Topics: Custom software & SaaS, AI automation
Want this done for your business? Get a fixed quote or book a free 30-minute call.
Quick answer: AI reads the enquiry (an email, WhatsApp message, call notes or photos) and drafts a quote by choosing items and quantities from your own catalogue. Ordinary code then calculates prices, discounts and tax, a person approves anything above a set value or anything the AI is unsure about, and accepted quotes flow straight into invoicing.
The slowest part of quoting is rarely the arithmetic. It's reading a rambling message, working out what the customer actually wants, remembering which package covers it and typing the same scope lines you've typed a hundred times. That interpretive work is exactly what language models are good at. The arithmetic, oddly, is what they're bad at.
So the design that works splits the job: AI for understanding, rules for numbers, people for judgement.
Which parts of quoting can AI take over?
The reading, structuring and drafting; not the pricing decisions.
| Step | AI | Rules in code | Person |
|---|---|---|---|
| Read the enquiry and attachments | Extracts scope, location, dates, constraints | ||
| Spot missing information | Drafts clarifying questions | Sends or edits them | |
| Choose line items and quantities | Selects from your catalogue | Checks items exist and quantities are in range | |
| Calculate prices, discounts and tax | Rate card, discount bands, GST | ||
| Write scope notes and cover message | Drafts in your house style | Reviews | |
| Approve and send | Applies value thresholds | Approves above threshold |
What data do you need before automating quotes?
A clean rate card and a memory of past quotes. The AI can only pick from what you give it, and it learns your style from examples. Before any build, gather:
- [ ] A rate card with every billable item, unit and price
- [ ] Standard packages and what each includes and excludes
- [ ] Rules for discounts, minimum charges, travel or rush fees
- [ ] A library of past quotes, marked won or lost, with notes on why
- [ ] Standard terms, exclusions and assumptions, as reusable text blocks
- [ ] Examples of tricky enquiries and how you priced them
This is often the most valuable part of the project. Many businesses discover their pricing lives in one person's head, and writing it down improves quotes even before any AI is involved.
Why should rates be calculated by rules, not by the AI?
Because language models can produce plausible-looking numbers that are wrong. They can slip on multiplication, round inconsistently, or "remember" a rate that changed last month. None of that is acceptable on a document a customer signs.
So the model never writes a price. It returns structured data, roughly "item code PAINT-INT-WALL, quantity 85, unit square metres, reason: three bedrooms described", and your code does the rest: looks up the current rate, applies permitted discounts, adds GST at the right rate for the market (18% in India, 10% in Australia), rounds consistently and totals. If the model names an item that isn't in the catalogue, the code rejects it and flags the draft.
A useful test during the build: feed the system the same enquiry ten times. The scope notes may be worded a little differently each run, but the total should be identical every time. If it isn't, a number is leaking out of the model somewhere, and that has to be fixed before any customer sees a quote.
How does a quote move from enquiry to approval to invoice?
Through one connected chain, so nothing is retyped.
- Capture. Enquiries from forms, email and WhatsApp land in your CRM; our guide to connecting website forms to your CRM covers that step.
- Interpret. The AI extracts scope and checks for gaps. If details are missing, it drafts questions for the customer; automating lead follow-up explains how to keep those replies prompt.
- Draft. Line items are selected, priced by rules and assembled into your quote template.
- Approve. Drafts under your threshold go straight to a quick review; larger or low-confidence drafts go to a senior person.
- Send and sign. The customer receives a link to view, accept and e-sign.
- Convert. Acceptance creates the deposit invoice in Xero or Zoho Books and the job in your operations system. Our guide to AI bookkeeping automation with Xero picks up from there for Australian businesses.
Sales teams can extend the same chain to call notes and CRM updates; see our guide to AI for sales teams.
Should you use an off-the-shelf quoting tool or a custom workflow?
Off the shelf if your services fit its model; custom when pricing logic or system connections are the hard part.
| Consider | Off-the-shelf quoting tool | Custom workflow |
|---|---|---|
| Pricing | Standard items and packages | Complex rules, measurements, bundles |
| Inputs | Mostly web forms | Email, WhatsApp, photos, call notes |
| Systems | Its own CRM or a few integrations | Your existing CRM, accounts and ops tools |
| Speed to start | Days | Weeks |
| Ongoing cost | Subscription per user | Build cost plus maintenance |
A focused custom quoting workflow typically sits in our simple-tool range: ₹1.5–5 lakh in India or A$10k–30k in Australia, plus around 15–25% of the build cost each year for maintenance. Try the subscription tools first; build only if they force you to price the way they think.
How do you stop AI adding wrong items or amounts?
Constrain what it can choose, check what it chose and show a person what changed.
- Allow only catalogue items; free-text lines are flagged for review.
- Set sensible quantity ranges per item and reject anything outside them.
- Ask the model to rate its own confidence and give reasons, and route low-confidence drafts to a person.
- Require approval above a value threshold you're comfortable with.
- Show reviewers the source enquiry beside the draft, highlighting each line's reason.
- Log every edit reviewers make; repeated corrections show where prompts or rules need fixing.
How do you measure whether it is working?
Measure before you switch it on, then compare. Track these for a month beforehand and every month after:
- Time to quote: from enquiry received to quote sent.
- Edit rate: the share of AI drafts changed before sending, and how heavily.
- Win rate: quotes accepted out of quotes sent.
- Pricing accuracy: quoted value compared with actual job cost once the work is done.
Faster quotes that win less, or win more but lose money, aren't progress. Our guide to the ROI of AI automation shows how to weigh the gains against the build and running cost.
Frequently asked questions
Can AI write quotes for my business?
Yes, as a drafting assistant. It can read enquiries, pick items from your catalogue and write scope notes in your style. Prices should come from your rate card through ordinary code, and a person should approve quotes before they're sent, at least until the edit rate is consistently low.
Is it safe to let AI set prices?
Not directly. Let the AI choose items and quantities, then calculate prices, discounts and tax with fixed rules. That keeps totals consistent and auditable, and stops a model from inventing or misremembering rates.
What software can automate quotes and proposals?
Dedicated quoting and proposal tools, CRM quoting modules and job management apps all automate parts of the process. A custom workflow adds AI interpretation of messy enquiries and connects to the systems you already use. Start with what your current tools offer before building.
How accurate are AI-generated estimates?
As accurate as your rate card, catalogue and past quotes allow, and only as good as the information in the enquiry. Track the edit rate on drafts and compare quoted values with actual job costs to find out for your own business.
Can AI-drafted quotes connect to Xero or Zoho?
Yes. Once a customer accepts, the quote's line items, customer details and tax codes can be sent to Xero or Zoho Books through their APIs to create the invoice automatically, so nobody retypes the figures.
Quote faster without guessing
If your team spends evenings turning enquiries into quotes, we can design the workflow, build the AI and rules layer and connect it to your CRM and accounts. See our custom AI automation work, or book a free call and we'll send a fixed quote, usually within one business day.