Local AI Automation Platform

UPDATE FOR EXISTING BIDDERS
Thank you for the replies so far. Several of you said the same thing: a paid discovery stage should come before a fixed build price. I agree, and I have changed this post to match. If you have already bid, you do not need to bid again. I am messaging shortlisted bidders with a scope guide.

WHAT I WANT BUILT
One dedicated, always-on PC in my office runs an AI operator. I give it an outcome in plain English. It plans the work, does it through APIs, browsers and desktop applications, checks the real result, and reports honestly.

Local AI is the default. Cloud AI is an option I can switch on per task type, under my own data and spending rules.

It must:
- Answer simple business questions from my own data, with sources.
- Take a goal with a budget and limits, and work inside those limits without step-by-step instructions. Limits are enforced in code.
- Keep each business in its own workspace: data, logins, rules and history. The design must grow past the first few workspaces on one machine.
- Operate websites and desktop applications that have no integration.
- Do rule-based admin from documents, descriptions and photos, and pass unclear cases to a person.
- Learn new procedures from instructions, documents and demonstrations, without a developer each time.
- Transcribe calls locally and show suggested answers on screen within a few seconds.
- Park a blocked task, keep working on others, retry later, and notify me on my phone when it needs a person.
- Provide spending limits, approval gates, an emergency stop, a full audit log, backups and a tested restore.

Fewer than 10 users. I buy the hardware after the discovery stage names the specification. Hardware budget is up to about A$15,000.

HOW THE WORK IS STAGED
0. Paid discovery: architecture, model choice, hardware specification, test plan, fixed quote for the rest, and a recorded proof run.
1. Core platform installed on my PC.
2. Web and desktop work, phone notifications and approvals.
3. Learning from demonstration and live call assistance.
4. Hardening, documentation and handover.

Each stage is a fixed-price milestone in AUD, paid when its agreed acceptance test passes. Tests use cases you have not seen. A model's own confidence score is not evidence.

WHAT I AM NOT ASKING FOR
I do not expect 100% autonomy on every unfamiliar task. I expect measured success rates and honest failure reports. CAPTCHA, MFA and identity checks pause the task and notify a person unless I have permitted an approved method.

WHAT YOUR BID SHOULD CONTAIN
1. A fixed price and duration for stage 0 only.
2. A rough price range and duration for stages 1 to 4.
3. Anything similar you have built that you can show me. This is not mandatory at this stage. Demos come later for shortlisted bidders.
4. Who writes the code, how many hours per week, and what else they are working on.
5. Whether your software runs on AMD graphics cards as well as NVIDIA.
6. The cost of an optional support plan: monthly, 6-monthly and 12-monthly.

I own all source code, configuration, data and documentation. The code lives in my repository from day one. Test accounts only during the build.

The intended result is a reusable local AI operator that performs directed work across multiple businesses, not a chatbot with a few fixed automations. Show More