AI for Business: #4 Selecting AI Projects

Identifying Game-Changing AI Opportunities

The discussion focuses on the importance of identifying impactful AI opportunities within a company, emphasizing the need for real problem-solving, technical feasibility, significant ROI, and requisite data and skills.

Key Points:

• This course caters to individuals without deep AI technical backgrounds but with a passion for leveraging AI to enhance business operations.

• The session delves into the art and science of discovering valuable AI use cases by exploring various brainstorming techniques, introducing the AI idea canvas as a collaboration tool, and emphasizing evaluation based on business and technical criteria using templates.

• A methodology called ISP (Ideate, Share, Evaluate, Pick) is introduced to guide through ideation from generation to selection. It emphasizes creative brainstorming aligned with business objectives and effective communication in AI projects.

• Collaboration between business and technical teams is crucial for selecting feasible AI ideas. Educating business personnel on AI enhances idea generation quality. Technical expertise ensures project feasibility.

• Involving a diverse group in idea selection optimizes outcomes. Lack of an internal technical team can be supplemented by engaging experienced AI consultants. Educating business staff on AI enhances their contribution to project ideation.

Effective Brainstorming Techniques for AI Ideation

This segment explores strategies for effective brainstorming in generating innovative AI project ideas by defining clear starting points related to business drivers, functions, or strategic objectives.

Key Points:

• Effective brainstorming hinges on well-defined starting points like business drivers (e.g., revenue), strategic objectives (e.g., reducing churn), or functions (e.g., marketing). These serve as foundations for structured ideation processes.

• Choosing a starting point allows focused brainstorming within specific AI use case patterns such as prediction, computer vision, text understanding, and content generation. Teams can generate ideas aligned with these patterns for targeted outcomes.

• Example use cases under different patterns include next best offer for prediction, visual product search for computer vision, customer sentiment analysis for text understanding in marketing contexts. Structured ideation enhances idea quality.

• Ideas like chatbots for lead capture under text understanding or blog generation under content generation exemplify potential applications within specific use case patterns. Structured brainstorming aids in maximizing ideation effectiveness.

• Steps for successful ideation involve forming hybrid teams of business and technical experts equipped with relevant knowledge; choosing suitable starting points; focusing on selected use case patterns; setting time limits; fostering creativity through structured processes.

Communication Strategies for Sharing AI Project Ideas

The importance of clear communication in sharing generated AI project ideas is highlighted as essential for gathering feedback early in the process to refine concepts effectively.

Key Points:

• Effective communication post-brainstorming is vital to gather feedback promptly and refine ideas efficiently during the early stages of project development.

• Clear methods of sharing ideas aid in garnering insights from diverse perspectives within the team while ensuring alignment with project goals.

• Timely feedback loops enable iterative improvements leading to more refined and viable project concepts before advancing further into implementation stages.

• Early engagement with stakeholders through transparent communication fosters collaboration and buy-in towards shared project visions among team members involved in the ideation process.

Evaluation and Selection Process

In this section, the speaker introduces the AI Idea Canvas as a tool to aid in organizing thoughts and ideas for evaluation and selection in AI projects.

AI Idea Canvas Utilization

• The AI Idea Canvas assists in structuring thoughts by focusing on key factors like data acquisition, expected value, end users, etc. Download the canvas from the provided link.

• Seven pillars within the canvas include problem/value articulation, specific idea identification, expected ROI determination, required data listing, AI patterns/tools recognition, end-user analysis, and proof of concept scope clarification.

• Cloud AI services from major providers like Amazon Web Services and Google Cloud can jumpstart AI projects efficiently by offering pre-built models for tasks such as document analysis, computer vision, speech recognition, etc.

Cloud AI Services for Project Acceleration

This part delves into leveraging cloud-based AI services from major providers to expedite project initiation without extensive development efforts.

Leveraging Cloud AI Services

• Major cloud providers offer off-the-shelf services like document analysis and computer vision that can significantly reduce project development time.

• These services handle common AI tasks effectively; customization may be needed based on project requirements but can save substantial time during exploration and proof of concept stages.

Team Collaboration and Idea Presentation

Team collaboration using the AI Idea Canvas is emphasized to generate diverse ideas efficiently.

Team Collaboration Strategies

• Divide team members into groups with at least one technical person per group to ensure technically feasible ideas are generated within specific domains or activities.

• Encourage teams to present their mapped out canvases for feedback and refinement before moving to the evaluation stage; audience interaction aids in idea enhancement.

AI Idea Evaluation Template

Introduction of the AI Idea Evaluation Template for selecting high ROI and technically feasible ideas.

Evaluating Ideas Effectively

• Use the template to evaluate ideas based on data readiness, business impact, technical feasibility, and expected adoption; assign scores accordingly for informed decision-making.

Understanding Model Fine-Tuning and Data Accessibility

In this section, the speaker discusses the importance of fine-tuning models based on data and the accessibility of data for AI projects.

Fine-Tuning Models

• Fine-tune models by using your own data to enhance accuracy.

• The more retraining or fine-tuning required, the lower the model's score.

• Pre-trained models are common for unstructured data like images, voice, text, and videos.

