Frequently Asked Questions About Wooldridge Connected AI Design Workflow
21 answers covering everything from basics to advanced usage.
// Basics
What does 'Connection Over Collection' actually mean?
It means mastering a few tools that connect to each other is better than accumulating tens of disconnected ones. The value isn't in any single tool — it's in how they link, because each connected tool makes the next one more powerful. Piling up tools that don't talk to each other just leaves you feeling behind despite having more than you need.
What is 'Context-First Execution'?
It's the principle that every downstream AI action is only as good as the context fed into it. You capture context early — in meetings and docs — so when you hand off to an AI agent, it never works from a blank slate. Capturing notes isn't busywork; it's what makes the agent's output actually useful.
What tools does this workflow recommend?
The reference stack is Granola for ambient meeting transcription, Notion as the Single Source of Truth, Claude Code as the AI agent with MCP connectors, Figma and GitHub as connected tools, and Spokenly for local voice-to-text dictation. The specific brands are swappable — what matters is that they connect via MCPs and produce structured, not raw, output.
Why is speaking better than typing for AI prompts?
Speaking is three to four times faster than typing, and hearing your words out loud reduces mistakes and surfaces things you want to include. For long, guidance-heavy prompts — especially corrective ones when the AI is being stubborn — dictation avoids the fatigue and errors of typing several hundred words manually.
// How To
How do I capture a meeting without disrupting it?
Run an ambient transcription tool like Granola locally in the background — do not join a bot to the call. Let it run quietly during the kickoff, client call, or research session. After it ends, retrieve the structured summary. For in-person interviews where you can't take notes, use the mobile version of the same tool.
How do I connect MCP connectors in Claude Code?
Navigate to Customize → Connectors → Browse Connectors, then search for and connect your meeting tool, document hub, design tool, code repository, and communication tools. Authenticate each one. This is a one-time setup per workspace. Once connected, the agent reads and writes across all of them simultaneously without any copy-paste.
How do I turn messy meeting notes into clean documentation?
Drop the structured meeting summary into your Notion hub, then use the built-in AI writing assistant to refine rough notes into a polished document. The intended pattern is jot things down quickly, then refine later with AI. Don't use this step to generate designs — keep it purely for clean, accurate, usable documentation.
How do I write a good context-grounded prompt for the agent?
Name the specific documents and tools to reference, then state what to produce. For example: 'Based on my discovery notes and PRD in Notion, generate a first-pass screen flow for the onboarding experience.' Let the MCP connectors pull the content — never paste it manually. Expect a first pass to iterate on, not a final deliverable.
// Troubleshooting
My AI agent isn't pulling context from my tools — what's wrong?
First verify your MCP connectors are actually live — connections can drop between sessions, so check them at the start of each major work session. Second, confirm your documents live in the Single Source of Truth, not scattered locally. Third, make sure your prompt explicitly names the tool and document to reference so the agent knows where to look.
Why is my voice dictation producing wrong text?
You're likely using a cloud-dependent or low-accuracy dictation tool. Switch to one that uses local AI models — they're faster and more accurate, which is critical since inaccurate transcription undermines the entire speed advantage of voice input. Prioritize local models over cloud-dependent dictation for prompt writing.
The AI agent's first output is off — is the workflow broken?
No — the workflow is designed to produce a first pass, not a finished deliverable. Treating the initial output as final is a common mistake. Review it, then dictate correction prompts via voice, referencing specific screens, components, and desired changes. Iterate until it's shippable. The speed advantage compounds during this correction loop.
Why does my agent seem to be working from a blank slate?
You've probably skipped the Single Source of Truth setup, leaving documents scattered locally or across platforms. Without a central hub, MCP connectors have nothing to pull from and the agent starts blind. Consolidate all specs, PRDs, and notes into one interconnected Notion workspace so both the agent and collaborators can reference it.
// Comparisons
How does this compare to keeping 15 separate AI design tools?
Fifteen disconnected tools leave you feeling behind despite having more than you need — they don't share context, so you copy-paste constantly. The Connected AI Design Workflow uses a few tools that feed each other via MCPs. Each connected tool makes the next more powerful, so a lean stack outperforms a large disconnected one.
How is this different from manually copy-pasting context into ChatGPT?
Manual copy-paste is slow, error-prone, and defeats the purpose of a connected workflow. MCP connectors let the agent pull context directly from Notion, Figma, and GitHub in real time and even write back to them. You reference documents by name in the prompt instead of pasting walls of text, and the agent can read multiple tools at once.
How does structured meeting output compare to raw transcripts?
A raw transcript is a wall of unusable text that dilutes and confuses. A structured summary organizes the meeting into key decisions, action items, and open questions — a real foundation. Only structured output should enter the workflow; feeding raw transcripts to your agent degrades every downstream result because context gets buried in noise.
Is this workflow better for solo designers or teams?
Both — but the Single Source of Truth scales the benefit for teams. Solo designers gain speed from connected tools and voice dictation. Teams gain shared context: add collaborators to the Notion hub so everyone and every AI agent references the same source, just like a shared design file. Fragmented docs break the workflow for teams especially.
// Advanced
Can the AI agent reference multiple connected tools in one prompt?
Yes. With MCP connectors set up, a single prompt can pull from your design spec in Notion, interview notes in another Notion doc, and files in Figma or GitHub simultaneously. You just name which sources to reference and what to produce, and the agent aggregates live context across your entire connected stack in one pass.
How do I keep MCP connectors reliable across long projects?
Verify connections are live at the start of each major work session — don't assume a one-time setup stays active. Re-authenticate if a tool has expired. Keep your Single Source of Truth well-organized so the agent can locate documents, and standardize how you name docs so your prompts can reference them consistently.
Can I use this workflow for in-person user research?
Yes. Use the mobile version of your transcription tool to capture the interview audio while you stay fully present to conduct the session. Afterward, it produces a structured summary. Refine it with Notion's AI writing assistant into a clean interview document, then share by adding your team to the workspace — feeding that context into future agent prompts.
Can the agent write back to my tools, not just read them?
Yes — MCP connectors are bidirectional. The agent can read from Figma, Notion, and GitHub and also write to them, such as updating a Notion doc or committing to a repo. This is why dictation extends beyond prompts to document updates and emails: the connected stack lets the agent act, not just answer.
How do I scale this workflow across multiple concurrent projects?
Set up a separate Single Source of Truth workspace per project so context stays isolated and clean. Since MCP connector setup is a one-time step per workspace, connect each project's tools once. Keep meeting summaries structured and named consistently so your agent can always reference the right project's documents without cross-contamination.