For businesses that run on documents
Win the work.
Bill for all of it.
The writing that wins your work and gets you paid, from your own project history, your own rates, your own words, with a human sign-off trail.
Your prices, exactly as you set them. Anything it can't back up from your own documents gets flagged rather than guessed. And a person signs before anything goes out.
Files deleted within 24 hours. Nothing you write trains anyone's model. Leave any time and take all of it with you. What it costs.
Every figure carries the line it came from. The one it couldn't back up, it held back, and told you why.
In one line
Lunar8Ops is Australian document-drafting software for businesses that win and get paid on paperwork: it drafts tenders, quotes, variation claims and reports from your own rates and past documents, and a person approves every one before it sends.
Two ways this goes
Which one sounds like your week?
Same product underneath. The difference is how much of the setup you want to do yourself, and that changes what it costs and where you should start.
I want my paperwork handled
Quotes, reports, letters and follow-ups pile up, and you write them yourself at night.
I bid and claim for a living
Tenders, prequalification schedules, variation claims and payment claims are how you get paid.
Not sure? Take the first one. It costs nothing, and nothing about it rules out the other.
Start to finish
Twenty minutes to set up.
Then a minute a document.
The whole product, in five steps. There is no step six, and none of it needs anybody technical.
Set it up
Four documents you already have. Confirm about twenty numbers.
20 min, once
It drafts
You ask, or it starts on its own when the work arrives.
seconds
Numbers are enforced
A figure that contradicts your rates is removed before the model sees it.
automatic
A person signs
It goes to whoever approves. Approving is what sends it.
a minute
It stays current
A rate moves and it asks you. One question, one click.
now and then
Steps three and four
are the two a chat window doesn't have. Nothing reaches a client without passing both. Then step five hands you back to step two. And that loop is the whole product.
What makes it different
Four things we do differently.
Not because they are bad. They are very good, they are cheap, and on a blank page they will out-write this. They are built to be capable and general. Every one of the four below needs the opposite: something that knows your business, and knows when to stop and ask you.
It knows your prices, and sticks to them
A figure that disagrees with your rate schedule is removed before the model ever sees it. That is a comparison between two numbers, not the model being careful, so it cannot come out wrong on a bad day.
It tells you when it can't back something up
Ask a chat tool about an accreditation you don't hold and you get a plausible sentence, because a plausible sentence is what it was asked for. This flags what it cannot evidence from your own material instead of writing around it.
It gets started before you do
A chat window is a thing you have to remember to open, and 4:47 on a Thursday is exactly when nobody does. Point this at an inbox or a folder and it starts when the work arrives, not when someone gets to it.
Someone you choose signs every document
The draft goes to whoever should approve it, and what it was drafted from is recorded. When a client asks who signed off the number in that claim, there is an answer.
The long version, including what each of them is genuinely better at: the comparisons ↗
The part people don't expect
The 4:47pm email is the one that gets away.
Not because it was hard. Because it arrived at 4:47 on a Thursday, and the person who could answer it properly was on the tools, with a client, or in theatre. So it gets answered on Monday, in a hurry, or not at all.
Point it at your inbox or a folder and tell it in plain words what to act on: this sender, this word in the subject, anything with an attachment. When one arrives it reads it, does the work, and leaves the draft waiting. A chat window can't help here: it is a thing you have to remember to open, and remembering is the part that goes wrong.
Assistants, not people. Set up a shared assistant for a job rather than a person, one anyone in the office can run, with its own rules and its own approver.
Or from your own software. There is an API, and it plugs into tools that speak MCP, so it can be triggered by whatever you already run.
whensomething arrives from the site supervisor
dodraft the claim from my rate table
thensend it to the contracts manager to approve
neversend anything by itself
Rules match on exactly what you wrote, so you can predict what it will act on. It only ever touches things that arrive after you switch it on, it answers a conversation once, and a rule that keeps failing turns itself off and tells you why.
So the morning starts with three decisions instead of three jobs.
Try it. The buttons work.
- 6:12amprocurement@aveo.example.au
New enquiry: quote request
Priced from your rate card: 3 items, $4,880 inc GST, with the two exclusions you always list.
