Pick by bid volume, spec complexity and who reviews. A handful of simple bids a month, checked line by line by you, suits an AI tool. Frequent bids, detailed specs or tender documents point to a part-time estimator. Most small firms end up with a hybrid: the tool measures, a person signs off every number.
The mistake is comparing a subscription with a salary. The comparison that matters is hours: how long a bid takes you now, how long it takes with a tool once you have checked it, and what your own hour is worth when you are the one checking. A tool that costs $119 a month and eats three evenings a month is not cheap, and a part-time estimator who saves you thirty hours may be.
The sum to do before comparing any prices
Four numbers decide this, and the subscription price is not one of them: bids per month, hours per bid by hand, hours per bid with a tool including the time spent checking it, and what an hour of the checker's time is worth. The tool route costs the subscription plus the checking hours. The estimator route costs the estimator's hours plus your sign-off. The hybrid costs both, with a smaller estimator bill.
Three invented firms show the sum. Each cell shows the monthly cost and, after the slash, the hours the owner spends. Assumptions: the owner's hour is worth $50 to $60 as stated, a part-time estimator costs $55 an hour, the tool costs the figure shown, and the owner spends half an hour signing off each bid that someone else prepared (a quarter-hour in firm C).
| Firm | By hand | AI tool, owner checks | Part-time estimator | Estimator using the tool |
|---|---|---|---|---|
| A: remodeller, 6 simple bids a month, 3.5 h by hand, 1.5 h with a tool, $119 plan, owner's hour $50 | $1,050 / 21 h | $569 / 9 h | $1,305 / 3 h | $764 / 3 h |
| B: mechanical subcontractor, 25 bids from drawings and specs, 5 h by hand, 3 h with a tool, $249 plan, owner's hour $60 | $7,500 / 125 h | $4,749 / 75 h | $7,625 / 12.5 h | $5,124 / 12.5 h |
| C: one-van specialist, 15 small quotes, 1 h by hand, 0.5 h with a tool, $119 plan, owner's hour $50 | $750 / 15 h | $494 / 7.5 h | $1,013 / 3.75 h | $719 / 3.75 h |
Read the table two ways. On dollars, the tool route is cheapest for all three. On the owner's hours, firm B cannot do it: 75 hours of checking is nearly half a month. The hybrid costs $375 more than the owner-checks route ($5,124 against $4,749) and gives back 62.5 hours, which is about $6 an hour for the owner's time. That is the real trade.
The subscription barely matters. A tool pays for itself once the hours it saves exceed its fee: 119 divided by (2 hours saved x $50) is 1.2 bids a month for firm A; 249 divided by (2 x $60) is 2.1 bids for firm B; and 119 divided by (0.5 x $50) is 4.8 bids for firm C. A one-van business doing three quotes a month should not pay $119. My rule of thumb: when checking would take more than about 20 hours of your month, someone else should do the checking.
Four estimating jobs and who does them better
"Estimating" is several jobs. An AI tool is strong at some and weak at others, and the weak ones are where money disappears.
| Job | AI tool | Part-time estimator |
|---|---|---|
| Quantity takeoff from clean plans | Fast; needs spot checks. Vendors' own scores differ by trade | Slower, but notices drawing oddities |
| Reading specifications, addenda and exclusions for scope traps | Weak to variable. Items specified but not drawn are the easiest to miss | Strong; this is the core of the skill |
| Pricing with your productivity rates and supplier prices | Only as good as the rate sheet you give it | Strong, if they learn your crews or ask |
| Access, sequencing, risk and the bid or no-bid call | Not its job | Helps; the owner decides |
One vendor's own wording is telling. Boon's page calls its product "your first AI estimator on staff" and says it "levels bids so your best estimator can focus on winning work". It publishes scores of at least 0.90 for electrical, 0.85 for architectural and 0.74 for mechanical drawings, and says "ask any vendor for theirs". That is a tool positioned beside an estimator. Handoff's pricing page shows another kind of limit: it lists AI takeoffs from residential plans up to 7,000 sq ft under its Scale plan. Know what the scope of a tool is before you build a process round it. For the bid-comparison side, which both routes still need, the tutorial on AI bid levelling and what it catches and misses is worth reading.
