Use purchasing patterns, comparable-account activity, and current stock information to identify eligible sales opportunities. Xect AI assembles a prioritized worklist with supporting evidence and proposed order details for the account team to review. Connect outputs from your predictive models to CRM tasks and draft-order workflows.
Your SAP Commerce order history already shows purchasing cadence by account, which products comparable accounts buy, and current stock levels. Today that information is often spread across reports and account knowledge that is difficult to track consistently across a large portfolio. Xect AI scores eligible account and SKU (stock-keeping unit) combinations against each account’s own purchasing pattern in MLflow, tracked per “ship-to” — the delivery location on the order — attaches supporting evidence from the events behind the score, and prepares a draft order for the account executive to review. A reorder opportunity is judged against each account’s own history rather than a shared benchmark, so seasonal, project-based or infrequent buyers are not scored against accounts with a different rhythm.
Xect AI collects purchasing activity from SAP Commerce, including orders, carts, zero-result searches and stock-status changes, and aggregates it per account, ship-to and SKU. Feature windows (trailing 30 and 90 days) and event versioning through the Business Operators Hub run in the background.
Xect AI assesses which accounts have an eligible opportunity: a reorder probability and expected date, look-alike accounts with a purchasing gap, and in-stock equivalents ranked by fit. Scoring runs against the MLflow @champion models in micro-batches of 50 to 200 accounts, and each score carries the model version and run ID.
When a purchasing pattern changes or a peer gap appears, Xect AI prepares the recommendation: what changed, against which pattern, what comparable accounts add, and the proposed order value. The Opportunity Agent writes the explanation from the events behind it, using definitions already in the Industry Knowledge Graph, such as ship-to and replenishment order.
Xect AI opens a CRM task, posts a card to Slack or Teams, and prepares a draft order in SAP Commerce through OCC for the account team to review. The draft order is not submitted without account-executive approval; the approval, the estimate and the outcome are logged for the next model run.
Each agent owns one step from a scored purchasing pattern to a reviewed draft order, and hands off cleanly to the next.
Watches each eligible account × SKU pair against its own reorder rhythm
Reads the scored output of ReorderPredictor@champion for each eligible account, ship-to and SKU, and raises an opportunity the day an expected order fails to arrive. “Late” is measured against the pair’s own history, not the catalogue average, so a 90-day buyer is not chased at day 40.
Adds the cross-sell, the substitute and the paragraph the AE reads
Pulls look-alike accounts from AccountSimilarity@champion, names the products peers buy that this account does not, checks stock and account-approved equivalence for the requested SKU, and writes the explanation with the estimate. Each claim in the paragraph points to the event or model output behind it.
Prepares the order in SAP Commerce and waits for the AE
Creates the draft order through OCC with the predicted lines and quantities, opens the CRM task and posts the card. The draft is a proposal: the account executive edits, approves or dismisses it, and either way the outcome feeds the next training run in MLflow.
Acme Industrial ordered Product A every 28–35 days; no order for 47 days. The cadence model raises it, the agent attaches the proposed order value, and the AE calls with a draft order instead of a question.
Accounts matched on purchasing pattern by the look-alike model define the gap. The AE sees which products peers add and how many of them do, so the pitch is specific to the account in front of them.
Stock status is an event like any other. When the requested SKU is unavailable, the approved equivalent already in stock is proposed in the same flow, with the same approval.
The reorder and look-alike models score the pattern in MLflow without an added language-model call; the agent then uses a model to write the explanation, and the account executive reviews, edits and approves the draft before it is submitted. The review and the outcome are traced in MLflow.
The model knows the rhythm. The agent writes the reason. The account executive makes the call, with the draft order already open.
This page covers reorder, cross-sell and substitution for B2B commerce, configured today for B2B e-commerce and distribution on SAP Commerce. The sectors below show where the same models and workflow apply with configuration for that industry’s data and approval rules. For the spend-decline side of the same accounts, see Revenue Leakage Detection; for the account-owner view, see the Account Manager Copilot.
Replenishment buyers on SAP Commerce order the same lines on their own rhythm. Xect AI models the rhythm per ship-to and SKU, and proposes the reorder the day it turns late.
See the industry pageProject-driven accounts buy in phases. Configured for phase-based buying, the look-alike model shows which products comparable projects add before the site manager asks.
See related distribution patternsConsumables have strict cadences and strict substitutes. The substitution model proposes only equivalents the customer configures as approved for the account, in stock, at the moment the requested SKU is unavailable.
See the Life Sciences industry pageThousands of low-value lines per account hide the pattern. Cadence scoring per SKU surfaces the maintenance reorder that is late and the part peers add to the same job.
Talk to us about MROWeekly cadences break quietly when a location changes supplier. A missed week is raised the same day, with the estimate and the account owner attached.
Talk to us about distributionOwns growth across their account portfolio and needs the reorder, the cross-sell and the substitute ranked and explained, with the supporting evidence — purchasing history, comparable accounts, stock status — ready to review alongside the draft order.
Owns the process between the model and the CRM, and needs every opportunity to arrive as a task with an owner, an estimate and a due date.
Owns the inbound order call and needs the approved equivalent, in stock, on screen before the customer hangs up.
Owns MLflow and needs the champion to reach a workflow without a ticket, and the outcomes to come back for the next training run.
How model outputs are connected to approved workflows.
The full Commerce Intelligence picture for SAP Commerce B2B: five queues, one event stream.
SAP Commerce, OCC, S/4HANA, Salesforce, HubSpot, Slack and Teams connect through standard connectors; interfaces without a ready-made connector are configured individually.
How draft orders, credit releases and other high-impact actions wait for one click with full context on screen.
Walk through a sample reorder or cross-sell opportunity, its supporting records, and the draft-order review process. Customer-specific analysis can be prepared after data and model requirements are agreed.