The answer
Mostly. Pulling deal details, assembling scope and pricing, sending and tracking can all be handed to software, and AI can draft a first version of the tailored summary from call notes. In our model about 70% of proposal time is automatable. The rep still owns the diagnosis and any pricing that breaks the rules.
The tasks, and how much of each is automatable today
Six tasks make up a proposal. Assembling scope, pricing and case studies from a template is the biggest block and about 75% automatable. Pulling deal data, sending, tracking and logging are 85 to 100% automatable. Writing the client's problem in their own words is half automatable. Discount approval stays mostly human.
| Task | Min / week | Automatable |
|---|---|---|
| Copy deal details and call notes into the proposalNames, company, goals and budget come straight from CRM fields. The 15% left is notes that live only in the rep's head. | 120 | |
| Assemble scope, pricing and case studies from templatesPackage rules pick the scope blocks and price table. Custom bundles still need a rep to arrange them. | 240 | |
| Write the client's problem and the proposed approachAI drafts well from a transcript. The rep must check it names the real problem, not the generic one. | 120 | |
| Get sign-off on non-standard discountsSoftware can route the request and show margin. The decision and the conversation stay with the manager. | 40 | |
| Send the proposal and track when it is openedProposal tools send, track views and alert the rep. Nothing to do by hand. | 36 | |
| Update proposal status in the CRMSent, viewed and signed events write back to the deal automatically. | 30 | |
| Whole process | 586 |
Hours saved per week
For a three-rep agency sending 12 proposals a week, proposals take about 10 hours of rep time by hand. Automating the tasks above returns about 7 hours a week, close to 340 hours a year. That is roughly a full extra day of selling time each week across the team.
Modelled on: A marketing agency with three reps sending about 12 proposals a week, each built by hand from an old proposal in Google Docs.
- 12 proposals a week, each taking about 45 to 50 minutes by hand.
- Most proposals reuse one of three service packages with standard pricing.
- About one proposal in four asks for a discount that needs a manager's sign-off.
- Call notes or transcripts exist for the discovery call, so AI has something to draft from.
Tools that do the job
You need a proposal tool with templates and variables, a CRM that holds the deal data, and something that joins them. PandaDoc or a Google Doc template covers the document. HubSpot or Pipedrive holds the data. n8n, Make or Zapier fills the template, and an AI agent drafts the summary from the call transcript.
- PandaDoc
- Proposal templates with variables, pricing tables, view tracking and e-signature.
- HubSpot
- Holds deal value, package and contact data that fills the template.
- Fireflies
- Records and transcribes the discovery call so there is real material to draft from.
- AI agent
- Drafts the problem summary and approach in the client's own words from the transcript.
- Make
- Watches for the 'proposal' stage, fills the template and writes status back to the CRM.
- Slack
- Routes discount requests to the manager with deal size and margin attached.
A simple workflow
The workflow starts when a rep moves a deal to 'proposal'. Software pulls the deal record and call notes, picks the right template, fills scope and price, and asks AI for a draft summary. The rep edits the summary, approves, and sends. Opens and signatures then update the CRM without anyone touching it.
- TriggerRep moves the deal to the 'proposal' stage. Pull the deal, contact and latest call transcript.
- AssembleChoose the template by package, fill names, scope blocks, prices and the two most relevant case studies.
- DraftAI writes a 150-word summary of the client's problem and approach, quoting their words where possible.
- ApproveRep edits and approves. If the discount is above the limit, the manager gets a Slack approval request first.
- SendSend from the proposal tool. Views and signatures update the deal and alert the rep.
- Hand onOn signature, start the contract and onboarding workflows so nothing waits for a person to notice.
Keep these human
- The diagnosis. A proposal that restates the client's real problem better than they could wins; a generic AI summary loses.
- Pricing outside the rules. Discounts are a negotiation, and the manager needs to own them.
Where it goes wrong
- Template sprawl. Twenty templates mean the automation picks the wrong one. Keep three packages and modular add-ons.
- AI summaries nobody reads. If reps approve drafts unedited, proposals start to sound the same. Require one edit before send.
- Stale prices. Hard-coded prices in old templates go out of date. Keep the price list in one place the template reads from.
What most guides miss
Most guides sell faster proposals. Speed matters less than what the proposal says. The useful automation is pulling the client's own words from the call transcript into the summary, because buyers sign proposals that describe their problem precisely. Template filling saves time; transcript-based drafting is what changes the close rate.
The evidence
- Sales reps spend 70% of their time on non-selling tasks, a share virtually unchanged from 72% in the previous survey.Salesforce, State of Sales, 6th edition (2024)
- Generative AI raised McKinsey's estimate of the technical automation potential of processing data from 73% to 90.5%.McKinsey & Company, The economic potential of generative AI: The next productivity frontier (2023)