What does SkyBuild AI do?

SkyBuild AI identifies which business work should be automated, engineers the AI system that performs it, integrates it into existing operations, and measures the outcome against a baseline recorded before deployment.

How much does an AI automation engagement cost?

Every engagement begins with a $2,500 five-day mapping engagement, credited in full against implementation. Single agent deployments start at $6,000, multi-agent business automation at $18,000, and transformation programs at $50,000. Ongoing AI Operations starts at $1,500 per month. All figures USD.

What happens if the automation is not worth building?

If the mapping engagement shows a payback period beyond eighteen months, SkyBuild AI advises against building. The client pays for the mapping only and keeps the process map, business case and baseline. These decisions are published on a public No List with the underlying arithmetic.

Where is client data stored?

Australian client data is stored in the Sydney region with tenant isolation enforced at the database row level. Client data is never used to train a model. Client-facing inference runs only on paid providers under a data-processing agreement with zero retention configured.

Who owns the system after it is built?

The client. Agent and system code is handed over on disengagement, along with the data. There is no automatic renewal, and a ten business day cure period applies.

Trust

Your data, your IP, your exit

Healthcare and financial services clients ask these questions during procurement. The answers are published rather than negotiated, because a commitment you have to extract in a meeting is not much of a commitment.

Follow one patient enquiry through the whole system. Every point it is touched — and every boundary it never crosses.

A patient submits an enquiry through your form.

WHAT IT HOLDS

namecontactfree text

WHERE

Your infrastructure

Isolation

Enforced at the row, not in the application.

Most multi-tenant systems filter by tenant in application code, which works right up until someone forgets a WHERE clause. Switch tenant below and watch what the others can see.

authenticated as

clinic_arecord_1041readable

clinic_arecord_1042readable

clinic_arecord_1043readable

clinic_arecord_1044readable

clinic_brecord_1141— filtered by row-level security —

clinic_brecord_1142— filtered by row-level security —

clinic_brecord_1143— filtered by row-level security —

firm_crecord_1241— filtered by row-level security —

firm_crecord_1242— filtered by row-level security —

firm_crecord_1243— filtered by row-level security —

firm_crecord_1244— filtered by row-level security —

firm_crecord_1245— filtered by row-level security —

Filtering happens in the database, not the application. An application bug cannot leak another tenant's rows, because the query never returns them in the first place.

Regulatory context

Health information carries a higher bar, and we build to it from the first commit.

Health information is sensitive information under the Privacy Act 1988, with a materially higher threshold for collection, consent and handling than ordinary personal information. Retrofitting that after someone asks is not a plan.

Question	Answer
Where does our data live?	Australian region, isolated at row level per tenant
Who can see it?	Named engineers on your engagement. Access is logged
Is it used for training?	No. Contractually, and across every provider in the chain
How long is it retained?	Defined windows per data class, with automated deletion
What happens if we leave?	Code and data handed over. Ten business days to cure, no auto-renewal
Insurance?	Professional indemnity and cyber cover. Certificates on request

Leaving

What the exit actually looks like.

Most contracts leave this deliberately vague, because vagueness is leverage. Ours is a timeline you can read before you sign.

Day 0

You give notice

No penalty, no minimum term, no automatic renewal

Day 1–10

Cure window

If the issue is fixable, we have ten business days to fix it

Day 10

Handover

Agent code, configs, prompts, eval suites — transferred to your repo

Day 10

Your data

Full export in an open format. Deleted from our side on request

After

You keep running

The system is yours. It does not stop working because we stopped

Where the line is

Free model tiers never touch client data.

Free inference tiers are funded by your prompts. They carry no service agreement, no data-processing agreement, and quotas that change without notice. They are entirely appropriate for our internal tooling and entirely inappropriate for anything of yours.

Every client-facing system runs on a paid provider under a data-processing agreement with zero retention configured. It costs more. It is not a place to save money.

How the systems are built See pricing

Your data, your IP, your exit

Follow one patient enquiry through the whole system. Every point it is touched — and every boundary it never crosses.

Enforced at the row, not in the application.

Most multi-tenant systems filter by tenant in application code, which works right up until someone forgets a WHERE clause. Switch tenant below and watch what the others can see.

authenticated as

clinic_arecord_1041readable

clinic_arecord_1042readable

clinic_arecord_1043readable

clinic_arecord_1044readable

clinic_brecord_1141— filtered by row-level security —

clinic_brecord_1142— filtered by row-level security —

clinic_brecord_1143— filtered by row-level security —

firm_crecord_1241— filtered by row-level security —

firm_crecord_1242— filtered by row-level security —

firm_crecord_1243— filtered by row-level security —

firm_crecord_1244— filtered by row-level security —

firm_crecord_1245— filtered by row-level security —

Filtering happens in the database, not the application. An application bug cannot leak another tenant's rows, because the query never returns them in the first place.

Health information carries a higher bar, and we build to it from the first commit.

Question	Answer
Where does our data live?	Australian region, isolated at row level per tenant
Who can see it?	Named engineers on your engagement. Access is logged
Is it used for training?	No. Contractually, and across every provider in the chain
How long is it retained?	Defined windows per data class, with automated deletion
What happens if we leave?	Code and data handed over. Ten business days to cure, no auto-renewal
Insurance?	Professional indemnity and cyber cover. Certificates on request

What the exit actually looks like.

Most contracts leave this deliberately vague, because vagueness is leverage. Ours is a timeline you can read before you sign.

Day 0

You give notice

No penalty, no minimum term, no automatic renewal

Day 1–10

Cure window

If the issue is fixable, we have ten business days to fix it

Day 10

Handover

Agent code, configs, prompts, eval suites — transferred to your repo

Day 10

Your data

Full export in an open format. Deleted from our side on request

After

You keep running

The system is yours. It does not stop working because we stopped

Free model tiers never touch client data.

Every client-facing system runs on a paid provider under a data-processing agreement with zero retention configured. It costs more. It is not a place to save money.