How do I identify the right AI tools for a workflow?¶
Short answer: Break your workflow into discrete steps using the 5-question framework (steps, decisions, data flows, context needs, failure modes), then map each step to AI building blocks like prompts, skills, agents, and connectors to see exactly what to build.
The Full Answer¶
You can't choose the right AI tools for a workflow you don't fully understand. The Business-First AI Framework addresses this with a structured approach: discover opportunities first, then deconstruct workflows into building blocks before selecting tools.
Most people jump straight to tool selection — "Should I use an agent? Do I need an API?" — before they've properly decomposed what the workflow actually involves. The result is over-engineered solutions for simple problems, or under-powered tools for complex ones.
The right approach is to deconstruct first, then map. Start by breaking your workflow into discrete steps using the 5-question framework: What are the individual steps? Where are the decision points? What data flows in and out? What context does each step need? What happens when this step fails? That last question surfaces the exception paths where the most important logic often lives.
Once you have the refined step-by-step breakdown, map each step to one or more of the seven AI building blocks:
| Building Block | What It Is | When to Use It |
|---|---|---|
| Model | The AI engine that processes inputs and generates outputs | When the task requires specific capabilities (reasoning, multimodal, speed) |
| Prompt | A well-crafted instruction that tells the model what to do | Single-step tasks with clear inputs and outputs |
| Context | Background information, reference docs, or examples | Steps that need domain knowledge or specific data |
| Skill | A reusable routine — give it inputs, it follows a defined process, it produces consistent outputs | Repeatable tasks you'll run many times |
| Agent | An autonomous AI that plans, uses tools, and executes multi-step work | Complex steps requiring judgment and tool use |
| MCP (Model Context Protocol) | A connector to external tools, APIs, or databases | Steps that need to read from or write to external systems |
| Project | A persistent workspace grouping prompts, context, skills, and agents | Organizing everything for a specific workflow |
Not every step needs AI. The deconstruction also classifies each step on an autonomy spectrum — from fully human to fully autonomous — so you can see which steps are candidates for AI and which should stay manual.
How to Get Started¶
Use the Deconstruct Workflows to run through this process interactively. Paste it into any AI tool and it will:
- Analyze your scenario — understand the workflow objective and rough steps
- Deep dive into each step — apply the 5-question framework
- Map to building blocks — classify each step and recommend the right AI tools
- Generate deliverables — produce a workflow analysis document and an executable prompt you can save and reuse
Start with a workflow you actually do
Real workflows produce the best analysis. The meta prompt is designed to work with rough, incomplete descriptions — don't over-prepare. Even "I onboard new clients and it takes forever" is enough to start. The model will find the hidden steps and assumptions you've internalized.
Key Takeaways¶
- Deconstruct the workflow before choosing tools — you can't pick the right AI building blocks for a process you don't fully understand
- Use the 5-question framework: discrete steps, decision points, data flows, context needs, and failure modes
- Map each step to one or more of the seven building blocks: Model, Prompt, Context, Skill, Agent, MCP, or Project
- Not every step needs AI — the autonomy classification helps you see which steps are candidates and which should stay manual
- Use the Deconstruct Workflows to run through this process interactively