// week by week
Eight weeks, no mystery
This is the standard shape of a first pilot build. Some finish sooner; complex integrations can run longer. Either way you’ll know at the end of each week what was done and what’s next.
-
Week 1
Discovery
We do: interview the people who do the work today, watch the process end to end, document every step, list the systems involved, and note where it stalls or gets re-done.
You provide: an hour or two with the people involved, and 20–50 real examples of the inputs — emails, forms, invoices — with personal details redacted where needed.
You see: a written map of the current process. Most owners find something in it they didn’t know.
-
Week 2
Analysis and scope
We do: rank the opportunities, estimate time and cost savings, design the build, and agree the success measure — minutes per item, turnaround, error rate — along with the before number.
You provide: a decision on what to build first, and who on your side owns it.
You see: a written scope: what’s in, what’s out, approval steps, data handling, running costs and the date we’ll measure.
-
Week 3
Connect and configure
We do: set up access with the narrowest permissions that work, connect the systems, and build the first version of the workflow against your sample inputs.
You provide: admin approval for the access we’ve listed — often your IT provider does this in a few minutes.
You see: the first outputs on your sample data, warts and all.
-
Week 4
Approvals, logging, edge cases
We do: add the approval step, write the logging, and work through the awkward cases — the invoice with two totals, the enquiry in the wrong inbox, the attachment that’s a photo of a page.
You provide: quick answers to “what should happen when…” questions. These become rules the build follows.
You see: a test set of inputs and the outputs you’ve accepted, which becomes part of the handover.
-
Week 5
Live, with training wheels
We do: switch it on for real work. Everything it produces waits for a person’s approval. We watch each run and fix problems as they show up.
You provide: the approver’s time, and honest feedback. “This draft is annoying because…” is the most useful sentence in the whole build.
You see: the work arriving already prepared.
-
Week 6
Tune and train
We do: adjust prompts and rules from the feedback, train the wider team, and write the runbook.
You provide: a short session with anyone who’ll approve or rely on its output.
You see: correction rates dropping week on week — or a clear explanation of why not.
-
Weeks 7–8
Measure and decide
We do: compare the live numbers with the before number, and write up what worked, what didn’t and what it costs to run.
You provide: a decision. Keep it as is, adjust it, extend it, or switch it off.
You see: the measured result and the full handover package.
— What usually slows a build down
- Access approvals. If nobody knows who the Microsoft 365 admin is, week 3 stalls. Find out before kick-off.
- Samples. We need real examples, not idealised ones. The odd cases are what break things.
- Moving targets. Changing the process mid-build is fine; changing it without telling us isn’t.
- No approver. If drafts sit unreviewed during week 5, we learn nothing and the build stalls.
— What it costs
A pilot build like this starts from $5,000 and typically sits between $5,000 and $15,000, depending on how many systems it touches and how messy the inputs are. If you’d like weeks 1–2 on their own first, that’s an opportunity audit, from $2,500. Prices are in AUD.