Book a scoping call →
Solutions by outcomeNever miss work26Every call, lead and enquiry answered, routed or captured.Get paid faster32Money collected, matched and reconciled without chasing.Fill the diary12Fewer no-shows, fuller weeks, less phone tag.Answer without staff16The repeated questions handled, the rest escalated.Join up the systems77Data entered once, moving everywhere it is needed.Verify and secure15Real customers in, fraud and fakes out.Prove you are compliant8Obligations met by default, evidence captured automatically.Run it for me12Someone watching, someone answering, every month.
Solutions by capabilitySMS & messaging28Voice & call flows22Verification & identity13WhatsApp & chat5Video & telehealth4Email & notifications4Flow automation3Unified conversations2Contact centre5Delivery & tooling2Compliance & records5Workflow automation49Trade & job management8Applied AI8Self-hosted platforms7Payments & billing29Run it for me4
All 198 builds Describe it in plain English Venture models Instant scope & price Free tools Search everything Connections Integration guides Workflow library By integration Partnerships Contract roles Honest comparisons Field notes Work we have built Colabs How it works Platforms & licensing Pricing Book a scoping call
Home / Guides / AI
Integration guide 6001 workflows7 min read

Automating AI in an Australian business

A language model step inside a workflow: classify, extract, draft or summarise, bounded to one decision.

Live example The copy-paste job, gone
  • TriggerForm submitted on your website at 21:14.
  • Step 1Validated, deduplicated against the last 30 days.
  • Step 2Customer created in the CRM, tagged by source.
  • Step 3Job drafted, owner assigned by postcode.
  • Step 4Acknowledgement sent to the customer.
  • 0.9sDone. Nobody opened a second tab.

What people actually automate with AI

The community workflow library holds 6001 workflows involving AI. Stripped of the novelty ones, they nearly all come down to the same thing: the judgement step: classify, extract, draft or summarise. The value is not the integration itself, it is that a person stops being the transport layer between two systems.

  • AI and Google Sheets appears in 2159 workflows. Usually the judgement step: classify, extract, draft or summarise, joined to a shared source of truth that stops being updated by hand.
  • AI and Gmail appears in 1515 workflows. Usually the judgement step: classify, extract, draft or summarise, joined to inbound triage and extraction.
  • AI and Slack appears in 981 workflows. Usually the judgement step: classify, extract, draft or summarise, joined to alerting the people who are already looking.
  • AI and Telegram appears in 944 workflows. Usually the judgement step: classify, extract, draft or summarise, joined to internal alerting.
  • AI and Google Drive appears in 764 workflows. Usually the judgement step: classify, extract, draft or summarise, joined to documents filed where they belong.
  • AI and Email appears in 288 workflows. Usually the judgement step: classify, extract, draft or summarise, joined to the manual step between two systems disappearing.

Three we would suggest first

Ranked by how often the pairing appears in the library, which is a better guide to what works than our opinion. Each one is a single workflow build: $700 to $2,500, then $290 a month.

[ BUILD 01 ]

AI to Google Sheets

Joins the judgement step: classify, extract, draft or summarise to a shared source of truth that stops being updated by hand. The version we build adds retries, alerting and a nightly reconciliation, so a silent failure cannot hide.

2159 workflows exist See them
[ BUILD 02 ]

AI to Gmail

Joins the judgement step: classify, extract, draft or summarise to inbound triage and extraction. The version we build adds retries, alerting and a nightly reconciliation, so a silent failure cannot hide.

1515 workflows exist See them
[ BUILD 03 ]

AI to Slack

Joins the judgement step: classify, extract, draft or summarise to alerting the people who are already looking. The version we build adds retries, alerting and a nightly reconciliation, so a silent failure cannot hide.

981 workflows exist See them

The workflows people actually use

These are the most-used AI workflows in the library. We do not sell them as-is: a template is a sketch, and the difference between a sketch and something you can run a business on is the part below.

  • AI Workflow Automation
  • WhatsApp Workflow with an AI Step
  • Telegram Workflow with an AI Step
  • AI Workflow Automation
  • Google Sheets Workflow with an AI Step
  • AI Workflow Automation

What we change before it goes live

A workflow engine holds the sequence, the retries and the credentials, so the logic lives in one visible place instead of scattered across four tools.

  1. Map the manual sequence, including who currently does each step
  2. Build it on your own instance so you own the automation outright
  3. Add queueing and retries so a transient failure never drops a record
  4. Alert on failure and version every change
The part most people skip: the failure path. We deliberately fail a step in staging so you can see exactly what happens to the record, who gets told, and how it is retried. That is the part you are actually buying.

Where these go wrong

Workflow tools make it easy to build something nobody can maintain. We version every workflow, document the field mapping, and keep credentials out of the canvas.

What it costs, and what happens to the unknown part

A single AI workflow is $700 to $2,500 to build and $290 a month to run, on your own instance. Where two or three workflows land together the build overlaps rather than doubling, and the managed fee steps rather than multiplying.

The part worth understanding is what happens to scope we have not built before, because that is where fixed-price work usually goes wrong. We do not fold a guess into the build price. Anything unrecognised is quoted as a separate, deliberately wide allowance plus a $1,500 discovery session, and the page shows you a budget figure that includes it. You plan against the number that covers the unknowns, not the number that flatters us.

The rest of the arithmetic: a single AI workflow is $700 to $2,500 to build and $290 a month to run, on your own instance. Where part of your request is not a known pattern, we price it as an allowance with a $1,500 discovery session rather than folding a guess into the build price. You can get a number for your own case in about a minute on the plain-English page.

Common questions

What can you actually automate with AI?

Anything the API exposes. The community library alone has 6001 workflows involving AI, and those are patterns other people have already used rather than ideas. Most requests we get are some version of the judgement step: classify, extract, draft or summarise.

What does a AI automation cost?

$700 to $2,500 for a single workflow on an instance that already exists, and $290 a month to run it. If part of what you want is not a known pattern, we add a clearly separated allowance and a $1,500 discovery fee rather than pretending we have priced it.

How long does it take?

Three to seven days for a single workflow, plus a day or two to stand up the instance if you do not have one. The variance is almost never the code; it is how fast credentials and an agreed field mapping arrive.

Do we own it?

Yes. It runs on your own instance, under your own accounts, and the repository is yours. The licence of the workflow engine expressly permits us to build and support workflows for you, and we do not resell it as a product.

Where to next

The AI integration page lists the workflows and what each one joins. If you would rather just say what you need, the plain-English page will scope and price it, including an allowance for anything we have not built before.

Want this for your AI setup?

Describe it in your own words and we will name the workflows, price the build and allow for the parts nobody has costed.

Describe it in plain English → See the AI workflows