Your Open Estimates Are a Gold Mine. Send an Agent to Work Them Tonight.

Open your field service software and look at the estimates that are still waiting for a decision. Start with those sent more than thirty days ago.

Some customers have hired someone else. Some have postponed the work. Others still need the job done, but the conversation stopped and nobody picked it back up.

Your team already invested time in those opportunities: answering the call, visiting the property, diagnosing the problem, and preparing the quote. Consistent follow-up helps you find out which ones are still worth pursuing.

An AI agent can help review eligible estimates, prepare relevant messages, interpret replies, and bring your team into the conversation when needed. With the right connections and rules, it can also send approved follow-ups and record activity in your system.

You can start tonight by reviewing a small group of estimates and testing the messages. Automated outreach comes after you confirm that the setup works.

Where does estimate follow-up break down?

An open estimate does not tell you why a customer has not approved the work. That is what follow-up helps you discover.

The homeowner may have questions about the scope. They may be waiting on financing, comparing options, or putting off a replacement while the existing equipment still runs.

Inside the business, the problem is often ownership. The technician moves to the next call. The office handles today’s schedule. Without a clear process, an estimate can sit untouched even when everyone intends to follow up.

AI can help keep that process moving. Your team still decides which opportunities deserve attention, what the business can offer, and when a person should respond.

What does AI add beyond an automated reminder?

A scheduled reminder follows a rule: if an estimate is still open after a certain number of days, send a message.

That may be enough for your first improvement. Check what your current platform already offers before adding another tool.

AI becomes useful when the workflow needs to interpret information. For example, an agent could:

• Draft a message using the approved description of the quoted work.
• Recognize that “We’re waiting until after the holidays” calls for a later follow-up.
• Flag a question about financing for the right person.
• Summarize the conversation so an employee can respond without rereading the entire thread.

The important question is which part of your follow-up process needs help. Start there, then choose the simplest setup that can do the job.

What needs to be connected before an agent can act?

For an agent to work directly with estimates in ServiceTitan, Housecall Pro, or another platform, your setup needs to support the actions you want it to take.

Reading an estimate, sending a text, recording a reply, and changing a status are separate capabilities. Confirm each one rather than assuming that a connection includes everything.

Before enabling automated outreach, establish:

• Data access: Which estimate details can the agent read, and how will it check the current status?
• Messaging: Which approved business email address or phone number will it use?
• Recordkeeping: Where will messages, replies, and handoffs be recorded?
• Ownership: Who receives a reply that needs attention, and how quickly should they respond?
• Controls: What stops a message if the customer approves, declines, opts out, or starts working with an employee?

If those connections are not ready, you can still begin with a staff-reviewed list and AI-drafted messages. That gives you a way to test the approach before automating more of it.

How should the workflow run?

1. Review the list before contacting anyone

Estimates older than thirty days can be a useful starting point for reviewing a backlog. Your normal follow-up process may need to begin much sooner.

Exclude estimates that are duplicates, already declined, associated with completed work, or being actively handled by someone on your team. Check for expired pricing and customers who should not receive further messages.

Also review estimates that present multiple options for the same job. Three quoted options should not produce three separate follow-up sequences.

An open status is a starting point for review, not automatic permission to send.

2. Set priorities that match your business

Choose a manageable group based on the work you want and can support.

You might start with recent replacement estimates, a particular service category, or customers who asked to revisit the project later.

Avoid treating the largest quote as automatically the best opportunity. Estimate age, customer history, current capacity, and whether the quoted work is still appropriate all matter.

3. Send a specific, neutral message

The message should identify the business, reference the work accurately, and make it easy to respond.

For example:

Hi [first name], this is [shop name] following up on the water heater replacement estimate we sent on [date]. Is that work still something you’re considering? We can have [team member] answer any questions or review the next step with you.

The agent should use customer-facing details your team has approved. Internal notes may include shorthand or information that does not belong in a message.

It should also avoid assuming that equipment has deteriorated, the customer is ready to buy, or an older price is still valid.

4. Answer only within defined limits

Give the agent a short list of questions it may answer using current, approved information.

