Entry point
Fit Call
A 30-minute call on one problem you want to fix. If your platform's AI already does the job, we'll say so.
Design · Build · Support
From first use case to a system your team runs.
We turn manual processes into AI workflows, connect your systems and support your team from the first test to everyday use.
Prove → Build → Run
Prove
1–3 weeks
Test one job on your own files, or on public or synthetic data, against an error rate agreed before work starts.
Deliverables
- An hours map and baseline, with each job marked build, queue or park (Admin Hours Audit)
- A measured error rate on 30–50 documents or questions your team already knows (Trial)
- Every failure the test found, listed
Exit criteria
- You decide go or no-go against the error rate agreed at the start.
- A no-go ends the work, and you keep the answer key and the results.
Build
4–6 weeks
Build one package where you chose to run it, then test it on your own examples before anyone relies on it.
Deliverables
- The working system, in your own account, managed by us, or inside a platform you already license
- An acceptance test report against the thresholds agreed up front, labelled "acceptance test (not independent)"
- A one-page data note, a runbook and a handover
Exit criteria
- The acceptance test passes the thresholds agreed before work started. If it doesn't, you decide whether to continue.
- Your team has the runbook and has been through the handover.
Run
Ongoing, with the exit built in
Keep the system running: managed by us in an Australian region, or supported in your own account, in the cloud region you choose.
Deliverables
- A re-test when the AI model or a vendor API changes
- Maintenance, whether we manage the system in an Australian region or support it in your own account, in the cloud region you choose
- The exit pack: code, data, configuration and runbook
Exit criteria
- Managed: cancel with 30 days' notice and take the exit pack.
- Supported: the system already runs in your account, so your team keeps it running with the runbook.
Services
Entry point
Admin Hours Audit
One week mapping where staff hours go, what your existing software could take over, and what to build, queue or park. It sets the baseline you measure against.
Entry point
Two-Week Trial on Your Own Files
Your team supplies 30–50 documents or questions it already knows the answers to. You get a measured error rate, then decide go or no-go.
Entry point
Feasibility Sprint on public or synthetic data
The same measured test, run on public or synthetic data while your approvals are in progress. It ends in a go/no-go decision.
Entry point
Independent Evaluation
Tests of one AI system you bought or had built elsewhere, on your own questions, in your own environment. You keep the evidence and the kit to re-run it.
After an Audit or Trial
Fixed-Scope Build
One Document Registers, Knowledge Assistant or Draft-for-Approval package, built in 4–6 weeks, acceptance-tested and handed over.
Secondary
Enterprise Program
Three 4-week phases: prove, harden, hand over. Its Pilot to Production variant takes a stalled pilot, adds tests and a baseline, and hardens it for go-live.
Run
Managed in an Australian region, cancellable on 30 days' notice with an exit pack, or supported in your own account, in the cloud region you choose. Both re-test when the AI model or a vendor API changes.
Entry point
Evaluation Partner
Independent AI evaluation, delivered as a workstream under an existing contract held by a prime, internal-audit co-source firm, law firm or systems integrator.
Handover & capability transfer
Staff briefings, runbook training, and harness training for your risk and data teams.
Where to start
View starting options
| Buyer | Start with | Then |
|---|---|---|
| Government agency (state) | Independent Evaluation Report on Copilot Studio agents leaving pilot, a vendor AI upgrade or tender AI claims, ideally as an Evaluation Partner workstream | Repeat evaluations |
| Government agency (national) | Independent Evaluation, or the Feasibility Sprint on public or synthetic data | A Knowledge Assistant or Document Registers pilot |
| Local government | Feasibility Sprint on public or synthetic data, or a small evaluation | Policy & Procedure Assistant |
| Mid-tier regulated financial institution | Model-Change Regression or a pre-contract Vendor AI Evaluation on one use case, as a workstream under your internal-audit co-source or law firm | Harness training for your second-line team |
| Accounting practice (50–300 staff, or a network) | Accounting Safe-Use Kit (Australia), then a Trial of the Trust Deed & Client Structure Register (outside your peak tax season) | The full register; ATO Letter Sorter (Australia, on request); Run |
| Residential property-management group (1,000+ managements) | Admin Hours Audit + Switch-On | A Trial of the Management Agreement & Rent Roll Register (Australia); on-request property packages |
| Mid-size law firm | Vendor AI Evaluation on your own matters | Ask the Firm Manual (on request); Private Workspace (on request) |
| Manufacturer (50–1,000 staff) | Trial of the SOP & Work-Instruction Assistant on one line | The full assistant; on-request Document Registers and Draft-for-Approval packages |
| Health service or GP group | Scribe & Clinical AI Evaluation | Policy & Procedure Assistant for non-clinical staff |
| Mid-tier miner, utility or renewables developer | Trial of the Approval & Permit Conditions Register on one approval (on request) | SOCI AI Risk Evidence (Australia, on request) |
| Independent school or group | Trial of the staff Policy & Procedure Assistant | AI Switch-On for the software you already use; an evaluation of AI tools you've deployed |
| Prime, internal-audit co-source firm, law firm or SI | Evaluation Partner workstream | Repeat workstreams |
Who you work with
View team roles
The engineers and evaluators who scope your work also build and test it.
