Watch the walkthrough on YouTube
Notion Custom Agents can draft proposals after sales calls by splitting the process into two focused workflows: one agent prepares the discovery call when a booking comes in, and a second agent drafts the proposal after the meeting is marked done. The proposal should still be reviewed by a human before sending.
The leverage is not fully automated selling. The leverage is eliminating the repetitive setup around meetings, CRM updates, proposal templates, dates, and follow-up drafts.
The sales process this solves
Consultants and agencies waste hours on sales admin that does not require strategy. A lead books a call. You create a meeting page. You check whether the contact is in the CRM. You research the company. You prep notes. After the call, you open the proposal template, copy over context, write the overview, estimate timeline, set expiration dates, draft the email, and then try to remember the follow-up.
Some of that work requires judgment. Most of it is structured movement between systems. That is exactly where a Notion Custom Agent can help.
Do not build one giant sales agent
The first architectural decision is the most important: use separate agents.
In the video, the proposal workflow is split into two jobs: a discovery call setup agent that runs when a call is booked, and a proposal drafting agent that runs when the meeting is marked done.
That split is reliability engineering. A single agent with too many steps has more ways to fail, loop, skip context, or create partial output. A focused agent is easier to test and easier to fix.
Agent 1: discovery call setup
The first agent runs before the call. Its job is to prepare the system so you are not walking into the meeting cold.
- Read the booking email from Cal.com or another scheduling tool.
- Extract the attendee name, email, company, budget, team size, and notes.
- Create a page in the Meetings database.
- Search the CRM for the contact and create one if no match exists.
- Research the company.
- Create a draft proposal page from a template.
- Link the meeting, contact, and proposal.
- Send a Slack summary with the relevant links.
That is the prep layer. It does not try to write the final proposal because the call has not happened yet.
Agent 2: proposal drafting
The second agent runs after the call. In Dave's setup, it is triggered when the meeting status changes to done. At that point, the meeting page should contain the call notes, transcript, or AI-generated summary. The agent can use that material to update the proposal draft with context from the actual conversation.
It can find the related proposal, review the meeting notes, fill in the discovery call overview, summarize the client's challenges, draft the proposed approach, add timeline and investment details when available, calculate proposal expiration and earliest start dates, draft an email for review, and leave a comment on the CRM page.
This is the right division of labor: the agent prepares the draft; the consultant reviews the substance.
Keep the CRM, meeting, and proposal linked
The article-level lesson is the relation chain: Proposal to Meeting to CRM. Once the records are related, the CRM becomes a history of the relationship. You can see the original inquiry, discovery call, proposal, comments, drafts, and follow-ups in one connected system.
A proposal workflow that only creates a document is weaker. A proposal workflow that updates the operating system is stronger.
Guardrails that matter
- Only trigger the proposal agent when meeting status changes to done.
- Do not run proposal drafting without meeting notes or a transcript.
- Search for the related proposal before creating a new one.
- Keep one proposal connected to one meeting unless you intentionally support multiple options.
- Do not invent budget, timeline, or scope.
- Leave evergreen proposal sections unchanged.
- Draft the email, but require human send.
- Add comments to CRM and proposal pages so the agent's work is auditable.
A proposal is a commitment. It contains scope, timeline, pricing, and expectations. The agent can save time by assembling a strong first draft, but you still need to check accuracy, fit, tone, and whether the scope is something you actually want to sell.
FAQ
Can Notion Custom Agents write complete proposals?
They can draft a proposal from structured meeting context and templates, but the final proposal should be reviewed by a human. Scope, pricing, and promises need judgment.
Why use two agents instead of one?
The triggers are different. One workflow happens when a call is booked. The other happens after the call is complete. Splitting them makes each agent easier to test and more reliable.
What is the biggest risk?
Letting the agent invent missing details. Proposal automation should never fabricate budget, scope, dates, or client priorities.
The Workcraft take
Proposal automation should not remove the consultant from the sale. It should remove the drag that keeps the consultant from selling. By the time the call is over, the proposal draft is ready for review, the CRM is updated, the email is drafted, and the next step is obvious.