Revenue Desk: a deal desk that runs itself
A deal desk in n8n, Notion and the Claude API that answers the Monday pipeline questions before the meeting starts.
The problem
Every Monday a sales leader spends an hour asking reps the same questions. What is real, what slipped, what is at risk. The answers live in call notes nobody logged, news nobody read, and a forecast built by hand.
I wanted a desk that does that prep on its own and leaves the judgment to people.
What I built
- R1Brief before every callEvery morning it finds deals with a meeting today or tomorrow, reads the company website and the last 30 days of news, and writes a brief on the deal: what changed, what to ask, what to avoid.
- R2Call notes to CRMA rep pastes raw notes into the meeting record. Within 15 minutes it writes the summary, next step, risks and a suggested stage, updates the deal and saves a recap email as a draft. It suggests a stage but never changes one. Moving a deal is the rep's call.
- R3News risk on every open dealEvery Sunday it checks the news on each open deal and flags only events that change budget, timeline or who decides. Each flag says what it means, what to do and the source.
- R4Monday pipeline reviewCode computes commit, best case, weighted pipeline and coverage. Claude writes the narrative and picks the deals that need attention, each with one action for the owner. It is not allowed to calculate anything.
What broke and what I changed
First versionThe news check flagged 10 of 12 deals. An index listing counted as an opportunity and a lawsuit as a risk.
FixA strict bar: only news that changes budget, timeline or who decides. The default answer is "none". Now it flags 4 of 12.
First versionClaude retyped a close date in the review and wrote the wrong year.
FixDates now come straight from the data, and the prompt forbids writing any date.
First versionThe Monday review ignored the news checks, so risk sat in one place and the forecast in another.
FixThe review now pulls in the news flags. That is how it caught $105,000 of commit sitting on two at-risk accounts.
What the test run produced
On the 11 October run against 12 open test deals, the review showed commit of $198,000 against a $300,000 target, best case of $411,000, weighted pipeline of $215,300 and coverage of 1.74x. It named five deals that need attention, and pointed out that $105,000 of commit sat on two accounts with a new CFO or company-wide layoffs. A spreadsheet does not tell you that.
These are process results on test data. The companies and the news are real. The deals and the seller, Northwind Analytics, are made up, and every record is labelled TEST.
The trade-off
It is only as good as what reps log. If notes never reach the meeting record, R2 has nothing to work with and the review falls back to stale fields. News coverage is also thin for small private companies.
Hiring for GTM engineering or RevOps? Need a revenue system built?
Email is fastest. I work evenings India time, which overlaps the US morning.