Edit firstNot this one - 9:47pmsam@northsidebuild.example.au
Re: this has slipped again
An unhappy client, answered the way you answer them: what happened, what you're doing, by when. No waffle up front.
Edit firstNot this one - 4:30pmtuesday-site-meeting.vtt
Follow-up from yesterday's meeting
Decisions, who owns what, dates. On the record while everyone still remembers agreeing to it.
Edit firstNot this one
Three pieces of work, already done, waiting on one decision each.
One job, start to finish
Where the figures came from, who signed, what landed in their inbox.
A single job followed through the three screens that matter. The business below is an example rather than a customer. Yours is built from your own documents, in your own format, on your own letterhead.
| Item | Rate | Certified from |
|---|---|---|
| Electrician: standard hours | $142.00 / hr | Rate schedule 2026, row 34 |
| Electrician: after hours | $213.00 / hr | Rate schedule 2026, row 35 |
| Imported fill & compaction | $2,180.00 / lot | Supplier quote, 14 Jul 2026 |
| Traffic control | $460.00 / day | Rate schedule 2026, row 12 |
A draft that contradicts a certified figure is corrected before the model sees it. That is a comparison, not the model being careful.
The asset
Your rates, with the row they came from
Every figure is certified against the document it was read out of, and the source travels with it into every draft. Nothing is typed in twice and nothing is guessed.
Variation claim VC-014: Lot 7 substation
waitingDrafted from Schedule of Rates · 4 sourced lines, 1 held back
Awaiting Dana R. (Contracts Manager)
Payment claim #38: Northside stage 2
approvedDrafted 08:14 · schedule due in 15 business days
Approved by Dana R. · 09:02
Prequalification response: Aveo panel
returnedDrafted from 6 past submissions · 2 claims flagged
Sent back by Sam K.: “use the 2025 ISO certificate”
Nothing sends itself. Every document carries who approved it, when, and against what.
The control
An approvals queue, not an outbox
Drafts wait for a named person. Approve, or send it back with a note that teaches it. The sign-off trail is what a head contractor's procurement team asks to see.
RIDGELINE ELECTRICAL
ABN 00 000 000 000 · Wacol QLD 4076
QBCC 000000
02 Aug 2026
Variation claim VC-014: additional excavation, Lot 7 substation
Further to site instruction SI-042 dated 28 July 2026, we claim the following variation to the contract sum. Rates are as per the schedule of rates annexed to the subcontract.
Ridgeline Electrical Pty Ltd · Payment terms 25 business days · This claim is made under the Building Industry Fairness (Security of Payment) Act 2017 (Qld)
The output
It comes out on your letterhead
PDF or Word, in the format your own past claims used, with your ABN and licence numbers where they belong. Not a chat transcript you reformat at 9pm.
How you sound, what you charge and the rules you would give a new hire all end up written down in plain English, where you can read them, correct any line and download the lot. It is not a model trained on your data. It is a document you own, and it gets thicker the longer you use it.
What it writes
If you write it more than once, it can write it.
9 document types are built in, with 60 more across 19 ready-made sets. Install one in a click, then change it to suit. Or describe the thing you write in a sentence and it builds that format, boxes and all. Everything exports as a PDF or Word file on your letterhead.
Quoting, claiming and reporting
- –Priced quotes and estimates from your own rates
- –Variation and progress claims
- –Tender and capability submissions
- –Inspection, defect and compliance reports
Client and case work
- –Referral letters and insurer reports
- –Progress notes and incident reports
- –Engagement letters and scopes of work
- –Monthly client summaries and reviews
Running the place
- –Minutes, decisions and owners within the hour
- –A written procedure from one walkthrough
- –Board and investor updates
- –The follow-up nobody gets around to
Worked through in detail for 11 industries, with the savings math shown: see yours ↗
Why you can leave it running
Careful by default.
Anything that drafts unattended has to be trustworthy in a way a chat window never has to be. This is the half of the product that does not demo well, and it is the reason this one can be pointed at a real inbox.