Engineering
Builds the system where you chose to run it, and puts the code, prompts and index in your repository.
Evaluation
Writes the test set with your team, agrees the thresholds before testing, and reports every failure.
Handover
Trains your staff on the runbook, and your risk and data teams on the test harness.
Where it runs
Delivered remotely from Sydney; on-site visits by arrangement.
View deployment and data locations
(a) In your own Microsoft or AWS account
Runs in your own Microsoft or AWS account, in the cloud region you choose. Our team reaches it remotely from Sydney, under the data-transfer terms in your contract.
(b) Managed for you in an Australian region
We run it for you in an Australian region, with an exit pack of code, data, configuration and runbook. If you're outside Australia, your data is transferred to Australia under the terms in your contract.
(c) Inside a platform you already license
Runs inside software you already pay for, such as Microsoft Copilot, Xero or Dext, where the vendor sets the processing location. Where we build part of it, such as a retrieval index, that part runs in your own account in the region you choose, or in ours in an Australian region. Our team reaches it remotely from Sydney.
The region note
- Where inference actually runs, which can differ from where the platform is hosted.
- The provider's retention, abuse-monitoring and human-review settings.
- What happens when the newest model isn't available in your chosen region.
- For a platform you already license: the vendor's published processing location, or "not published", with its source and an as-at date, re-checked quarterly.
Independence policy
View our independence policy
- We never issue an independent evaluation of a system we built, configured or advised on for the same client.
- We take no resale margin or referral fees from vendors we evaluate, and we disclose any vendor relationship.
- Acceptance tests on systems we build are labelled "acceptance test (not independent)".
- Our sample report on our own demo system is always captioned as not independent.
- We stay vendor-neutral: any vendor technical certification is disclosed, and we don't resell under a vendor partner tier.
How we de-risk
View delivery safeguards
Acceptance tests
Thresholds are agreed with you before work starts, and the report shows every failure.
Go/no-go gates
Each step ends in a decision point, and you decide whether to continue.
Exit pack
Code, data, configuration and runbook, so you can move a managed system or run it yourself.
You own it
Your code, prompts, evaluation tests and index live in your own repository.
Questions
What happens on a Fit Call?
A 30-minute conversation about one problem you want to fix. If your platform's AI already does the job, we'll tell you, and you won't need us for it.
What if a Trial misses the agreed error rate?
You decide whether to continue. The go/no-go is a decision point, and a no-go ends the work with the answer key and the results in your hands.
Who owns what you build?
You do. Your code, prompts, evaluation tests and index live in your own repository, whether your team runs the system or we do.
Can you independently evaluate a system you built for us?
No. We never issue an independent evaluation of a system we built, configured or advised on for the same client, and our own tests on it are labelled "acceptance test (not independent)".
What happens if we stop a managed Run?
Give 30 days' notice and you receive the exit pack: code, data, configuration and runbook. We don't offer escrow, because the exit pack covers continuity.
Do you resell AI licences?
No. Switch-On configures features you already own, and we take no resale margin or referral fees from vendors we evaluate.
What don't you offer?
In-house adversarial testing, a fractional AI lead, or AI policy-paperwork packages. A standalone AI platform or model gateway is on request only, as a Private Workspace or for a critical-infrastructure entity. We also don't rebuild what your software already does, such as bookkeeping capture, tenant reply drafting or KYC and ID verification.
Bring us one manual process.
Talk to us about a Fit Call