Bid volume and spec complexity on one grid
The bands below rest on the 20-hour capacity rule. A check takes about an hour on a simple, site-measured quote and about two and a half hours on a complex bid from drawings and specifications, so the same 20 hours covers very different numbers of bids. If your checks take longer or shorter, rescale the bands.
| Who does the checking | Simple scopes: site-measured, repeatable (about 1 h a check) | Complex scopes: drawings, specifications, addenda (about 2.5 h a check) |
|---|---|---|
| You check every bid, with an AI tool | Up to about 20 bids a month | Up to about 8 a month; outsource the complex ones one at a time, or use the tool for takeoff only |
| Someone else checks and you sign off | About 20 to 40 a month, with an office manager or foreman checking | About 9 to 20 a month, with a part-time estimator using the tool |
| An estimator does the work, using the tool | Over about 40 a month, with a part-time checker | Over about 20 a month, a part-time or full-time estimator |
Six triggers push a job into the complex column whatever the volume: addenda and revised sheets; alternates and allowances; scanned or hand-marked drawings; unit-price schedules; coordination between trades; and formal tender forms with compliance statements. If three or more apply, price the job with a person in the loop from the first hour. The tutorial on an AI bid or no-bid screen helps you decide which invitations deserve that effort at all.
What each route costs, from published prices
These are USD list prices from the vendors' own pages in October 2026, plus what two outsourcing firms advertise on their own sites. Several vendors publish none.
| Route | Price signal | Note |
|---|---|---|
| Manual digital takeoff | Metres.ai: $20 a month or $220 a year | Its page describes takeoff on uploaded plans; I found no mention of AI |
| AI takeoff tools | Exayard: $99 or $249 per user a month. BuildVision AI: from $299 a month. Togal.AI: Growth $299 per user a month billed yearly | Exayard's plans allow about 20 or 50 sheets per seat each month |
| AI estimating for residential contractors | Handoff: Flex $119, Pro $239, Scale $719 a month, billed annually | Seven-day free trial on Flex; plan takeoff listed under Scale, for residential plans up to 7,000 sq ft |
| Takeoff service with human QA | iBeam: done-for-you annual licence from $25,000 a year; do-it-yourself software from $8,000 a year per trade (HVAC, plumbing and steel only) | Says done-for-you takeoffs arrive in 24 to 72 hours, reviewed by its staff |
| Outsourced estimating firms | One firm advertises estimates "averaging $200" depending on project size; another gives $1,375 as its typical flat fee per estimate. Both list monthly packages from $1,500 | Asking prices on the two firms' own pages, not market rates. The second says its fees track drawing count, trade sections and document quality, not contract value |
| Part-time estimator you hire | Your own rate. The sums above use $55 an hour as an illustration only | Ask for a trial bid |
A gap that wide says "an estimate" means different things to different sellers: takeoff only or takeoff plus pricing, a small job or a full tender set. Run firm A through both. Six simple bids a month at $200 each is $1,200, plus the owner's three hours of sign-off ($150): $1,350, close to the $1,305 part-time route in the first table. At $1,375 an estimate the same six bids would cost $8,250, a price that only makes sense for full tender sets. Ask what a "bid" includes, how many revisions are covered, and who owns the file. For the broader price bands, see how much estimating software costs.
Then send every firm, and any part-time estimator you interview, the same one-page brief, so the prices that come back compare like with like. Filled in for an invented garage conversion, it reads:
- Job: single garage converted to a home office, about 260 sq ft; sheets A101 to A103, S101 (a steel beam) and E101.
- Scope wanted: quantities and prices for carpentry, drywall, finishes and electrical at our rate sheet, which is attached. The beam supply is an allowance line, not priced.
- Deliverable: our estimate layout, with assumptions and exclusions listed, and one revision included.
- Addenda: priced separately, and agreed with us before any work on them starts.
- Timing: bid due Thursday; draft back to us by Tuesday.
- Files: tell us how long you keep our plans and whether anything is used to train software.
A reply that prices only some of those lines is not cheaper, just smaller. Put each quote on the same scope before you compare the totals.