Numbers can't drift. A figure that contradicts your rate schedule is removed before the model ever sees it. Claims we can't evidence from your own material get flagged for you to confirm, and nothing leaves without a person approving it. The flag plus your sign-off is the control, not the model's good behaviour.
Nothing sends itself
Every outbound action (an email, a filed document, a calendar invite) waits in an approval queue. Approving is what sends it. That is the execution path, not a setting someone can leave off.
The right person signs
Drafts can route to an approver group rather than back to whoever triggered them, so a manager signs off what their team drafted.
Personal details never land
Anything arriving from a connected mailbox has personal details replaced before it is stored. A value that never enters storage cannot appear in a draft, however it is prompted.
There are 6 of these in total. The rest, the full data posture, the subprocessor list and the DPA are on one page you can forward to procurement.
Fair questions
The ones we always get.
We're already on Microsoft 365. Isn't this what Copilot does?
Keep it. Copilot is better than us at finding what is already in your SharePoint, and that is a different job. Ask it what your standard hourly rate is and it will find three answers from three different years and let you pick. This one won't, because a figure that contradicts your rate schedule is removed before the model sees it. Copilot also never notices anything on a Tuesday. It answers when asked. And when a client asks who approved the number in that claim, it has no answer. It was never built to have one.
How long does setup actually take, and who does it?
About twenty minutes of your time. Send two or three documents you have already sent clients plus your current price list; it reads roughly twenty numbers out of them and shows each one beside the line it came from for you to confirm. You are checking, not authoring. If you would rather not do even that, we do the whole week for you at a fixed price and you are in the room for one day of it.
What happens when our rates change?
It notices and asks you. Connect email or file storage and when a number it holds looks like it has moved, you get one question and one click. The old figure stays on record against the work already quoted at it, so a price change cannot quietly rewrite a job you have already sent. That is what stops this becoming another system somebody has to remember to maintain.
Isn't this just ChatGPT with extra steps?
A chat tool is something you have to remember to open, and it starts from nothing every time. This holds your rates, your reference material and your rules permanently. A figure that contradicts a price you have certified is removed before the model ever sees it. That is a comparison, not the model being careful, so it cannot come out wrong. Anything it can't evidence from your own material gets flagged for you to confirm, it watches the work arriving without being asked, and it routes what it produces to the person who should sign it. None of that is prompting. It is the part around the prompting.
We already use job management, practice management or an ERP. Where does this sit?
We don't touch your job system. We don't want your data out of it, we don't replace it. We do the writing it was never built to do. Your system owns the job, the schedule, the register and the ledger. It does not write the tender answer, the client report, the referral letter or the awkward follow-up, because it was never built to. There is no connector to any of them, and that is the guarantee rather than a gap.
What can it actually write?
9 document types are built in and there are 60 more in 19 ready-made sets covering trades, care, health, property, agencies, accounting, recruiting, sales and working for yourself. If yours isn't there, describe it in a sentence or hand it two examples of something you already write, and it builds that format, including the boxes you fill in each time.
Who owns what it learns?
You do. Everything it learns is written down in plain English where you can read it, change any line, and download the lot. Uploaded files are deleted within 24 hours of being read. Nothing you upload or create ever trains a model. Cancel and you keep everything; delete and it is gone immediately.
How do we know it's actually worth the money?
It keeps score, and it separates what it measured from what it estimated. It counts the documents it produced, prices them against assumptions you set and can change, lets you tick off the work you won, and emails you the summary monthly whether or not it flatters us. Where it had to guess, it says so. It won't state a win rate at all until enough outcomes are recorded to mean anything.
Can we run it on our own model provider account?
Yes, on the paid plans. Connect your own provider key and every request goes direct to them under your agreement, not ours. Your volume becomes a matter between you and your provider, with no cap of ours in the way.
Two ways to find out whether it works for you.
Both cost nothing and neither needs a card. Take whichever describes the week you have just had.
Quotes, reports and letters
Open an account and put one real job through it. Nothing to install, nobody to talk to.
Start free · no cardTenders and payment claims
Tell us your company name and we read what you have already published against the policy that is current. One page back within a business day.
Get a free gap report