How Construction Companies Can Automate RFQ and Customer Emails with AI

There is a quote request sitting in your estimating inbox right now. It came in as an email with a PDF attached, from a GC you'd like more work from, and nobody has opened it because your estimator is walking a site and the office manager is chasing a permit.

That gap is what this post is about. Construction companies rarely lose bids because they can't do the work. They lose them in the days between an email arriving and a person getting to it.

Disclosure: we make InboxPilot, an AI email assistant that drafts replies to RFQs and customer emails inside Gmail and Outlook. This post explains the workflow with sourced numbers, and it says plainly where a tool like ours does not belong: no AI should be sending your final pricing unsupervised.

TL;DR

• In 2018, FMI and PlanGrid's Construction Disconnected study found construction professionals spend 35% of their working hours on non-optimal activities, including 5.5 hours a week just hunting for project data. See why teams lose hours to email.
• Most quote requests still move outside bid portals. Loopio's 2026 cross-industry benchmark of 1,500+ response teams found only 45% of submissions go through portals; for subs and suppliers, what fills the gap is email. See the five workflows.
• You probably can't hire your way out. In the AGC's 2025 workforce survey, roughly four in five construction firms had unfilled salaried positions, the office roles that answer email.
• AI can classify, extract, draft, and route RFQ email, grounded in your own rate sheets and capability docs, with a human approving every send.

Why construction teams lose hours to email

In 2018, FMI and PlanGrid surveyed 599 construction leaders for their report Construction Disconnected and found that 35% of a construction professional's working hours, about 14 hours a week, went to non-optimal activities: 5.5 hours looking for project data, 4.7 hours on conflict resolution, and 3.9 hours dealing with mistakes and rework (FMI/PlanGrid, Construction Disconnected, 2018). The same report put the cost of that lost time at $177.5 billion in US labor for the year.

That study is old enough that we're stating its date, which is more than most articles quoting it do. The newer data suggests the problem grew. In 2021, Autodesk and FMI's follow-up study, Harnessing the Data Advantage in Construction, surveyed more than 3,900 industry professionals and estimated that bad data may have cost the global construction industry $1.85 trillion in 2020 (Autodesk + FMI, 2021).

And where does project information actually sit? A lot of it sits in email. In 2025, Microsoft's Work Trend Index measured the average worker receiving 117 emails a day and getting interrupted every two minutes during core hours (Microsoft WorkLab, Breaking Down the Infinite Workday, 2025). That's all industries, not just construction. Now add the construction-specific twist: the people receiving those emails are often in a truck, on a scaffold, or mid-pour.

So the inbox is where RFQs, bid invitations, sub questions, and document requests all land, and it's staffed by whoever happens to be at a desk.

The five email workflows that eat the day

Ask a construction office manager what's in the inbox and you'll get some version of the same five piles:

Here's the part most software guides skip: this traffic mostly does not run through bid portals. Loopio's 2026 RFP benchmark, built on data from more than 1,500 response teams across industries, found only 45% of submissions go through online portals, down from 51% the year before (Loopio, 2026). Public work and the biggest GCs use portals. Nearly everyone else works out of the inbox. That's why construction email automation has to start inside Gmail and Outlook, not inside another platform your customers won't adopt.

What AI can actually take off your plate

The useful question isn't whether AI works on email. It's which parts of the pile an AI assistant for construction companies can genuinely own. The adoption data says teams have already decided: in the same 2026 Loopio benchmark, nearly 80% of response teams had brought generative AI into their process, and in Autodesk's 2025 State of Design and Make survey of 5,594 industry leaders, 69% said AI will enhance their industry, a 12-point drop from the year before, so the enthusiasm is real but no longer naive (Autodesk, 2025).

1. Classify what just arrived

The AI reads each incoming email and tags it: RFQ, bid invitation, sub question, document request, follow-up, or none of the above. That sounds minor. It isn't. Triage is the step that currently requires a person to open and read everything, and it's the reason time-sensitive quote requests sit unread next to newsletters.