That might include confirming which equipment was quoted or sharing an approved link to request a callback. Questions about changed scope, pricing, financing eligibility, technical recommendations, or availability should go to a person unless you have explicitly built and tested those capabilities.

For an older estimate, the appropriate response may be: “Our team can review the quote and confirm the current details with you.”

5. Hand off with enough context to act

When a customer wants to move forward, raises a concern, or asks something outside the agent’s instructions, notify a named employee.

The handoff should include the estimate, the conversation, the customer’s question, and any promised follow-up.

Pause automated messages while that employee handles the conversation. Customers should not receive another reminder after they have already asked your team to call.

What should the follow-up sequence look like?

Start with a short sequence your team can review and adjust.

First message: Confirm whether the work is still under consideration. Reference the estimate and offer help with questions.

Second message: Offer a useful next step. About a week later, ask whether the customer would like a call to review the scope or confirm current quote details. Include seasonal information only when it is accurate and relevant to that job.

Final message: Explain that reminders will pause. Let the customer know how to reach the team if they want to revisit the work.

For example:

We’ll pause our follow-up messages about this estimate. If you’d like to revisit the work, reply here or call [number], and our team can confirm the current details.

Only say you are closing an estimate if your process actually changes its status.

If the customer asks you to follow up next month, record that preference. If they decline the work or opt out, stop the applicable messages. Configure approved contact hours so a workflow prepared tonight sends at an appropriate time.

What rules should be written down?

Before the agent contacts customers, give it clear boundaries:

• No invented details. Use the estimate and approved business information. Missing information goes to the team.
• No unauthorized discounts or negotiation. Route price discussions to the designated employee.
• No unsupported promises. Do not guarantee dates, equipment availability, results, or warranty coverage.
• No technical diagnosis through follow-up messages. New symptoms or changed conditions need staff review.
• No misleading identity. Clearly identify the business and answer honestly if asked about AI.
• No continued reminders after a stop condition. Approval, refusal, opt-out, or employee takeover should trigger the appropriate pause or exit.

These instructions should govern what the agent can do, not just how it writes.

Test ten estimates before expanding

Choose ten estimates your team has reviewed and confirmed are appropriate for follow-up.

Have the AI draft messages first. Check whether each message describes the work correctly, uses the right customer information, and offers an appropriate next step.

Then test how the workflow handles replies such as:

• “We already had this done.”
• “Can you lower the price?”
• “Call me next month.”
• “Yes, we want to move forward.”
• “Stop texting me.”

Confirm that each reply triggers the right action, reaches the right person, and stops further reminders when required.

Once that works, run a small live batch during hours when someone is available to handle replies. Review the results before increasing volume.

How do you know it is working?

Track the estimates included in the test and the activity tied to them. Three measures provide a useful starting point.

Customer replies. Count the responses and what they tell you. A customer who confirms they hired someone else helps clean up your records, even though it is not a sale.

Approved work and scheduled jobs. Track which estimates move forward after outreach. Keep requests for a callback separate from approved estimates and confirmed bookings.

Completed work and collected revenue. Follow those opportunities through completion and payment. An approved estimate is not yet collected revenue.

Compare the results with your existing follow-up performance. Record staff time, tool costs, and any errors or complaints alongside the outcomes.

A booking after a message is useful evidence, but it does not prove the agent alone caused the sale. Consistent tracking will help you judge whether the workflow is improving results.

Start with the estimates you already have

Tonight’s first step is to pull the list and choose ten estimates worth reviewing.

Check their status. Confirm the quote details. Assign someone to handle replies. Then use AI to prepare messages your team can approve.

That small test will show you what needs to be connected, clarified, or corrected before the agent takes on more responsibility. It may also reveal that an existing reminder feature can solve part of the problem immediately.

Where does estimate follow-up tend to stall in your business: choosing whom to contact, sending the messages, or responding when customers reply?

Related reading:

• How to Set Up Your First AI Agent With Grok: A 5-Step Guide for Contractors
• Your Customer Just Asked AI Which Contractor to Call — Were You in the Answer?