Skip to content

⚙️ Discovery & Strategy — Activities

Purpose

Overview of the core activities and role assignments during the Discovery phase, from problem exploration to data evaluation and risk assessment.

When to use this?

You are in the Discovery phase and want to know which activities to perform — from problem exploration and data evaluation to risk assessment and Collaboration Mode selection.

🎯 Objective

Execute the Discovery phase activities to produce a validated Project Charter, completed Data Evaluation, and initial Risk Inventory that together form the basis for the Gate 1 Go/No-Go decision.

✅ Entry Criteria (Definition of Ready)

  • A business sponsor has formally initiated the project and allocated budget.
  • At least one concrete use case has been identified for investigation.
  • The team has access to stakeholders for interviews and data sources for preliminary assessment.

⚙️ Core Activities

1. Problem Exploration

We define the challenge from the end user's perspective, not from the technology's perspective. This is the most critical activity in Discovery — a poorly defined problem leads to wasted effort in all subsequent phases.

Steps:

  1. Interview stakeholders: Conduct structured interviews with at least three people who experience the problem daily. Document their pain points, workarounds, and desired outcomes.
  2. Articulate the question: Write a single-sentence problem statement that a non-technical manager can understand. Avoid technical jargon. Example: "Customer service agents spend 40% of their time searching for policy information" instead of "We need a RAG system."
  3. Assess AI suitability: Ask explicitly: "Can this be solved without AI?" If the answer is yes — through process improvement, better search, or simple automation — pursue that route first. AI adds complexity and cost; use it only when justified.
  4. Define success indicators: Establish measurable outcomes before any technical work. What does "solved" look like? Examples: "Reduce search time from 5 minutes to 30 seconds," "Achieve 90% accuracy on invoice categorisation," "Cut customer wait time by 50%."

Avoid solution-first thinking

Do not start with "We need a chatbot" or "We need fine-tuning." Start with the problem. The solution emerges from the problem analysis, not the other way around.

Practical Example

Situation: A regional hospital network noticed that emergency department nurses spent an average of 12 minutes per patient gathering preliminary intake information before a doctor could see them. During peak hours, this contributed to wait times exceeding 90 minutes. Approach: The AI Product Manager conducted structured interviews with six triage nurses, two ER physicians, and three patients. Instead of starting with "we need an AI intake assistant," the team articulated the problem as: "Nurses spend 40% of their shift on repetitive intake questions that could be collected earlier." The Data Evaluation revealed that the hospital's EPR system contained 18 months of structured intake data (Access: ✓, Quality: 15% missing fields in free-text notes, Relevance: high correlation between intake completeness and time-to-treatment). The Risk Pre-Scan classified the project as Limited Risk under the EU AI Act, since the system would support — not replace — clinical decision-making. The team selected Collaboration Mode 2 (Advisory): the AI would draft an intake summary, but a nurse would review and confirm every field before it entered the patient record. Result: The Project Charter was approved at Gate 1 with a clear problem statement, a validated data assessment, and a defined Collaboration Mode. The Risk Pre-Scan (template) identified GDPR considerations for patient data, triggering an early consultation with the Guardian. The Data Evaluation Report flagged the 15% missing-value issue, which was scoped as a data cleaning workstream in the Cost Overview.

2. Data Evaluation

An analysis of the required information across three dimensions. Without adequate data, no AI solution can succeed — regardless of the model or approach.

Access

  • Question: Are we legally permitted and technically able to access the data?
  • Check: Legal rights (contracts, licences), API availability, database access credentials, security clearance, data residency requirements.
  • Decision criterion: If access cannot be secured within the project timeline, the project cannot proceed. Document the blocker and escalate to the Business Sponsor.

Quality

  • Question: Is the data complete and consistent?
  • Check: Completeness (are there gaps?), accuracy (does the data reflect reality?), currency (is it up to date?), duplicates (are there conflicting records?), format consistency.
  • Decision criterion: Data with more than 30% missing values or systematic inconsistencies requires a data cleaning workstream before the Validation phase. Factor this into the Cost Overview.

Relevance

  • Question: Does the data contain the answer to the question?
  • Check: Correlation with the objective (does the data relate to the problem?), representativeness (does it cover the target population?), temporal alignment (is the data from the right time period?).
  • Decision criterion: If the data does not contain signals relevant to the problem, the AI cannot learn the task. Consider alternative data sources or reframe the problem.