Data Accessibility

• Assessing data availability, labeling, and cleanliness is crucial.

• Insufficient historical sales data can hinder forecasting accuracy.

• Lack of labeled data poses challenges in tasks like defect detection.

• Data may be inaccessible due to security or privacy issues.

Factors Influencing Data Quality for AI Projects

This part delves into factors affecting the quality of data used to train AI models.

Evaluating Data Quality

• Clean, accessible, labeled data enhances model training effectiveness.

• Rate data quality high if using pre-trained models with labeled data.

• Medium rating for available but not perfectly clean or labeled datasets.

Quantifying Business Impact in AI Projects

The speaker explores methods to quantify the business impact of AI ideas.

Business Impact Metrics

• Measure business potential through revenue/cost savings, time efficiency, customer experience enhancement, risk management improvement, and specific business KPIs.

• Financial return and time saved are key metrics for quantifying impact.

• Customer satisfaction and risk management are additional measurable aspects.

Utilizing Business KPIs for ROI Assessment

This segment emphasizes leveraging existing business KPIs to assess ROI in AI projects.

Leveraging Business KPIs

• Use established business metrics relevant to each industry to gauge ROI effectively.

Business Impact of AI Use Cases

In this section, the speaker discusses various AI use cases and their potential business impacts, categorizing them as high, medium, or low based on their ROI and strategic importance.

High Business Impact Use Cases

• Predictive maintenance: Implementing predictive maintenance could save 20% of maintenance costs (approximately $6 million per year) and reduce production downtime by 35%. Initial investment required is around $500k.

• Smart cross-selling and upselling: Using machine learning for recommendations could increase average basket size by 10%, resulting in a $7 million uplift. Initial investment needed is about $350k.

Medium Business Impact Use Cases

• Automating rail asset inspection with AI: Automation can save $350k for a $70k investment, providing a 5x ROI. Considered medium impact as it's not a strategic area.

• Customer chatbox: Expected to save 20% of support team's time, resulting in $1.5 million cost savings annually with a 6x ROI. Deemed non-strategic but still impactful.

Low Business Impact Use Case

• Resume scanning for job applicants: Scanning resumes faster using AI technologies may not have significant ROI due to low volume (30 resumes per day). Considered low impact due to lack of strategic importance.

Feasibility Assessment for AI Projects

The speaker outlines three key elements to assess the feasibility of an AI project: technical complexity, availability of resources and skills, and implementation time.

Technical Complexity

• Some AI ideas are more complex than others, requiring expertise in areas like machine learning algorithms such as time series forecasting and predictive analytics. For instance, building a stock price prediction system from scratch is more complex than using pre-trained models for vehicle detection.

Availability of Resources and Skills

Building AI Teams and Assessing Feasibility

In this section, the speaker discusses how to build AI teams and assess feasibility when implementing AI projects.

Building AI Teams

• Importance of understanding whether expertise is in-house or outsourced.

• Customization and fine-tuning requirements determine the need for deep learning and machine learning skills.

• Consideration of time needed for implementation varies based on project complexity.

Assessing Feasibility

• Practical examples of assessing feasibility based on customization needs.

• Examples ranging from highly feasible (invoice processing automation) to low feasibility (real-time traffic forecasting).

Factors Influencing Expected Adoption of AI Projects

This section explores factors influencing the expected adoption of AI projects within organizations.

Expected Adoption Factors

• Importance of organizational support and adoption for successful AI project implementation.

• Three key elements for high adoption: alignment with organization strategy, team excitement, and a clear Champion.

Strategies for Driving Adoption of AI Projects

Strategies are discussed to drive adoption of AI projects within organizations effectively.

Driving Adoption Strategies

• Emphasize end customer value, specific use cases, education, and sharing success stories.

Google Sheets for Idea Evaluation

In this section, the speaker discusses how to use Google Sheets for idea evaluation and scoring.

Using Google Sheets for Idea Evaluation

• The scoring in Google Sheets increases from low to high values, with low getting one point, medium getting two points, and high getting four points. Factors that score high should have a higher magnitude to push the overall score of an idea.

• The idea score is calculated by multiplying each factor's score by its weight and summing these weighted factor scores. Modifying the weight of each factor can reflect its importance relative to others.

• To add more ideas, copy and paste the last row and drag the last score cell down. It is recommended to start with a quick win when beginning an AI journey – an idea with good business impact that is feasible.

Starting Your AI Journey

This part emphasizes starting with achievable ideas in your AI journey to build momentum and generate excitement companywide.

Beginning Your AI Journey

• Start with a quick win – an idea that has good business impact but is feasible. Avoid complex ideas initially; focus on building momentum and showing success gradually.

• The objective is to create excitement across the company by taking small steps at a time. Success in initial projects encourages further investment in AI initiatives.

Closing Thoughts on AI Idea Generation

The closing thoughts highlight the importance of generating solid AI ideas, evaluating them effectively, and preparing for implementation in upcoming episodes.

Importance of AI Idea Generation

• Understanding how to create solid AI ideas based on multiple criteria sets a strong foundation for selecting initial AI projects.