Ask a partner how long a proposal takes and they will say two hours. Watch the process and you will see something closer to two days spread across a week, most of it waiting rather than writing.
This gap between perceived and actual time is where automation has a legitimate case. But not every step in the workflow should be automated, and the order matters.
The workflow, step by step
- A call happens and notes are taken, sometimes in a document, sometimes in a notebook.
- Someone researches the prospect: website, recent news, relevant prior work.
- A similar past proposal is located and copied.
- Sections are rewritten for this client, with scope and pricing adjusted.
- The draft goes to a partner for review, often with a comment like 'can you check the numbers'.
- It is sent, and the version history becomes unclear.
What to automate
The research and retrieval steps are high volume, low judgement, and easy to verify. That is the profile of good automation candidates.
- Transcribe and structure the call notes into a consistent brief.
- Gather public context on the prospect with sources attached.
- Retrieve the closest prior proposals and reusable sections.
- Assemble a draft skeleton with placeholders where human judgement is required.
- Route the draft to the right reviewer with the source material linked.
What to leave human
Scope, pricing and the promise you make to a client are judgement calls with commercial consequences. Automating the assembly is not the same as automating the decision.
Automate the preparation. Keep the commitment with a person.
The approval step is not overhead
In practice, firms that automate this workflow keep an approval queue for two reasons. First, it prevents a confident-sounding wrong draft from reaching a client. Second, it produces a record: who approved what, when, and what changed between the draft and the sent version. That record is valuable long after the proposal is decided.
Measure before you build
Take a baseline first: how many proposals per month, how long from call to sent draft, how many revisions, and how often the same sections are reinvented. Without that baseline, a pilot can feel successful and prove nothing. With it, you can decide in four weeks whether to expand or stop.