التجارة
أدره نيابة عني
Let Sela learn your store or restaurant, test product placement against your goal, and keep a history of results and scheduled briefings.
آخر تحديث: 2026-10-08
AI Store Manager, called Manage it for me in the workspace, analyses your catalog, inventory, orders and storefront activity. A review means one analysis of those business records, with observations and suggested next steps. You choose a business goal and whether product placement tests need your approval. Sela keeps the evidence, results and business memory together in Store → Manage it for me.
Requirements#
Use a Restaurant or Standard Store workspace with a live storefront and an active plan. The Large bundle includes the manager with its storefront. Advanced and existing platform agreements can add it on Inbox and Operations; Scale and Enterprise include it with an active storefront. Prices and applicable annual terms are on the billing page. You do not need AI Assist for manager reviews. See Plans and billing.
Your first activation starts one 30-day preview, with no automatic charge when it ends. Pausing, resuming and renewing do not restart it. When the preview ends, buy the manager add-on or use an included plan to continue. Your profile, reports and history stay saved. Existing managers use their original creation date for this preview.
Each storefront includes one scheduled analysis per UTC calendar day and 10 extra analyses you request per UTC calendar month. The dashboard calls these analyses reviews. For example, request another analysis after adding products or changing your menu. Reviews triggered by choosing a goal count as manual reviews. Queuing reserves a review, including one later cancelled or affected by a provider failure. Retrying the same active review does not reserve another. The dashboard shows your remaining manual reviews. Once those are used, automatic briefings continue.
Manager reviews use no shopping AI credits or AI wallet funds. The customer shopping assistant still uses its separate AI allowance. If the manager's provider fails, it can prepare a clearly labelled review from verified business records.
Members need permission to manage settings to enable the manager, change its goal or approve a test. Other permitted members can review its dashboard.
Start learning#
- Open Store → Manage it for me.
- Enable the manager. It starts in Learning, or can be ready to choose a goal when your existing business evidence is sufficient.
- Review the business profile and observations. Existing orders help establish demand; storefront visits and product activity add new evidence as customers browse.
- When the manager has enough evidence, choose a goal.
The manager keeps separate learning and testing stages. Learning describes what happened in your business. A test compares different product placements to find whether a change helps.
Choose your goal#
| Goal | What the manager considers |
|---|---|
| Sell more | Products with demand and available stock that could benefit from more visible placement. |
| Launch new products | Recently added orderable products or dishes. |
| Clear slow-moving stock | Tracked inventory with low sales and stock available. Restaurants use available dishes with low demand when unit stock counts are unavailable. |
The manager uses your published catalog and current availability. It cannot make a sold-out item orderable or invent a price.
Control its work#
Choose Review first to approve a proposed placement test before it starts, or Automatic merchandising to let the manager start reversible placement tests for your selected goal.
Customers in a test see either your current layout or a version with selected products featured near the top. A customer's assignment remains consistent during that test. Prices, checkout rules and saved storefront design stay under your control.
The dashboard shows visitors, ordering customers, ordered revenue and the observed conversion difference. Ordered revenue represents orders placed, before costs and payment collection. Cancelled and rejected orders are removed from the measurement.
Low traffic or overlapping results can leave a test Inconclusive. The manager can stop a harmful test and preserves the current version when the evidence does not support a change. An observed improvement does not guarantee future sales. The manager saves what happened so future decisions can use the result.
In Review first, use Use winning version after a test wins. In Automatic merchandising, the manager can apply a supported winner. Later tests compare against that version. Restore original storefront removes the manager's placement changes.
Approval and adoption controls appear only while the experiment still matches your manager settings, live storefront, current catalog and subscription. A winner must also remain supported by current measurements, including order cancellations. Earlier results remain in the history when their actions are no longer available. Experiment deadlines and access checks use server time, even when a customer's device clock is inaccurate.
Use Pause to stop management or stop the current test from its card. The original design remains available. Editing the storefront yourself invalidates the earlier test's design assumptions.
Routine reviews keep a current proposal available for approval, or continue the running test. Changing the goal or merchandising mode stops the earlier proposal and lets a fresh review prepare work for your current settings. An unstarted or invalidated proposal does not count as a completed test.
Stopping a candidate or choosing Restore original storefront prevents the next routine review from immediately proposing that same unchanged layout again. Selecting a goal again, changing the merchandising mode or changing the relevant catalog releases that earlier stop. Changing only the briefing schedule keeps it in place. Reports distinguish work awaiting approval, owner-stopped work, a completed measurement period and an adopted layout.
Briefings and business memory#
Choose daily or weekly briefings, the local hour, and your business timezone. Weekly briefings run on Mondays. A briefing explains the evidence, work performed and proposed next steps. Saved reports remain available in the dashboard alongside daily, weekly and monthly business memories. Later reviews retain earlier recorded decisions in those memories and label shortened history. The profile uses a recent reporting window; each report shows the period it covers. Activity totals in those memories use UTC calendar periods, while your briefing delivery follows the selected business timezone.
Use the analysis control when you want an updated profile before the next scheduled briefing. If the AI service cannot complete its analysis, the dashboard reports the problem and can provide a summary based on verified business records.
Discounts and marketing#
The manager can prepare discount and marketing opportunities for your review. Approving a discount creates an unpublished offer draft for the selected product. Review its price, availability, branches and terms in your workspace's offer tools before publication. Approving a marketing opportunity saves your decision; prepare the audience and campaign through your connected channel tools before sending.
A WhatsApp contact link on the storefront does not by itself authorize marketing messages. Campaigns still require a connected sending channel, a suitable audience and the channel's messaging rules. See WhatsApp.
If management stops#
Check that the storefront is live, the subscription still qualifies and AI access is available. A paused manager or an unavailable storefront stops applying experimental placement. The dashboard keeps your prior results and reports for review.
لم تجد ما تبحث عنه؟ تواصل مع الدعم.