When missing information would materially change the outcome, Laureo asks for it in one card. A card can combine choices, numbers, dates, text, presets, defaults, help text, custom options, and inline validation. Your answers persist across refresh and devices until you select Continue.

A field that has a sensible default arrives already filled in with it, and the word Assumed sits beside the label while it still carries that default. So what would happen if you simply selected Continue is visible before you select it, and the word disappears the moment you change the value, because at that point it is your answer rather than an assumption.

One card covers everything missing, and one Continue sends the whole set back at once, rather than a question per turn. After you continue, the form is replaced by a short line naming what you answered and what was assumed, so the conversation keeps a readable record of the terms the run is about to act on.

When several materially different strategies exist, work spans multiple domains, or approval checkpoints shape the outcome, Laureo proposes an execution plan. The plan shows the objective, deliverables, defaults, affected records, ordered steps, and exact approval checkpoints.

An exact action approval is bound to the persisted run and conversation, exact targets and recipients, content version, risk, and a fingerprint of its parameters. A material edit invalidates the old approval and creates a new preview.

The decision saves immediately, but the web request does not own the effect. A background worker rechecks your current membership, seat, role, plan, grant, feature, record scope, record version, recipients, and approval fingerprint before it acts.

Batch previews identify whether a subset is allowed or all items must succeed together. Every item receives an explicit decision; editing one item creates a new preview for that item. Repeated clicks, tabs, retries, or worker restarts cannot claim the same effect twice.

For example, ask: "Create an AI Consulting Retainer service at $2,500 and prepare a quote for review." If no stronger record or organization default exists, Laureo proposes:

• product type: service;
• currency: USD;
• quantity: 1;
• title: AI Consulting Retainer Quote;
• expiry: 14 days; and
• terms: full payment is due in advance.

The preview identifies each assumption. Laureo does not invent a price or financial commitment, so a missing unit price appears in the grouped clarification. After the required approval, the compound action creates exactly one product and one quote and returns one prominent Open quote action.

Ask Laureo to create a booking page and it can prepare the page through the Scheduler's canonical service. The completed artifact provides one primary Open booking page action to the exact page editor while keeping this conversation available.

"Help me plan a campaign" returns grounded strategy and, when useful, exactly one Create campaign draft for review recommendation. It does not create or schedule anything until you select that action.

"Create and schedule a campaign" continues through authorized preparation and draft creation. Missing audience, recipient consent, date, time, or time zone appears in one clarification because Laureo cannot infer them. Scheduling or sending receives an exact preview and approval.

After substantive work completes, Laureo may offer zero or one next action that materially advances the same goal. Opening or reviewing a newly created artifact has priority. A recommendation is suppressed while work is running, waiting, or showing an error.

Recommendations never execute automatically. Laureo deduplicates completed or dismissed suggestions and remembers Not now by action family. Suggestions start on. Turn them off with Suggest one next action in Chat Settings, either from the gear on the Laureo AI page or at Settings, then Laureo AI preferences.

Laureo maintains a classified inventory of current app actions and their Assistant and Agent policy. Only capabilities with reviewed schemas, permissions, scope, risk, retry, receipt, and destination behavior are operator-enabled. Some classified actions remain partial or unavailable while that review continues.

Human-only boundaries include entering external credentials, completing external consent, recent identity or multi-factor verification, and provider identity checks. When a step requires direct interaction, Laureo provides the exact settings page or record destination and keeps your conversation context ready for your return.