Published · Updated
7 MIN READ
AI Proposal and Quote Automation: The Workflow
AI proposal and quote automation is the workflow in which AI reads the customer's request, separates items and conditions, pulls the price from the company's rules and drafts the proposal for a person to review before it is sent. AI handles the reading and the writing; the math still comes from the price list and from whoever approves.
The time this work takes shows up in sales research. Salesforce's State of Sales report, released in February 2026, surveyed 4,050 sales professionals in 22 countries, Brazil among them, and concluded that the average rep spends only 40% of their time selling. The rest goes to support tasks, and putting together quotes is one of the most repetitive.
At Wama, responding to a quote request is treated like any other process: it is only worth automating what has volume, clear rules and someone to check it. It is the same filter we use to decide which process to automate with AI first, applied to the path between the request and the response.
What does AI proposal and quote automation do?
The automation covers five steps, from the incoming request — after AI lead qualification — to the outgoing proposal. In each one, AI plays a different role, and in none of them does it set the price on its own.
Receives the request: email, website forms, attached spreadsheets or WhatsApp messages go into a single queue.
Reads and extracts: AI identifies the customer, items, quantities, deadline and requested conditions, even when the text comes in disorganized.
Builds the quote: the extracted items are matched against the price list, inventory and discount rules that already exist in the system.
Drafts the proposal: AI writes the summary of the request, the scope and the conditions in the company's official template.
Goes through review: a person checks price, deadline and scope before sending.
Step 1 depends on the channel. Anyone who receives many requests by message needs to look at the rules and costs of an AI agent on WhatsApp. Step 3 depends on what AI can read in the sales system, and CRM integration with AI is what sets that limit.
Where does the price come from when AI builds the quote?
The price always comes from the company's rules, never from the AI model. Price lists, discount policy, shipping and taxes live in a system or spreadsheet with an owner; AI looks up that data and shows the math, but it does not make up amounts or round margins on its own.
The division of labor is clearer side by side:
Part of the response | Who does it | Why |
|---|---|---|
Reading the request and listing the items | AI | Free text, varied formats, high volume |
Unit price, discount and shipping | Company rules | The same request must give the same result |
Proposal text and scope summary | AI, with an approved template | Repetitive writing that varies by customer |
Price or deadline exceptions | A person with approval authority | A commercial decision with risk |
That separation also protects the company from a silent mistake: a well-written proposal with the wrong number. When AI needs to look up price lists in different systems, the design of the connection matters as much as the model, and MCP for companies is one way to connect AI to that data with controlled permissions.
Human review and approval limits
Every automated proposal needs a checkpoint before it goes out, and the rule for who checks what should be written down before the first proposal is generated. Review that depends on common sense disappears in the busiest week.
The cases that call for approval from someone with authority are usually few and predictable:
a discount above the rep's limit;
an item outside the catalog or a custom scope;
a delivery deadline shorter than standard;
a new customer with no payment history;
a clause that differs from the approved template.
At Wama, every output produced with AI goes through human review before it reaches the client, and that is the rule we recommend for sales proposals. The warning also shows up in market projections: Gartner predicts, in a November 2025 press release, that by 2028 AI agents will outnumber sellers ten to one, but fewer than 40% of sellers will say they improved productivity. Gartner's own recommendation is to fix data and process before scaling. The rules on who uses which tool belong in the company's AI usage policy.
How to measure the gain before automating?
Measure the current process for two to four weeks before turning on any AI. Without that snapshot, there is no way to say afterward whether the automation saved time or just moved the work somewhere else.
Four numbers are enough to start: quote requests received per week, minutes spent on each response, time between the request and the send, and proposals that came back because of price or scope errors. Illustrative example, unrelated to any real client: a team that answers 40 requests per week and spends 50 minutes on each one adds up to about 33 hours a week on this task alone.
In Wama's method, this measurement is part of the AI Assessment, which produces the custom Implementation Plan and the baseline to compare before and after. What that work delivers is in the text on the AI assessment for companies.
Measurement also reveals the most common obstacle. In the same Salesforce survey, 51% of sales leaders using AI said disconnected systems slow down their initiatives. If the price list lives in a spreadsheet only one person understands, that is the first job in the stages of implementing AI in companies.
Common mistakes
The mistakes below show up on pages about the topic and in the design of the workflow itself. Almost all of them come from starting with the tool instead of the process.
Starting with the text generator. A beautiful proposal without the price list connected becomes checking rework, not a gain.
Letting AI calculate the price. A language model is not a rules engine. Price, discount and tax come from the system, with AI only showing the math.
An outdated price list. The automation repeats the price list's error in every proposal, faster.
Discounts with no owner. Without written approval limits, the exception becomes the rule within a few weeks.
Customer data in unapproved tools. In a Thomson Reuters survey reported by The Logic in June 2026, about 34% of 1,816 professionals in fields such as law, accounting and compliance said they use unauthorized AI at work. A quote request carries personal and commercial data, and the rules of LGPD (Brazil's data protection law) and artificial intelligence apply to every field AI reads.
Frequently asked questions
Can AI send the proposal without review?
It is not recommended. Even with the price coming from the price list, AI can misread an item, a deadline or a condition the customer asked for. Automatic sending only makes sense for standardized, low-risk responses with a written rule, and even then with samples checked every week by someone in sales.
Do I need a CRM to automate quotes with AI?
Not necessarily, but you need a single source of prices and rules. It can be the CRM, the ERP or a well-maintained spreadsheet, as long as it has an owner and a current version. Without that source, AI drafts proposals with numbers nobody can guarantee, and the time saved turns into manual checking.
How much does it cost to automate proposals and quotes with AI?
It depends on request volume, the systems that need to be connected and the AI model used at each step. The recurring cost usually comes from tokens, integrations and maintaining the rules. The full math, with exchange rates and taxes, is in the guide on how much it costs to implement AI in Brazil.
Where does Wama start a project like this?
Wama starts with the AI Assessment, the first stage of the journey, which shows where AI speeds up slow, time-consuming and error-prone processes and produces the custom plan. It is a small, paid step, and its fee is deducted from the plan's first installment when the company moves forward.
What kind of company gains the most from automated quotes?
The company that answers many similar requests, with a defined catalog and price list, such as distributors, manufacturers and recurring service providers. Custom projects, where each proposal is a new study, gain less from quote automation and more from assisted drafting of the text.
If your sales team loses hours putting together quotes and you want to know whether that process is ready for AI, talk to Wama about an assessment of your case.
