Frequently asked questions.
The technical detail behind how Iceberg works: where the data comes from, how the matching and attribution are done, and what happens to your information.
What is Iceberg, technically?
Iceberg is a data service your AI assistant connects to over the internet, using MCP (Model Context Protocol), the open standard that lets assistants work with outside tools and data. Once connected, your assistant can do more than fifty new things: search every government contract, watch your renewals, match your contacts to buyers, hold your pipeline, qualify a tender, prepare a drafting pack. There is no Iceberg app. Your assistant is the interface.
Which assistants does it work with?
Anything that supports MCP, which today includes Claude, ChatGPT, Copilot and Cursor, with more arriving steadily. Everything works everywhere as text. If your assistant can draw charts and diagrams, Iceberg's data can also come back as market maps and dashboards.
How do I set it up?
Your organisation gets a private access key and a short setup guide. Adding it to your assistant takes a few minutes, once, and then it is there for good. A one-click connection is on the way. Setting up your market needs no configuration at all: you describe what you sell in a sentence and refine it in conversation.
Where does the data come from?
Three public sources, refreshed nightly: every contract notice ever published on AusTender, including amendments; every live tender, which we keep permanently even after AusTender removes it; and the public directory of government organisations, roles and appointments. On top of that sits the data only you can see: your contacts, companies, pipeline, notices and notes.
How do you know who owns a contract?
Honestly: it is an inference, and we present it as one. AusTender does not name a responsible person for any contract. Iceberg attributes each contract to the division or branch that owns the work by connecting the contract to the government directory, which succeeds for roughly nine in ten contracts, and presents the head of that unit as the most likely owner. When it cannot resolve ownership it says so rather than guessing.
How does contact matching work?
Government people are matched to the directory by name, role and agency, with a confidence level, and your assistant asks you to confirm anything uncertain. A match is never silently assumed. People at companies are matched through their email domain to the supplier register, which links one contact to that company's entire government contract book. You can also record people who appear in no public directory at all, on your own knowledge, and they count in every answer about who you know.
Do you integrate with my CRM?
No, and that is deliberate. Your assistant already has access to your CRM, spreadsheet or LinkedIn export through connections you control, so it reads your contacts and loads them into Iceberg itself. We host no integrations, hold no credentials to your other systems, and store only what gets loaded: names, roles, companies, and the notes you choose to keep.
Can our agents use it without a person?
Yes. Iceberg has no login screen an agent could get stuck at: anything your organisation connects, a teammate's assistant or a scheduled agent, uses the same connection under the same per-organisation fair use. One customer's agents qualify and rank every newly published tender on a schedule, so the team starts the day with a shortlist. Heavy automation counts toward your organisation's fair-use allowance, and Iceberg tells you before you reach it.
What about my other subscriptions: lead lists, tender alerts, meeting tools?
The same pattern. Whatever your assistant can read, it can load into Iceberg: companies from a lead service, deals from your CRM, notices from a tender alert subscription, summaries from a meeting tool. Each record is matched into the market on the way in, re-syncing updates rather than duplicates, and every record keeps a note of which source it came from. Whether a given service permits its data to be stored elsewhere is governed by your agreement with that service.
Does Iceberg write my tenders?
No. Your assistant does the writing, in your account, with your AI provider. Iceberg supplies what the writing should be built from: the incumbent and pricing history, the agency's buying pattern and vocabulary, and where you stand. Your own case studies and past bids stay in the systems your assistant already reads, and it drafts from both at once. Iceberg also instructs the assistant, in plain text you can read, never to invent capability or experience.
What do you store from what I load?
Only what your assistant chooses to save: contacts, companies, deals, notices, interaction summaries and notes. We never receive or parse your files, and transcripts stay in the tools that made them; Iceberg holds the summary. Nothing you load is ever pooled with other customers or used to help anyone else.
Who can see my data?
Your organisation, and nobody else. Every customer's data is isolated at the database level, hosted in Australia, and never used for AI training. Access is through your organisation's private key, which you can rotate or revoke.
What happens if I leave?
You can export your data at any time, and it remains yours after you go. If your subscription lapses, the connection pauses and your data is preserved for 30 days so you can return without starting over.
How is usage measured and billed?
Per organisation, not per person. Every tool call your assistant makes is metered at 20 cents. $50 a month is the minimum and includes your first $50 of use (250 calls); above that, calls are billed on your next invoice, up to the monthly cap you set yourself, $50 or anything above it. A call that fails is never billed, asking the same thing twice within a couple of minutes is never billed, and at your cap Iceberg stops and says so rather than billing past it. You can see the meter at any time by asking your assistant, or on your account page.
Something we haven't covered? Contact us