Quick answer. A well-scoped first automation for an Australian SME typically runs 2-3 weeks from kickoff to going live: a discovery call and process mapping in the first few days, a clickable prototype and design review in week one, build and testing against real data in week two, and a supervised go-live with a human safety net in week three. The process that takes longer than 3 weeks is usually one that wasn't scoped tightly enough at the start. Here's what that first automation looks like, week by week.
Key takeaways
- A well-scoped first automation runs 2-3 weeks: discovery call and process mapping in the first few days, a clickable prototype in week one, build and testing in week two, and a supervised go-live in week three.
- In a real-shaped example, a Melbourne trade business cut its average inquiry-to-quote time from around 6 hours to under 30 minutes for standard jobs.
- By the end of week three, quotes for standard jobs were going out automatically in under 30 minutes, freeing up 10+ hours a week previously spent on quoting.
- A system that only works on the easy 80% of cases isn't ready to go live — it needs a clear, visible path for the other 20% to route to a human rather than guessing.
- Most Australian SMEs add a second automation within two to three months of the first one going live, because the second project moves faster once the team has seen the first one work.
Why We Time-Box the First Automation
The single biggest risk in a first AI project isn't the technology — it's scope creep before anything ships. Left unchecked, a "simple quoting automation" quietly grows to include the CRM overhaul, the new pricing model everyone's been meaning to formalise, and a reporting dashboard nobody asked for. None of that is automation's fault; it's what happens when a project doesn't have a firm shipping date.
Time-boxing the first automation to 2-3 weeks forces a narrower, sharper scope: one process, one measurable outcome, live in front of real use as fast as honestly possible. Here's what that actually looks like, week by week, using a real-shaped example — a Melbourne trade business automating lead intake and quoting.
Days 1-2: The Free Discovery Call
Everything starts with a free 30-minute call. This isn't a sales pitch dressed up as a discovery session — it's genuinely trying to answer one question: is there a process here worth automating, and is a 2-3 week first project realistic for it?
In our example business, the conversation surfaces that quoting takes 30-45 minutes per job, happens 25 times a week, and slow turnaround is costing jobs to faster-responding competitors. That's a strong first candidate — high frequency, high manual time, a clear and measurable outcome (turnaround time).
If the process isn't a good fit — too infrequent, too exception-heavy, or the data needed doesn't exist yet — we'll say so. Being told a project isn't right for this stage is a better outcome than three wasted weeks.
Days 3-5: Mapping the Real Process
Before any building starts, the actual process gets mapped in detail — not the process as described in the owner's head, but the process as it genuinely happens, including the messy bits. For the trade business, this means sitting with the office manager who currently handles quoting and walking through the last 10 real quotes: where the information came from, what took the longest, what went wrong last time, what "urgent" actually means in practice.
This step also settles the one number that matters. In this case: average time from inquiry to quote sent, currently around 6 hours (often longer overnight or on weekends), target under 30 minutes for standard jobs.
Week 1: Design and Clickable Prototype
With the process mapped, week one is spent designing the new workflow and building a clickable prototype — not a working system yet, but something real enough to put in front of the actual office manager and ask, honestly, "would you actually use this?"
For the quoting automation, the prototype shows: how a new inquiry gets captured, what information the system pulls automatically versus what still needs a human, and what the quote looks like before it goes out. Getting this in front of a real user this early — rather than after the system is built — is exactly the step that prevents the stalled-pilot problem we cover in why most SME AI projects stall after the pilot. Feedback from this session often reshapes the design meaningfully, and it's far cheaper to reshape a prototype than a finished build.
Week 2: Build and Test on Real Data
This is where the actual system gets built — connected to the tools the business already uses, not a demo environment. For our trade business, that means connecting to their existing job management software and pricing sheet, and building the logic that generates a quote from a job description.
Critically, testing happens against real historical inquiries, not a curated clean sample — including the awkward ones: the customer who described the job vaguely, the job that needs a site visit before pricing is possible, the request that arrived at 11pm on a Sunday. A system that only works on the easy 80% of cases isn't ready to go live; it needs a clear, sensible path for the other 20% that routes to a human rather than guessing.