2. Extract the requirements

For a quote request, the assistant pulls out what the estimator needs to see in ten seconds: scope, site, dates, deadline, and what's missing. If the sender never attached the grading spec, the draft reply asks for it, which is exactly what your best office manager would do.

3. Draft the reply from your own documents

This is where generic AI tools fall down. The 2018 FMI study found 52% of rework globally traces back to poor project data and miscommunication, and an AI reply that invents an insurance limit or quotes last year's rates is a liability, not automation. Grounded drafting means the assistant writes only from documents you gave it: your rate sheet, capability statement, service area, insurance and licensing, standard exclusions, past quotes. When the answer isn't in those sources, it escalates instead of guessing.

4. Answer the questions you've answered a hundred times

"What are your insurance limits?" "Do you cover the north county?" "What's your lead time on formwork?" These have fixed answers that live in documents. They're the highest-volume, lowest-risk place to automate quote requests and routine questions, and they're where teams see time savings first.

5. Route the rest to the right person

Not everything should get a draft. Keyword rules send RFQs to the estimator, invoices to accounting, and scheduling to the super. Negative keywords do the opposite job: anything mentioning a claim, dispute, back charge, or change order skips automation entirely and lands in front of a human with no draft attached.

The workflow: from new RFQ to approved reply

Here's the whole loop for RFQ response automation, end to end:

1. An RFQ lands in estimating@ at 6:40am while the crew is loading out.
2. The AI reads and classifies it as a quote request, labels it, and flags the bid deadline it found in the body.
3. It checks your documents: services and capabilities, service area, rate sheet, insurance and licensing, and the last three quotes you sent this GC.
4. It drafts a response that confirms fit, states availability, asks for the missing plans, and commits to a proposal date.
5. A person approves, edits if needed, and sends. Total elapsed time: minutes, and none of them the estimator's.

What that looks like in practice:

Incoming, 6:40am: "We're bidding the Ridgeline distribution center and need a number for site concrete: footings, slab-on-grade, and dock aprons. Bid is due the 28th. Can you turn this around, and are you available to start in October?"

AI draft, 6:41am: "Hi Mark, thanks for the invite on Ridgeline. Site concrete is squarely in our wheelhouse and we have crews opening up in October. To get you a solid number before the 28th, can you send the structural drawings and the slab spec? We carry $2M general liability and can name you as additional insured before mobilization. You'll have our proposal at least two days ahead of the deadline."

Notice what the draft doesn't do. It doesn't invent a price, promise a start date the schedule can't support, or paste in boilerplate. Every claim in it traces to a document, and a person still reads it before it leaves the building. If your bids are 200-question proposal documents rather than emails, that's a different problem with a different toolset.

What changes when replies go out the same day

Faster response times. The classic data here is old, so we'll date it honestly. In March 2011, Harvard Business Review published an audit of 2,241 US companies ("The Short Life of Online Sales Leads") that found the average first response to an inbound lead took 42 hours, 23% of companies never responded at all, and firms replying within an hour were roughly seven times more likely to qualify the lead (HBR, 2011). That study measured web leads, not RFQs, so treat it as directional. The direction isn't controversial: the first credible answer shapes the shortlist.

Fewer missed opportunities. Sit with that 23%-never-replied figure for a second. GCs behave as if they know it: standard practice is to invite several subs per trade precisely because many invitations go unanswered. Being the sub that always answers, even when the answer is a fast and polite no-bid, is a cheap reputation to buy, and it's the one that gets you invited back.

