Legal
AI disclosure
Draft for legal review. This page is a plain-language starting point and will be replaced by a counsel-approved version before any customer signs up. Questions: [email protected].
This is an AI system
Each govconagents.ai customer works with one or more AI agents: software that uses a large language model to read, search and write. They are not people. Customers give their agents names (our first customer calls its DIBBS Agent Brenda), but they are always AI, and every agent's email signature labels it as an AI assistant.
What the agents do on their own
- Read public government data (for example DIBBS, SAM.gov and award data) and screen it against the customer's rules.
- Read the emails and documents in the one shared mailbox the customer connects.
- Search the customer's records and public sources, for example for suppliers and contact details.
- Write drafts: RFQs and follow-ups, quotes, invoice packets, renewal answers, CPARS responses and proposal sections.
- Recommend decisions, such as a bid price, with their reasons.
- Answer questions in Microsoft Teams about the work they have done.
What people always do
- Review and send every email.
- Decide every bid, price, representation and invoice, and submit it in the government system with their own login.
- Approve any change to how an agent works. The agent proposes a change in Teams and applies it only after a named approver says yes.
The model and your data
The agents run on an AI model from OpenAI through its business API. Under those terms customer data is not used to train the model. On the Managed plan, model processing stays in the United States.
Known limits
AI output can be wrong. An agent may misread a document, pick an out-of-date contact or suggest a price that misses something. That is why drafts wait for review and every action is logged with its reason. Check important facts, especially part numbers, approved sources and prices, before you act on them.
Report a problem
If an agent does something wrong or unexpected, tell us at [email protected] or in your support channel. We review each report and fix the rule, playbook or code that caused it.