By the end of week two, the system handles a defined majority of real cases correctly and routes everything else — clearly and visibly — to a person.
Week 3: Supervised Go-Live
The system goes live, but with a human safety net for the first stretch: every automatically generated quote gets a quick human glance before sending, or the system runs in parallel with the manual process for a few days so results can be compared side by side. This is the difference between "it worked in testing" and "we trust it enough to run unsupervised" — and it's worth not skipping, even though it's tempting to declare victory the moment the build is technically finished.
By the end of week three, quotes for standard jobs are going out automatically in under 30 minutes, freeing up 10+ hours a week previously spent on quoting, and the office manager is reviewing exceptions rather than building every quote from scratch.
The Week-by-Week Summary
| Timeframe | What happens | Who's involved |
|---|---|---|
| Days 1-2 | Free discovery call, initial fit check | Owner, Sketchli |
| Days 3-5 | Map the real process, agree the one number that matters | Process owner, Sketchli |
| Week 1 | Design the workflow, build and test a clickable prototype with real users | Process owner, team, Sketchli |
| Week 2 | Build the system, connect existing tools, test against real historical data | Sketchli, spot-checks from process owner |
| Week 3 | Supervised go-live, human safety net, measure against the agreed number | Process owner, team, Sketchli |
| Ongoing | Monitor, refine, and identify the next process to automate | Owner, team, Sketchli |
What Can Push This Beyond 3 Weeks
Being honest about the exceptions matters more than pretending every project fits neatly into a template. A few things reliably add time:
- The process genuinely isn't mapped anywhere, and even the person doing it can't fully describe it consistently — this needs to be resolved before building, not during.
- The data needed doesn't exist in a usable form — for example, pricing that lives entirely in someone's head rather than any system.
- Multiple stakeholders disagree on what "good" looks like — this needs resolving before the prototype stage, not after go-live.
None of these make automation impossible — they just mean the honest timeline is longer than 2-3 weeks, and it's better to know that in week one than discover it in week three.
Why the Compressed Timeline Matters
A short first project isn't about rushing — it's about getting a real, working result in front of the business fast enough that the next process becomes an easier conversation. Once the quoting automation is live and the office manager has 10 hours a week back, automating invoicing or onboarding next is a much easier internal sell than it was before anything existed to point to. For more on how AI transformation compounds over successive projects, see What is AI transformation?
FAQ
Is every process a fit for a 2-3 week timeline?
No, and it's important to say so upfront rather than promise a timeline that doesn't hold. High-frequency, well-understood processes like quoting, invoicing, and basic customer service queries usually fit. More complex, exception-heavy processes — like end-to-end project management for custom manufacturing — often need a longer first phase, or benefit from being broken into smaller pieces.
What does the business need to prepare before week one?
Mostly, access — to the relevant systems (CRM, accounting software, job management tool) and to the person who actually does the process day to day. You don't need clean data or a fully documented process; part of week one is figuring out exactly how messy the real process is.
What happens if the prototype feedback in week one is negative?
That's a good outcome, even though it doesn't feel like one. It's far cheaper to redesign a clickable prototype than a finished, connected system. A prototype that gets meaningfully reworked in week one after real feedback is doing exactly its job.
Do we need to pause our normal operations during the 2-3 weeks?
No. The manual process keeps running exactly as it does today throughout design and build. Nothing changes for the team until go-live in week three, and even then it typically runs supervised alongside the existing process for a short period before fully taking over.
What happens after the first automation goes live?
Ongoing optimisation — monitoring how the system performs against the agreed number, refining it based on real results, and identifying the next process worth automating. Most Australian SMEs we work with add a second automation within two to three months of the first one going live, because the second project moves faster once the team has seen the first one work.
Next step
If you want to see whether your business's highest-friction process fits this kind of timeline, run our free AI Readiness Check, or read more about how much AI automation costs before you start. When you're ready, book a free 30-minute call to scope your first automation.
Want to take your business to the next level with AI? Contact us or chat on WhatsApp.