Less admin work, which you couldn't hire for anyway. In 2025, the AGC of America and NCCER's workforce survey of nearly 1,400 firms found 92% of firms trying to hire had difficulty finding qualified workers, and roughly four in five had openings for salaried positions, the office and professional roles that answer email (AGC/NCCER 2025 Workforce Survey, 2025). Deloitte's 2026 Engineering and Construction Outlook projects the industry needs about 499,000 additional workers in 2026 and puts nearly $124 billion of output at risk from labor gaps (Deloitte, 2025). Automating the inbox isn't about replacing your office manager. It's about the second office hire you can't make.

Setting up RFQ email automation without risking a bad quote

Three steps, and the guardrails matter more than the features:

1. Connect the inbox. Whatever address the quote requests hit: estimating@, bids@, info@. Gmail or Outlook, and a shared mailbox works fine.
2. Feed it your documents. Rate sheet, capability statement, service area, insurance and licensing, standard exclusions, and a handful of past quotes you were proud of. This is the grounding step; the drafts are only as good as what you give them.
3. Set the guardrails. Trigger keywords like RFQ, quote, bid, and proposal route into drafting. Negative keywords like claim, dispute, back charge, and change order route straight to a person, no draft. Leave draft-and-approve on. The AI writes, you send.

That last guardrail isn't a setting we invented for this post. Drafts-with-approval is how InboxPilot works by default: the AI draft is the unit we meter every plan by, and sending stays a human act.

On InboxPilot, that setup runs on any plan: Free covers 25 AI drafts to test it on real traffic, Hobby is $29 per month, and Standard at $149 per month covers 1,500 drafts, unlimited inboxes, and knowledge base grounding, which is the feature doing the heavy lifting in step 2. Annual billing takes roughly 20% off. Full details on the pricing page.

See how InboxPilot helps construction teams respond to RFQs faster →

The construction page includes real draft examples for RFQs, subcontractor emails, vendor pricing, and scheduling requests.

Frequently asked questions

What is RFQ response automation?

RFQ response automation is software that reads an incoming request for quote, pulls out the requirements, and drafts a reply using your own documents: rate sheets, capability statements, insurance certificates, past quotes. The strongest setups keep a person in the loop, so the AI drafts and a human approves before anything sends.

Do construction RFQs really arrive mostly by email?

The best data we have says most arrive outside portals. Loopio's 2026 benchmark of more than 1,500 response teams found only 45% of RFP submissions go through online portals, down from 51% the year before. That benchmark is cross-industry, not construction-specific, but for subs and suppliers the channel filling the gap is almost always email.

Can AI send quotes without a human reviewing them?

It can, but it shouldn't. Pricing, scope, and exclusions carry real contract risk, so the sane default is draft-and-approve: the AI writes the reply, a person edits and sends it. Negative keywords add a second guardrail by pulling sensitive threads such as claims, disputes, and change orders out of automation entirely.

How fast should a contractor respond to an RFQ?

Same day is a reasonable bar, even if the real number comes later. Harvard Business Review's 2011 audit of 2,241 companies found the average first response to an inbound lead took 42 hours and 23% of companies never replied at all. That study measured web leads, not RFQs, but the direction holds: slow answers lose work.

What does construction email automation cost?

Far less than proposal software. Email-layer tools publish their pricing: InboxPilot starts free with 25 AI drafts, paid plans start at $29 per month, and Standard at $149 per month covers 1,500 drafts, unlimited inboxes, and knowledge base grounding. Enterprise RFP response platforms, by comparison, are mostly quote-only and often run five figures a year.

Related reading

• Best RFQ software for suppliers: response platforms compared, and why the email intake layer is a separate purchase.
• Best shared inbox software for Gmail and Outlook: where estimating@ actually lives, and how teams coordinate on it.
• Instant lead replies for sales teams: the same speed-to-respond problem, framed for inbound leads.

---

Statistics are attributed to their original source with the year of publication stated, including where that year is older than the article. The FMI/PlanGrid figures are from 2018 and the HBR response-time audit is from 2011; both are labeled as such above rather than passed off as current research. InboxPilot pricing was verified against the pricing page in July 2026. InboxPilot is the publisher of this article.