AI Assistant

The AI Assistant is the front door to everything in Alloovium. It’s a conversational interface, but unlike a general-purpose chatbot it is always grounded in your workspace: your documents, your project’s live records — variations, RFIs, the programme, the timeline — your meetings and your connected email. Every answer links back to the passage or record it was derived from, so you can verify the source rather than take its word.

And it doesn’t stop at answers. The same ask bar takes instructions: draft a document, email the subbie, invite someone to the project, share a file with another company, flag an activity at risk, set up a routine that does the checking for you. Actions are never fired silently — the assistant drafts what it intends to do and waits for your confirmation.

Conversations are multi-turn: ask follow-ups, refine a draft, change your mind — the assistant keeps the thread’s context throughout.

Questions can be broad (“what are the key deliverables in this contract?”) or highly specific (“what’s the liquidated damages rate in clause 12?”) — and they aren’t limited to documents. The assistant answers from every source the project keeps:

Scope follows where you are: asking inside a project searches that project; asking from the company level — or saying “across all projects” — widens to everything you have access to. You can narrow too: name a folder and the answer draws only on that folder and everything beneath it. Retrieval is permission-aware throughout — the assistant only ever reads what you could open yourself.

When the follow-through is a write — a message, a meeting, a to-do, an invitation, a document share, a schedule risk flag — you can ask for it in the same sentence you’d use with a person: “text Wayne the crew count”, “invite Pat to the project”, “send the signed amendment to Lumino”. The assistant drafts the action, shows you exactly what will happen, and executes only when you confirm.

See Taking Action for the full list and how the approval flow works.

Ask for a document — a delay notice, an RFI response, an evidence pack, a position paper — and the assistant writes it from the project record, with a citation on every factual claim and your library templates matched automatically. The draft opens beside the chat, where you refine it conversationally (“keep it to one page”, “add a cost table”) and save it to the project when it’s right.

See Drafting Documents.

Anything you ask repeatedly can become a routine. Describe the recurring job in one sentence — “every morning, check the inbox for new claims and send me a summary” — and the assistant builds the routine: trigger, steps, conditions and delivery, laid out for your approval before it’s switched on. It will even notice repeated asks and suggest automating them with an Automate this chip.

See Creating Routines from Chat.

Every answer carries numbered citations linking to the exact evidence behind it. Document citations open the file at the highlighted passage; [T#] citations point at live records — timeline events, action items, meetings — and deep-link to where those records live. One answer can carry both kinds.

In addition to querying documents already in your project, you can attach files directly to a chat message — useful for a one-off question about a file without adding it permanently to the project. Supported formats include PDF, DOCX, XLSX, PPTX, CSV, TXT, PNG, JPG, and TIFF. Attached files are processed for the duration of the conversation and are not stored in the project unless you explicitly add them.

Speak instead of typing: click the microphone in the ask bar, dictate your question or instruction, and the words appear as text you can edit before sending — particularly handy on site or for longer briefs. See Voice Input. For hands-free work away from the screen entirely, the assistant also answers site calls and SMS.

Chat threads can be shared with other members of your team. Recipients see the full exchange — questions, answers, and citations — read-only, without re-running the queries. Use the Share button at the top of the conversation, or the conversation's three-dot menu in the sidebar.

For complex queries that take longer — synthesis across a large number of documents, multi-step questions — you can run them in background mode. Submit the question and go work on other things (or close the tab) while the assistant processes it. You’ll get an in-app notification when it completes, and optionally an email, so you can pick up where you left off.