The prompt
Paste this into Claude, ChatGPT, Cursor, or any MCP-compatible client connected to your Bourd account. Replace the highlighted placeholders with your own values before sending.
Build me a quarterly content roadmap for the {{ workspace name }} workspace.
Look across every prompt. For each prompt where {{ your domain }} is absent or cited below the top five, estimate the size of the opportunity using response volume and how dominant the leading competitor is. Cluster the gaps into three to five topic themes, and for each theme propose one pillar page and two to three supporting pieces.
Return the output as a ranked backlog. For each item include:
- Theme
- Proposed pieces
- Estimated effort in t-shirt sizes
- Prompts the piece is expected to move
- Recommended sequence across weeks 1 to 12
Flag anything that would need original research versus what can be written from existing expertise. 2 placeholders to fill in. The agent will call: list_prompts, list_citations, list_competitors, and 1 more.
What this workflow does
Reads every prompt in your workspace and pulls citation data across every tracked engine. Flags every prompt where your domain is missing or ranked outside the top five sources. Groups those gaps into three to five themes, proposes one pillar piece (the main, in-depth article) and two or three supporting articles per theme, sizes the effort, and lays out a 12-week sequence with the reasoning for the order. The output is a ranked backlog. Pair the agent with a Linear or Notion connector and it writes the backlog there directly, or paste it into a sheet.
What "opportunity size" actually means
Three signals, in this order. First, how often the question gets asked. A prompt that has been answered 200 times matters more than one answered three. Second, how dominant the leading competitor is. One competitor cited 80% of the time is harder to displace than five competitors splitting the field. The fragmented field is the easier win. Third, whether neighbouring questions share a buyer concern, because one well-built pillar piece can move several at once. The agent shows the math behind each rank, so you can argue with the order instead of accepting it on faith.
The clustering judgement call
Clustering decides the shape of the output. It's also where the agent goes wrong most often. Ask for three to five themes. Fifteen is too many: the agent stopped clustering and just labelled each prompt. The right granularity: each theme is a topic one writer could become an expert on over a quarter, with at least three prompts feeding it. If a theme has only one prompt, fold it into the nearest neighbour or treat it as a one-off. Push back if the grouping looks soft. "Why are these two prompts in the same theme?" is a fair question, and the agent should point to a shared buyer concern or shared cited sources, not just topical overlap.
Why it works
The ranking logic is mechanical once the data is in front of you. The agent applies it consistently across every prompt, which is the work humans skip on a busy week. Bourd ships the raw inputs: which URLs are cited, how often each prompt runs, and which competitors win the source list. The agent does the cross-tabulation the dashboard skips.
Sequencing across 12 weeks
Highest-confidence wins go first. Pieces aimed at fragmented questions (where no single competitor dominates) ship in weeks one to four. The pillar pieces for the largest themes ship in weeks five to nine, because they take longer to write and they unlock supporting work in the back half. Supporting articles and refreshes fill weeks 10 to 12. Tell the agent to flag anything that needs original research or a survey, because those eat lead time you cannot recover. Flag dependencies too: a supporting piece linking to an unwritten pillar should ship after the pillar.
How to extend it
Chain into Linear or Notion so each backlog item lands as a ticket with the prompt list, target H1, competitor URLs, and recommended angle pre-filled. Writers do not need to open Bourd to start. Re-run the analysis monthly and diff the roadmap, so work that became moot (a competitor just shipped the same piece) drops off and the roadmap stays a living document. Pair with Google Sheets if leadership wants the backlog as a sortable spreadsheet.
When to re-run it
Monthly, lightly: ask the agent for the three biggest changes since the last run instead of a full rebuild. Quarterly, in full: rebuild from scratch, because the gaps in April will not be the gaps in July. Run it on the same cadence as your planning cycle. That's what keeps the roadmap honest.
Common failure modes
Two patterns to watch for. The agent over-weights large themes, ranking a five-prompt cluster above a three-prompt one even when the smaller cluster has buyers closer to a purchase decision. Counter it by adding 'score commercial intent separately, not just volume' to the prompt. The agent also reaches for pillar pieces where a supporting article would be cheaper. Five pillars in a quarter is unrealistic for most teams. Cap pillars at two per quarter and have the rest support existing pages.
Tools the agent calls
The agent picks from these 4 Bourd MCP tools based on the prompt. You do not call them directly.
list_prompts list_citations list_competitors list_insights
Pairs with
MCP agents chain across servers. Add these alongside Bourd and the same prompt can hand output straight to the tools you already work in.
Notion Linear Google Sheets Airtable