. Investing in the Right AI Talent

There are fewer than 10,000 people in the world currently qualified to do state-of-the-art AI research and engineering.

Understanding Job Titles. Different roles are required for successful AI implementation, including data science team managers, machine learning engineers, data scientists, researchers, data engineers, and distributed systems engineers. Understanding these roles helps in recruiting the right talent.

Seeking Key Characteristics. Beyond technical skills, look for candidates with mathematical aptitude, curiosity, creativity, perseverance, rapid learning abilities, and a passion for your problem. These characteristics are essential for success in AI.

Optimizing Recruiting Strategies. Tailor your recruiting approach to the level, background, and career goals of your prospects. This includes university partnerships, hackathons, specialized training programs, and emphasizing your company's unique advantages.

8. Strategic AI Implementation: Planning for ROI

Presenting a clear ROI on AI initiatives is the best way to persuade executive stakeholders.

Ranking Business Goals. Prioritize business goals and identify opportunities for AI adoption that align with your company's strategic plan. Common goals include increasing revenue, cutting costs, and entering new business lines.

Performing Opportunity Analysis. Use analytical frameworks like Gap Analysis and SWOT Analysis to evaluate current enterprise workflows and technologies. This helps in identifying areas where AI can deliver the highest return on investment.

AI Strategy Framework. Use the AI Strategy Framework to evaluate each opportunity based on strategic rationale, opportunity size, investment level, ROI, risk, timeline, and stakeholder buy-in. This framework provides a structured approach to decision-making.

9. Data: The Fuel for AI, Not Reality Itself

Data is a human invention.

Data as a Construct. Data is not an objective representation of reality but a human construct shaped by the choices of what to measure and how to interpret the results. Recognizing this helps in avoiding the trap of blindly trusting data.

Common Data Mistakes. Be aware of common mistakes in data collection and processing, including undefined goals, definition errors, capture errors, measurement errors, processing errors, coverage errors, sampling errors, inference errors, and unknown errors.

Establishing Ground Truth. Strive to establish ground truth, which is observable, provable, and objective data that reflects reality. This serves as a benchmark for assessing the performance of AI algorithms and ensuring accurate results.

10. Building and Assessing Machine Learning Models

Machine learning is a powerful tool, but it is not right for everything.

AI Is Not a Silver Bullet. Machine learning is not a universal solution and is best suited for specific types of problems. It's crucial to understand the capabilities and limitations of different algorithms and choose the right approach for each task.

Assessing Model Performance. Use evaluation metrics like accuracy, precision, and recall to assess the performance of machine learning models. Understand the trade-offs between these metrics and choose the ones that align with your business goals.

Avoiding Common Mistakes. Be aware of common mistakes in building machine learning models, including underfitting and overfitting. Implement a rigorous validation and test process to mitigate these issues and ensure accurate predictions.

11. AI for Enterprise Functions: Streamlining Operations

Enterprise functions represent one of the easiest entry points for deploying AI within your own company.

Targeting Inefficiencies. AI can be effectively deployed to streamline processes and automate repetitive tasks in enterprise functions. This frees up employees to focus on higher-value activities and improves overall efficiency.

Examples of AI Applications. AI can be applied to various enterprise functions, including finance and accounting (expense management, spend analysis), legal and compliance (contract review, due diligence), human resources (candidate matching, intelligent scheduling), business intelligence (data wrangling, analytics), software development (rapid prototyping, intelligent programming assistants), marketing (digital ad optimization, personalization), sales (lead qualification, sales analytics), and customer support (conversational agents, social listening).

Overcoming Obstacles. Despite the potential benefits, companies face obstacles in adopting AI, including HiPPO resistance, lack of data, high costs, and integration challenges. Addressing these obstacles requires executive commitment, technical understanding, and a strategic approach.

12. Ethical Responsibility: AI for Good

We believe that business leaders have an ethical responsibility to workers to minimize and ameliorate the potential disruptions that AI may bring to the workplace.

Workforce Transition. Business leaders have an ethical responsibility to minimize the potential disruptions that AI may bring to the workplace. This includes investing in workforce training and creating new opportunities for employees.

Continuous Learning. Workers will need continuous access to high-quality training in order to stay competitive in the AI-powered economy. This requires a commitment to lifelong learning and the development of new skills.

Upholding Human Values. Business leaders must ensure that AI is designed and deployed in a benevolent fashion, upholding human values and avoiding harm to customers, employees, and society. This requires vigilance and a commitment to ethical practices.

Last updated: February 23, 2025