In a small business, a sales proposal might take a morning. Someone asks the owner about the price, they agree and the email goes out. When five people are preparing proposals, the same process can leave a queue of questions waiting for one person.
AI might help with the writing. The proposals could still get stuck at approval. Before looking at draft generation, I would trace one proposal that recently reached a customer.
Where did the work stop?
Use the messages and documents to follow what happened. When did the request arrive? When was the draft ready? What was everyone waiting for after that?
The explanation may be quite ordinary. The salesperson did not know what discount they could offer. The owner needed to check costs, but a colleague kept them in a private spreadsheet. Everyone worked quickly on their part, while three days passed between steps.
This example gives you something specific to change. The people preparing proposals could get access to the cost information. The owner could agree a discount range for routine proposals and approve the exceptions.
Where would AI fit?
Once the information is available and the rules are clear, it makes sense to try generating a draft proposal. Whoever is responsible for the price still checks it. Unusual requests need someone who can make a decision about them.
After the change, I would measure the time from the original request to the email reaching the customer. If writing drops from an hour to ten minutes but the proposal still goes out three days later, the main problem remains.
Looking at work this way is useful even if you never add a new tool. It tells you which decisions the team can make themselves and where they need the owner involved. As a business grows, that distinction matters.