Aeglio AI
How to use Chat with Aeglio
Chat with Aeglio helps you explore the business data you are already allowed to see. It can find records, follow relationships and combine exact read-only calculations into a clear answer with links back to Aeglio.
The short answer
Open Chat with Aeglio, choose an available model and ask a concrete business question. Aeglio plans which approved read-only tools it needs, checks every result against your permissions and streams the answer. Record titles in the response link back to their source. The assistant cannot change your data or approve, send, pay or delete anything.
Ask questions that connect several parts of the business
You do not need to know which report or table contains the answer. Ask in ordinary language. Chat with Aeglio can move from a ranked result to a related record, such as finding the highest-revenue client and then inspecting that client’s newest project.
A useful question states the subject, period and comparison you care about. For example: “Which three clients generated the most invoiced revenue this year, and which active projects belong to the leading client?”
- Invoices and payments: totals, balances, overdue status, reminders and linked clients.
- Projects and time: budgets, members, tasks, recorded hours, billable value and uninvoiced work.
- Expenses and documents: payment, approval, extraction, duplicate and accounting-synchronization state.
- Deals, approvals, banking, travel, expense reports, integrations and the current user’s action queues.
Aeglio chooses tools, then checks the result
The language model does not receive database access or write its own SQL. It chooses from a registered set of bounded tools. Aeglio executes those tools with team scope and the same Laravel policies used by the application, returns the permitted observations and lets the model decide whether another lookup is needed.
Exact totals come from deterministic Aeglio calculations rather than guesses from document text or semantic similarity. Semantic search is useful for finding a relevant record; the structured record remains authoritative for amounts, dates, statuses and relationships.
- The planning loop can combine several domains and refine a search using record names it discovers.
- Repeated calls, tool steps, returned sources and context size are bounded.
- Large datasets are not silently truncated into misleading totals. Aeglio asks for a narrower question when a safe calculation cannot be completed.
Follow linked records and verify important answers
When the answer names a relevant Aeglio record, its title can link directly to that record in a new tab. This keeps the conversation readable while making the underlying invoice, expense, project, client or other source easy to inspect.
AI output can still be incomplete or wrong. Verify financial, tax, payment and accounting decisions against the linked record and the appropriate Aeglio report. If the assistant cannot access enough authorized data, it should say so instead of inventing an answer.
Your permissions remain the boundary
A conversation belongs to one user in one team. Other users, including team administrators, cannot open it. Search is filtered by team and every retrieved record is authorized again before its content can be sent to the selected AI provider.
The assistant only sees records you can already view in Aeglio. Changing a role or record permission changes what can be retrieved. Attachments inherit access from their parent invoice or estimate, and content found inside a file is treated as untrusted data rather than an instruction.
- Chat with Aeglio is read-only. It cannot create, update, approve, send, pay or delete records.
- Credentials, private profile data, raw banking payloads and integration secrets are excluded from tool output.
- One team’s records and semantic index entries are never used to answer another team’s question.
Choose a model and continue a conversation
Each user can choose from the models allowed for the team’s configured managed or connected AI provider. The choice is saved for that user and for the active conversation, so a better model can be selected when a question needs deeper reasoning.
Recent messages are sent as conversation context. As a conversation grows, Aeglio stores a compact continuity summary of user-stated focus, preferences, decisions and named records. Current business facts are retrieved again instead of being preserved as stale conversational memory.
Understand storage, credits and deletion
Chat with Aeglio is available to eligible Team workspaces. Managed requests use the team’s AI credits. If a team owner connects a supported provider, that provider’s account, model, processing region and retention settings apply instead.
Aeglio stores conversation messages, linked source references, model metadata and reported token usage. New activity extends the default 90-day conversation retention period. You can permanently delete your own conversation at any time. Conversation contents are not added to the semantic index.
Frequently asked questions
- Can Chat with Aeglio change a record?
- No. It is deliberately read-only and has no tools for creating, editing, approving, sending, paying or deleting records.
- Can it see records that I cannot open?
- No. Every result is team-scoped and checked against your existing record permissions before it can enter the answer context.
- Why does an answer sometimes say there is not enough data?
- The relevant record may not exist, may not be indexed yet, may be outside your permissions or may exceed a safe calculation bound. Aeglio should state that limitation rather than inventing an answer.
- Does it remember the whole conversation forever?
- No. Recent context and a compact continuity summary support longer chats. Conversations expire after the configured inactivity period, 90 days by default, and you can delete your own conversation sooner.
- Which AI provider is used?
- Aeglio-managed AI uses Mistral’s EU regional endpoint. A team owner can instead connect a supported provider such as Mistral, OpenAI, Anthropic or an approved OpenAI-compatible service.