How can I automate sales quotes with AI?

You can use AI to read customer enquiries, collect the information you need and prepare draft sales quotes using your price list and templates. You review and approve them before they are sent, without starting from scratch every time.

An email arrives: “Could you send me a quote for…?” You open an old proposal, change the name, look up the right prices, update the services and check that none of the previous customer’s terms are still there. If that sounds familiar, you have already spotted a process worth considering for automation.

AI can help with the writing. But the bigger opportunity is connecting the steps: from the enquiry and customer details to the finished proposal and the CRM update.

Can AI create quotes from the emails I receive?

Yes, provided the solution is connected to your email and you have defined which enquiries it should process. It can identify what the customer needs, organise the information and prepare a first draft. When details are missing, it needs to flag them, not guess.

Here is a hypothetical example, not a result from an Ergonya client.

You run a technical services company. A customer asks for an equipment installation and mentions the location and their preferred date. They do not specify how many units they need or the technical requirements.

A useful workflow would:

1. Record the enquiry and link it to the correct customer, if there is a verified match.
2. Separate the information you have from the details that are missing.
3. Draft a short clarification email for review.
4. Once the details are complete, retrieve prices from the approved source and prepare the quote.
5. Show you the draft alongside the original request, ready for approval before sending.

The difference is practical: instead of gathering the information and writing from scratch, you start with organised material. You remain responsible for the commitments you make to the customer.

What do I need to prepare before I start?

You mainly need an up-to-date price list, a good quote template and clear rules about who approves what. You do not need to have a technical solution in mind.

Gather:

• A quote template: the sections it includes, your preferred tone and any terms that must always appear.
• Your current source of prices: a file, ERP or another system, with a clear owner responsible for keeping it up to date.
• The required details for each service: what you need to know before you can price the work.
• Approval rules: who reviews the quote and who can authorise a special discount or commit to a delivery date.
• A few representative enquiries: straightforward requests as well as cases that currently need clarification. Use synthetic or appropriately de-identified information for the initial test.

If your team uses three different price lists and nobody knows which one is current, resolve that first. Automation needs a clear starting point.

How do I make sure it uses the right prices?

Do not leave the language model to invent prices. Design the solution to retrieve them from a specific approved source and perform calculations using predefined rules.

The following split is a practical design suggestion:

What the quote needs	How to handle it
Understanding the enquiry	AI organises the text into specific fields.
Prices and calculations	The system uses an approved price list and rules.
A description of the solution	AI drafts the wording using your template and confirmed details.
A discount or special term	The authorised person makes the decision.
Sending	The system sends only after that version has been approved.

If a service is not in the price list or the system cannot access the source, the correct behaviour is to ask for help. It should not use a plausible price from an old proposal.

You also need to see where prices came from and which details changed. No design removes the need for testing and accountable review.

Will it send quotes on its own?

It does not have to. You can automate preparation and keep the final send behind an approval step. That is the approach we recommend for an initial pilot.

Human approval means the process pauses until the responsible person reviews, edits or rejects the proposal. Microsoft describes this mechanism in Power Automate: a review of AI-generated text, followed by the next action only in the approval branch. Microsoft Learn: using an AI prompt with human review.

Agree on three simple rules for your process:

• If the reviewer does not respond, send a reminder or notify a backup reviewer. Silence is not approval.
• If the price, scope of work or recipient changes after approval, review the new version again.
• If sending fails, check what happened before retrying to avoid duplicate emails.

The goal is to remove repetitive preparation, not oversight of your commercial commitments.

Can it connect to my CRM and the tools I already use?

It may be possible, but your specific tools need to be checked. Connectivity depends on the available integrations or APIs, access permissions and potentially your subscription plan. Do not assume that every tool connects to everything else.

Before choosing a solution, clarify:

• Where do enquiries arrive: email, a form or a manual entry by your team?
• Where do customer details and prices live?
• Where will you view and approve drafts?
• Which system will store the final quote and its status?

Think about failures too. If the CRM is updated but the email is not sent, the outstanding action needs to be visible and someone needs to be notified.

This is also the logic behind the enquiry-to-quote workflow described by Ergonya: collect information, prepare a draft, get approval and update the systems. The exact implementation depends on your business.

Do I need a custom app, or will an existing tool do?

If you prepare a small number of quotes using a consistent template, start by reviewing what you already have. A CRM with quote creation and approvals may cover much of the need without a new application.

Connecting existing tools may be enough when the steps are predictable. A custom solution is worth considering when you have unusual pricing rules, multiple roles, data spread across systems or a need for a dedicated review interface.

You do not necessarily need an autonomous AI agent. Anthropic distinguishes predefined workflows from agents that choose their actions dynamically and recommends starting with the simplest solution that meets the need. Anthropic: Building effective agents.

In plain language: decide how you want the process to work first. Choose the technology afterwards.

How will I know whether it actually saves time?

Measure the work required from receiving the enquiry to sending the quote, before and after the change. Include review and corrections, not just the time AI takes to generate text.

For an initial test, choose one type of quote and disable real sends. Test complete enquiries, missing details, special discounts, changes after approval and technical failures.

Record:

• How many minutes of active work your team spends on each quote.
• How long the customer waits, including time waiting for approval.
• How many drafts need substantial correction.
• Whether prices are correct and unapproved sends are blocked.
• The cost of usage, support and oversight.

Check the actual outcome: was the correct record created, the right field updated, the send blocked? The distinction between what AI says happened and what actually happened is also central to Anthropic’s approach to evaluation. Anthropic: Demystifying evals for AI agents.

If you save time drafting but lose it making corrections, improve the solution before expanding it. Freed-up hours are time available for other work, not automatically an equivalent reduction in expenses or an increase in sales.

Where should I start in my own business?

Take a recent quote and trace its journey: how the enquiry arrived, who gathered the details, where the prices came from and what delayed sending. That is a more useful first step than a list of AI tools.

If you want help deciding what is worth automating, the Ergonya Blueprint examines the process, the alternatives, a prototype and the business case before implementation. The recommendation may be an automation, a custom application or better use of a tool you already have.

Tell us how you create quotes today. Describe where enquiries come from, which tools you use and which step takes the most time. You do not need to know the technical solution or send sensitive customer information in your first message.