Quick Summary
AI Canvas 2.0 is a strategic planning framework developed by Kellogg School of Management for planning AI initiatives. It consists of three phases: Define (Find it) - understanding the problem and business value; Design (Bottle it) - developing MVP or POC; Deploy (Scale it) - productionizing and managing implementation. The framework uses 10 narrative questions and 9 blueprint cells to ensure comprehensive AI project planning from conception to deployment.
Understanding AI Canvas 2.0
The AI Canvas 2.0 is a comprehensive, structured framework designed to frame the business case for a single AI initiative, guiding it systematically from conception through deployment. It is an evolution of a previous version, providing a complete view of the entire life cycle of an AI project on one page.
1. What Does the AI Canvas 2.0 Do?
The AI Canvas 2.0 describes the three major phases an AI initiative progresses through, along with the critical decisions and considerations necessary at each stage.
The framework is structured into nine cells across three phases, corresponding to the rows:
| Phase | Core Focus | Blueprint Cell Categories |
|---|---|---|
| Define (Find it) | Understanding the problem, the use case, and the business value. | Business and customer problem, Jobs to be done (JTBD), Business impact. |
| Design (Bottle it) | Developing a Minimum Viable Product (MVP) or Proof of Concept (POC). | Data and model management, Context and fine-tuning, MVP and rapid prototyping. |
| Deploy (Scale it) | Productionizing, scaling, and managing the implementation at an industrial level. | Testing and scaling, Implementation and change management, Governance and risk mitigation. |
2. How Is the AI Canvas 2.0 Used?
The framework can be applied in two complementary formats: the Narrative Format and the Blueprint Format.
A. Narrative Format:
This format is used to tell a compelling story about the initiative by addressing ten questions:
- The business problem to be addressed.
- Why the problem is important.
- How an AI-based solution will address the problem.
- The key benefits to the business and customers.
- The quick financial impact estimate (e.g., in dollars).
- The composition of the required cross-functional team.
- The necessary resources (headcount and budget).
- The 90-day proof-of-concept plan.
- The hurdles/risks that will be faced.
- The plan to mitigate those hurdles and risks.
B. Blueprint Format:
This format organizes the details into nine cells to ensure all necessary components of the project's life cycle are covered systematically.
- Define Phase (Find it): Focuses solely on the problem, not the solution. It defines the Jobs to be Done (JTBD), emphasizing the outcomes the end-user wants to achieve. It also quantifies the Business Impact (revenue enhancement, cost reduction, risk mitigation, productivity improvement).
- Design Phase (Bottle it): Addresses the solution's technical components. This includes determining the Data and model management strategy (data sources, cleanliness, accuracy, access control) and the Context and fine-tuning approach (using prompts, proprietary data, or custom models). It also details the plan for building the MVP using rapid prototyping and an agile approach.
- Deploy Phase (Scale it): Focuses on execution and organizational integration.
- Testing and scaling involves productionizing and integrating the AI system with core enterprise systems of record.
- Implementation and change management is crucial, addressing how adoption will be driven, buy-in secured, and how resistance to changing workflows will be managed (people don't like change).
- Governance and risk mitigation establishes procedures for data and model governance, addressing issues like bias, compliance, safety, and data privacy.
3. Why Is the AI Canvas 2.0 Important and Useful?
The AI Canvas 2.0 is highly important because it ensures that organizations move beyond unorganized proofs of concept (POCs) to strategic, scalable AI deployment.
- Systematic Framing: It provides a comprehensive framework for framing the entire life cycle of an AI initiative on one slide, ensuring no critical aspect—from problem definition to change management—is overlooked.
- Focus on Value: It forces leaders to quantify the Jobs to be Done and the Business Impact early in the "Define" phase, ensuring the project is grounded in solving a specific business or customer problem that is worth solving, preventing "a solution looking for a problem" pathology.
- Risk Mitigation: By dedicating specific cells to Governance and Risk Mitigation and outlining potential hurdles in the narrative format, it promotes proactive risk management concerning bias, compliance, and adoption barriers.
- Cross-Functional Alignment: The framework inherently requires input from multiple stakeholders (traders, commercial teams, production, IT, external experts) early on, promoting the cross-functional collaboration necessary for successful scaling.
- Roadmap Development: It is used as a tool to develop a roadmap for an AI initiative business case.
4. When Should the AI Canvas 2.0 Be Used?
The AI Canvas 2.0 should be used when framing the business case for any single AI initiative. It is an essential step after identifying high-potential areas (e.g., using the AI Radar 2.0) and before significant resources are committed to development. It is designed to guide the project through the Define, Design, and Deploy phases, making it relevant for the initial planning and throughout the execution phases.
5. Example Completed Diagnostic (Case Study: Palm Oil Optimization)
A detailed case study involves a company in Indonesia that manufactures palm oil.
Business Problem: How to optimize the company's margin by determining the optimal product mix of downstream products (e.g., cooking oil, biodiesel). The margins are highly dynamic due to changing prices, logistics costs, and government policy/levy changes.
| Canvas Element (Blueprint/Narrative) | Description/Details from Case Study |
|---|---|
| Business Problem (Define) | Optimizing the product mix (what products and in what mix) to maximize the spread between input and output prices due to volatile market conditions, destination disparity, and fluctuating logistics/duties. |
| Jobs to be Done (Define) | Sourcing and logistic planning, capacity planning, production planning, sales planning, shipment planning, and scenario planning. |
| Business Impact (Define) | The company handles 10 million metric tons per year. Making an additional $10 per metric ton translates to $100 million dropping straight to the bottom line annually. The estimated value over 24 months was $200 million. |
| Solution (Narrative) | Building an AI-based solution to run scenario analysis and recommend the optimal product mix, literally on a daily basis. |
| Data and Model Management (Design) | Required data included historical price data, margin data, and spread data, which had to be accurate and span a certain period of time to train the model. |
| Context and Fine-Tuning (Design) | They required workshops with internal and external stakeholders to add appropriate constraints to the model. |
| MVP and Rapid Prototyping (Design) | A 90-day pilot focusing on one factory that was representative of the company's downstream products was chosen as the proof of concept, utilizing agile sprints. |
| Testing and Scaling (Deploy) | Plans included running a parallel run with human-in-the-loop to ensure the algorithm outperformed manual decision-making and comparing the results before full implementation. |
| Change Management (Deploy) | Required change management workshops with key stakeholders to ensure adoption and buy-in. |
| Governance and Risk Mitigation (Deploy) | Hurdles included data leakage (proprietary trading expertise) and the need for the right technology partner. Mitigation involved proper cybersecurity procedures and establishing governance mechanisms for model failure and data leakage. |
The AI Canvas 2.0 systematically applied here ensures that the company addresses not just the technology needed (the model and data), but also the massive business return and the crucial organizational factors (people, change, risk) required for successful deployment.