An AI agent working on your data, in two weeks.
One workflow, one agent, running in your own Google Cloud project against your real systems. You finish with evidence of whether it works, what it costs to run, and what production would take.
Straight to a senior GCP architect. No SDR, no slide deck.
Clients include
A good fit if
- You have a specific workflow in mind that an agent could take on, not a general AI strategy question.
- The data and the systems the agent needs to act on exist and you can grant access to them.
- You want evidence before committing a platform budget or a production roadmap.
Probably not a fit if
- You are still deciding which problem to point AI at. Start with a call, not a build.
- The data the agent needs does not exist yet or would need a migration first.
Deliverables, not effort.
Everything is built in your environment and handed over in your repositories. Nothing runs on our infrastructure.
A working agent in your project
Deployed in your Google Cloud project, connected to the data sources and tools agreed at scoping, and usable by your team from day ten.
An evaluation set and scores
Test cases written with your subject matter experts, and the agent scored against them. You see where it is right, where it is wrong, and how often.
Cost controls
Budgets and alerts on the project, and the measured cost per task, so the run cost of production is a number rather than a guess.
Guardrails and human approval
Limits on what the agent can do on its own, with approval steps for actions that change data or reach customers.
Code and infrastructure you own
Agent source and Terraform in your repositories, with a short runbook for the team that will look after it.
A production readout
What worked, what did not, and a fixed quote for production rollout. If the answer is that it should not go to production, we say so.
Built with the Google Agent Development Kit on Vertex AI and Cloud Run. The model, Gemini or Claude on Vertex AI, is chosen against your evaluation set.
No surprises on the way.
Two weeks of build, with scoping agreed and access in place before the clock starts.
- Before we start
Scope and access
We agree the one workflow, what a good result looks like, and which systems the agent touches. You arrange access to a Google Cloud project and the data.
- Week 1
End to end, early
The agent is connected to your data and completes the workflow end to end, roughly. Your experts write the first evaluation cases against it.
- Week 2
Measure and harden
We improve the agent against the evaluation set, add guardrails and cost controls, and hand over the code, the scores and the production quote.
Not included
- Production rollout, which is quoted separately once you have seen the results.
- Ongoing operation of the agent after handover.
- Cleaning or migrating data beyond the sources agreed at scoping.
- Google Cloud and model usage charges, which are billed to your account.
What we need from you
- A named owner who can make decisions during the two weeks.
- Access to a Google Cloud project and the systems the agent uses.
- A few hours of subject matter expert time to write and review evaluation cases.
Questions we get asked
What happens if the agent does not work well enough?
You get the evaluation results either way, and they tell you why. Finding out in two weeks that a workflow is not ready for an agent is a useful result, and cheaper than finding out six months into a platform build.
Does our data leave our environment?
The agent runs in your Google Cloud project and model calls go through Vertex AI in the region agreed at scoping, including Australian regions. We do not copy your data to our systems.
Which model do you use?
Whichever scores best on your evaluation set at an acceptable cost per task. That is usually Gemini or Claude on Vertex AI, and the choice is shown in the readout with the numbers behind it.
Can we do more than one workflow?
Yes, as separate Agent MVPs or as part of a production rollout. Keeping each MVP to one workflow is what makes two weeks and a fixed price possible.
Fixed price, fixed date, before you sign.
Rapid MVP
4 weeks
A$16,000
Cloud Foundations
Fixed date, set at scoping
From A$45,000
AWS or Azure Migration
Set after assessment
Funded
Managed SRE and Agentic SOC
Ongoing
From A$5,000 / month
Prices are set in AUD; other currencies are fixed conversions as at 15 September 2026, shown for guidance, with the contracted currency confirmed before you sign.
Talk to an architect who has done this before.
Bring your current setup and the outcome you need. You will get a view on the approach, the risks and roughly what it costs.
Straight to a senior GCP architect. No SDR, no slide deck.
Not ready to talk? See how we migrated Hapana off AWS →
Or call +61 2 8359 9507 · Hello@aviato.consulting