3. Risk Inventory

An initial scan for legal and ethical obstacles. This activity determines the regulatory burden and governance requirements for the project.

  • EU AI Act Classification: Use the Risk Pre-Scan to determine whether the system falls under Minimal, Limited, or High Risk. High-risk systems require full compliance documentation, conformity assessment, and post-market surveillance.
  • Privacy & GDPR: Identify which personal data is being processed. Document data flows, lawful bases for processing, and data subject rights. Consult the Guardian for privacy impact assessment.
  • Ethical Questions: Can the system discriminate or cause harm? Consider fairness across demographic groups, transparency of decisions, and potential for misuse. The Guardian conducts the initial ethical scan.
  • Organisational Risks: Do we have the right people, skills, and resources? Identify gaps in data science expertise, MLOps capability, or domain knowledge. Plan for hiring, training, or external support.

4. Collaboration Mode Assessment

We determine the intended human-AI relationship for this project. The Collaboration Mode is a design constraint that drives governance, validation depth, and monitoring requirements throughout the lifecycle.

Steps:

  1. Review the five Collaboration Modes — from Mode 1 (Instrumental) to Mode 5 (Autonomous).
  2. Assess the task characteristics: complexity, risk level, volume, and consequences of error.
  3. Select the intended mode. Guideline: Discovery phase projects typically target Mode 1–3. Mode 4–5 require proven reliability from earlier phases.
  4. Record the mode in the Project Charter with rationale.
  5. Note the validation requirements for the selected mode → See Evidence Standards.

Start low, scale up

Begin a use case in Mode 2 (Advisory) to collect data and build trust. Only when quality is proven (>90% on the Golden Set) should you transition to Mode 4 (Delegated).

1b. Project Type Classification

Two project types at a glance

  • Type A — Building with AI: The development team uses AI tools and agentic AI as part of the development process. The end product itself does not need to contain AI.
  • Type B — AI in the Product: The end product integrates AI functionality for end users.

Before proceeding with the core activities, determine the type of AI project. The Blueprint distinguishes two fundamentally different project types:

Characteristic Type A — Building with AI Type B — AI in the product
Description AI/agents are used as development tools (code assistants, test generation, documentation automation) The end product contains AI capabilities for end users (recommendations, classification, generation, agentic workflows)
Risk profile Standard software risks; AI errors affect the development process, not the end user AI-specific risks; errors directly affect end users, customers or business processes
Collaboration Mode Typically Mode 1–2 (the developer reviews AI output) Mode 2–5 depending on risk and volume (full lifecycle required)
Blueprint scope Selective: use Risk Pre-Scan, Governance Model and relevant cheatsheets Full: all phases, Gate Reviews, Collaboration Modes and monitoring apply

This Blueprint is primarily designed for Type B projects

Type A projects (building with AI) may use selected modules but do not require the full lifecycle. Classify your project deliberately — a wrong classification leads to either unnecessarily heavy governance or insufficient safeguards.

Not sure? If the AI system generates output that is directly seen or used by end users without human intervention, it is a Type B project.

👥 RACI

Role Responsibility in Discovery
AI Product Manager Accountable: Owner of the business case and problem articulation.
Data Scientist Responsible: Performing the Data Evaluation.
Business Sponsor Consulted: Validates the problem and the value hypothesis.
Guardian (Ethicist) Consulted: Conducts the initial ethical and legal scan.
Stakeholders Informed: Are kept informed of findings.

✅ Exit Criteria (Gate 1 — Go/No-Go Discovery)

The Discovery phase activities are complete when:

  • Problem statement is documented and validated by stakeholders.
  • Data Evaluation report is completed with positive findings on Access, Quality, and Relevance.
  • Risk Pre-Scan is completed and risk level is classified.
  • Collaboration Mode is assessed and recorded.
  • Project Charter draft is ready for Gate 1 review.

Collaboration Mode: [Mode X — Name] as determined during this phase. Required validation for this mode: → See Evidence Standards.

📦 Deliverables

  1. Project Charter (draft) — scope, objectives, stakeholders, Collaboration Mode.
  2. Risk Pre-Scan — initial risk classification.
  3. Data Evaluation Report — Access, Quality, Relevance assessment.
  4. Collaboration Mode Assessment — documented rationale.

Next step: Complete the Goal card and run the Collaboration Mode Assessment. → Use the Project Charter as your starting point. → See also: Objectives | Collaboration Mode Assessment | Risk Pre-Scan

Version: 1.1 Date: 07 May 2026 Status: Final