Test the tool against five of your own past bids
Before you commit to either route, run five finished bids through the tool and compare its total with your final priced bid. If the tool only does takeoff, compare quantities instead. Invented figures show a finished test.
| Bid | Tool total | Your final bid | Gap | Cause |
|---|---|---|---|---|
| Bathroom remodel | $14,200 | $15,900 | -10.7% | No floor protection or disposal; generic tile rate |
| Garage conversion | $38,500 | $41,200 | -6.6% | Did not know about the steel beam |
| Deck | $21,800 | $22,100 | -1.4% | Rounding in fixings |
| Kitchen | $46,000 | $52,300 | -12.0% | Cabinet allowance and an electrical upgrade |
| Roof repair | $9,400 | $9,200 | +2.2% | Slight over-allowance for waste |
| All five | $129,900 | $140,700 | -7.7% |
The result is not "the tool is 7.7% out". It is that the tool under-priced every job with hidden scope, structure, services or allowances in it, and matched the simple ones. Four of five bids came in below your price, so left unchecked this tool would have cost you margin on the jobs that were hardest to price. The rule that falls out: tool-only drafts are acceptable for repeatable scopes such as the deck and the roof repair, and anything with structure, electrical work or allowances gets a human read before it goes out. The tutorial on testing a remodelling AI estimator against your past jobs goes further into this method.
The sign-off: who reviews, and what they check
The test for a reviewer is simple: can they say "I would have built it this way" about every line? If the owner is the only person who can, the owner is the reviewer, and the volume rule above says how many bids that is sustainable. If a foreman or office manager can, they check; if neither can, a part-time estimator is the answer, not a better tool.
A one-page review sheet keeps the check honest. Filled in for an invented kitchen bid, it reads as follows.
| Check | What the reviewer does | Result on this bid |
|---|---|---|
| 1. Quantities | Spot-check the three biggest lines against the plans | Cabinets 18 units, matches. Tile 412 sq ft against 405 by room, fine. Drywall 668 against 640: a return wall was counted twice; corrected |
| 2. Scope gaps | List what the draft does not price | Disposal, floor protection, permit fee and panel capacity were missing; $1,870 added |
| 3. Rates | Compare labour hours per unit with your last three jobs | Tiling at 0.35 h per sq ft against 0.45 in your history; changed |
| 4. Access and risk | Stairs, parking, working hours, occupied house | Occupied house with second-floor access; protection allowance added |
| 5. Margin and decision | Compare margin with target; decide to bid or not | 18% against a 22% target after changes; bid with a note on the allowance |
One rule goes with the sheet: no bid leaves the building until a named person has written "checked" and the date on it. That sounds fussy until the first bid goes out without it.
A second reader for AI drafts: a prompt for missing scope
Whoever reviews can use an assistant as a second reader. Its job is to find gaps, not to price anything.
You are a second reviewer for a small contractor. Below is a draft bid summary produced by an estimating tool, and the list of sheets the tool was given. Do not rewrite the bid. Do three things: (1) list scope items that this kind of job normally needs that the draft does not mention (for example disposal, protection, permits, temporary works, access); (2) list sheets or documents a job like this normally has that are NOT in the list I gave you; (3) flag any line whose unit rate looks unusual against the other lines, and say why. If you are unsure, say "ask the owner". Do not invent prices.
JOB: 2,400 sq ft restaurant refit, occupied building, night work only.
SHEETS GIVEN: A101 floor plan; A201 elevations; P101 plumbing plan.
DRAFT LINES: [paste lines with quantities and rates]An illustrative reply would say that the draft does not mention a night-work premium, protection of neighbouring tenants, waste removal, temporary power and lighting, fire-stopping, or permit and inspection fees. It would note that a job like this normally comes with a reflected ceiling plan, a finishes schedule, a door schedule, mechanical and electrical plans and specification sections, none of which were given. And it might flag that the ceiling framing carries one flat-ceiling rate across the whole area, which is worth a question in a refit.
Treat the reply as a checklist generator and not an authority. Delete items that do not apply, chase the missing sheets, and remember that the assistant does not know your rates, so an "unusual rate" flag is only a prompt to look. Remove client names and addresses before pasting anything into a consumer chat account.
The first month's set-up work
Neither route is free to start. In the first month, my estimate is about ten hours to turn your last twenty bids into a rate sheet (labour hours per unit, markups and allowances), two hours to write an exclusions library you paste into every proposal, an hour to adapt the review sheet to your trades, and a few trial bids that you price twice, once by hand and once through the route you are testing. That is roughly fifteen hours. It is worth doing before you decide, because both a tool and an estimator are only as good as the rates and exclusions you hand them.
A hybrid week for a six-person firm
Take a six-person firm sending about five bids a week on a mix of remodels and small extensions. The week runs like this, with a part-time estimator for six hours.
- Monday. The office logs every invitation and sends a quick screen to the owner: bid, no bid, or ask a question. The tutorial on bid invitations and follow-up covers that log.
- Tuesday. The office manager runs takeoffs and drafts through the tool. At about 1.5 hours a bid, that is 7.5 hours for five.
