Every week there's a new app claiming AI will transform your business. Most of them slap a chatbot on top of a spreadsheet and call it a revolution. If you've been burned by tech hype before, your scepticism is well-earned.
So let's be completely straight about what Dockett is, what it isn't, and exactly where AI sits in it.
The short version
Dockett uses AI in three specific places. Everything else — the job tracking, the client re-engagement, the benchmarked pricing, the overdue invoice flags — runs on your own data, organised properly. No language model involved.
Where AI is actually used
1. Voice-to-quote extraction
When you tap the microphone button on a quote or invoice, describe the job out loud, and hit stop — that transcript gets sent to Claude AI (made by Anthropic) to extract structured line items. Descriptions, quantities, unit prices. It turns speech into a formatted quote draft before you've left the driveway.
This is genuinely useful. Not because it's AI, but because dictating is faster than typing when you're standing on a job site. The AI is just the mechanism that turns words into a table. You review everything before saving.
The text transcript of your voice recording. Your audio is transcribed on-device first — only the text leaves your phone. Anthropic processes it to extract line items and does not use it to train AI models.
2. Weekly insights copy
Once a week, Dockett generates a short plain-English summary of your business activity — revenue this week, jobs completed, outstanding invoices, how you're trending. This summary is written by Claude based on your anonymised business metrics.
The AI here isn't making decisions. It's translating numbers into a sentence or two that's easier to read than a table. The underlying data — what you invoiced, what you're owed — is all from your own Dockett records.
3. SMS draft suggestions
When Dockett suggests you follow up with a past client or chase an overdue invoice, it offers a pre-drafted message. That message is written by Claude based on the context: the client's name, the job type, how long it's been. You edit it if you want. You tap send when you're happy with it.
AI as a time-saver, not a decision-maker. You're still the one who decides whether to send it.
Where AI is not used
This is the larger part of the product.
| Feature | How it works |
|---|---|
| Client re-engagement triggers | SQL query on job history — finds clients not contacted in 10+ months |
| Seasonal outreach prompts | Pattern matching on your own past job dates — no model involved |
| Overdue invoice flags | Simple date comparison — due date vs today |
| Calendar conflict detection | Date range overlap logic |
| Benchmarked pricing | A curated data table of Australian market rates by trade, job type, and state |
| Job pipeline tracking | Structured database — status fields, timestamps, relationships |
| Profit per job type | Arithmetic on your invoiced amounts vs recorded costs |
| Referral prompt timing | Triggered when invoice.status changes to 'paid' — no AI needed |
The re-engagement logic that tells you "Sarah Smith is due for a follow-up — her bathroom reno was 11 months ago" is not AI. It's a database query. Same with the seasonal prompt that says "AC season is coming, you did 8 jobs last March." That's pattern matching on your own history.
We built it this way deliberately. SQL is deterministic, fast, and works the same way every time. There is no hallucination risk in a date comparison.
Why this distinction matters
Most "AI apps" in the trade space use AI as the product. The pitch is: "Our AI will manage your business." That's a marketing claim, not an engineering decision.
Dockett's design principle is the opposite: use the right tool for each problem. When a job needs deterministic logic — flag this invoice as overdue, find clients not contacted in 10 months — we use deterministic logic. When a job genuinely benefits from language understanding — turning a rambling voice note into structured line items — we use a language model.
The result is a product that's more reliable than an all-AI system, and more useful than a dumb database. Your re-engagement prompts fire correctly every time because they're based on real dates, not a model's interpretation of what "recently" means.
What Dockett actually is
The one-line version: a business operating system for Australian tradies that helps you win more jobs, charge correctly, and get paid faster.
The three places most tradie businesses leak money simultaneously:
- Past clients who'd rebook but never got followed up. Dockett knows the re-engagement window for every job type and tells you when it's time.
- Undercharging because there's no easy way to know the market rate. Dockett shows you what tradies in your state are charging for the same job.
- Slow invoicing and manual payment chasing. Voice note on site becomes a sent invoice in under a minute. Payment reminders go automatically.
None of that is AI hype. Most of it is just your own data, finally organised in a way that tells you something useful.
The honest conclusion
If you've tried apps that promised AI-powered everything and delivered a chatbot that couldn't remember your ABN — we get the scepticism. You're right to be tired of it.
Dockett uses AI where it earns its keep and nothing else. The voice-to-quote feature is the one you'll notice most — it genuinely saves time on site. Everything else is data doing what data does when it's properly collected and used.
We're built for Australian tradies who want their business to run better, not for investors who want to see the word "AI" in a pitch deck.
14-day free trial. No credit card. Make up your own mind.