- Wednesday. The part-time estimator reviews each draft for half an hour: quantities, scope gaps, rates. That is 2.5 hours, with the remaining hours left for the complex bid of the week.
- Thursday. The owner spends fifteen minutes a bid on the sign-off sheet and the margin call: about 1.25 hours.
- Friday. Bids go out and follow-ups are scheduled.
That is about 11.25 hours of drafting, review and sign-off for the five bids, or 14.75 hours if you count all six of the estimator's hours, against roughly 22.5 hours if each bid took 4.5 hours by hand. The owner's share falls to about an hour and a quarter. The real gain is not the hours saved. It is that no bid depends on one person's eyes alone.
The ceiling nobody checked
Picture a firm that used a tool to take off drywall for a restaurant refit. The tool priced every ceiling as flat, because the coffered ceiling in the dining room appeared only on the reflected ceiling plan, a sheet nobody had uploaded. The owner skipped the review because the draft looked complete and the last four had been fine. The bid went out about $4,800 short.
The review sheet above would have caught it at check 2, scope gaps, by asking "which sheets did the tool actually see?" Put that question on the sheet. A missing sheet looks exactly like a good result.
Judging the choice after three months
Log three numbers for every bid: the hours actually spent, the gap between the priced cost and the real cost on jobs you win, and the number of corrections the reviewer made. After three months, compare the log with your five-bid test. If corrections per bid keep falling, the tool or the estimator is learning your way of working. If they do not, change the input before you change the person: add rates, exclusions and past examples. And if your own checking hours are still above 20 a month, hire the human and keep the tool.
Interview questions for a part-time estimator
Whether you hire a person or buy per-estimate help, ask for a trial bid on a finished job and score it against your own result. Then ask these.
- Which takeoff and estimating software do you use, and can you work in mine or deliver in my rate-sheet layout?
- How do you deal with addenda and revised sheets, and what do you do when a drawing is unclear: ask or assume?
- What is your turnaround for a typical bid, and what happens in a deadline week?
- How do you show your assumptions and exclusions, so the file can go straight into a proposal?
- Who owns the rate sheet and the past-bid history, and what leaves with you if we stop?
Score each answer 0 to 2, giving a maximum of 10, and add up to 2 points if the trial bid lands within 5% of your own result on a finished job. A score under 8 out of 12 means keep looking. For the AI vendor, ask the same questions in different words: where do my rates live, can I see how each line was derived, what happens to my plans, and what can I export if I leave?
AI estimator or human estimator: common questions
Can an AI estimator replace a full-time estimator?
Not for tender work with specifications, addenda and alternates, where the costly errors are items that are specified but not drawn. For repeatable residential scopes, a tool plus an owner who checks every bid can replace a lot of the estimating hours. Test it against your own past bids before you decide, and keep a person signing off either way.
How many bids a month justify a part-time estimator?
My rule of thumb is capacity, not price: if checking the AI's output would take more than about 20 hours of your month, someone else should do the checking. At two to two and a half hours of checking a bid, that is around eight bids a month for complex work, and roughly twenty for small, repeatable quotes.
Does a part-time estimator need to use my software?
Not necessarily, but agree the hand-over format first. Ask for quantities and prices in your own rate-sheet layout, with the assumptions and exclusions listed, so the file drops into your proposal. A trial bid on a finished job shows quickly whether their output fits the way you work.
Is my plan data safe in an AI estimating tool?
Ask each vendor how long plans are kept and whether they are used to train the product, because the pages I read did not all say. If the answer is vague, run the trial on a redacted copy. Client drawings sometimes carry confidential details, so treat the upload like sending them to a stranger.
Further reads
- Can AI Produce Accurate Estimates for a Small Building Firm? — How accurate AI estimates are for a small building firm.
- AI Take Off Software: Trial Three Tools on One Real Plan Set — Trial three takeoff tools on one real plan set.
- Painting Estimating Software: Which Ones Use AI and What They Cost — Seven tools compared on measurement, photos, price book and proposals, with a worked monthly bill and the minutes each must save to pay off.
- Which AI Tools Should a General Contractor Buy? What to Skip — A buy-first order, a skip list with triggers, and a hours-it-must-save test for a general contractor, priced for a six-person remodelling firm.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: vendor pages for Handoff, Boon, Exayard, BuildVision AI, Togal.AI, Metres.ai and iBeam (including its pricing page), and the own sites of two outsourced estimating firms (tylerestimating.com and estimators.us). All read in October 2026.