Monitors engagement metrics per subscriber to identify customers approaching email fatigue – declining opens, increasing deletes, rising complaint rates. Before they unsubscribe, the workflow reduces their send frequency or pauses them from batch campaigns automatically.

Delivers a daily briefing on every active campaign: email, paid, SMS, push. Aggregates open rates, click rates, conversion, and revenue attribution – with AI attribution beyond last-click. Flags campaigns significantly over or under benchmark.

Tracks loyalty program health across enrollment, engagement, redemption, point liability, and tier distribution – with trend analysis and early warning flags for structural problems (low redemption, high point liability, tier stagnation).

Analyzes performance of PDPs, landing pages, editorial content, and campaign creative – ranking by conversion contribution and surfacing underperformers that need refresh. Connects content quality signals to downstream conversion and AOV.

Runs Shapley value attribution across all marketing channels weekly to show true multi-touch contribution – beyond last-click. Surfaces which channels are over-credited (usually last-touch) and under-credited (usually upper-funnel), with specific budget reallocation recommendations.

Tracks customer acquisition cost by channel, cohort quality, and LTV – showing not just how much it costs to acquire a customer, but which channels acquire customers worth keeping. Surfaces channels in diminishing returns and recommends reallocation.

Measures promotion performance in near-real-time: true uplift vs. baseline, pull-forward (sales borrowed from future periods), and cannibalization across non-promoted products. Distinguishes genuine incremental revenue from discount cost that merely accelerates purchases.

Monitors cart abandonment rates in near-real-time against a rolling baseline. When abandonment spikes, diagnoses probable causes (site performance, competitor price shock, checkout friction, coupon failure) and recommends an intervention strategy.

Re-runs customer segmentation weekly and compares segment membership against the prior week. Alerts when a segment grows or shrinks significantly – often the earliest signal of a broader trend (acquisition quality shift, retention problem, pricing effect).

Monitors brand and product mentions across social channels. When volume spikes (viral moment or PR crisis), fires immediately: sentiment breakdown, top themes driving the spike, and response strategy recommendation based on your brand's social playbook.

Analyzes the past month's resolved tickets for recurring questions with no knowledge base coverage. For each gap, auto-drafts a FAQ entry in your brand voice – ready for SME review and publication.

Measures first-contact resolution rate by issue type, channel, and team – tracking which issue types are systematically failing FCR and feeding repeat contacts. Surfaces root causes (knowledge gaps, policy ambiguity, routing errors) with specific improvement recommendations.

Delivers a weekly performance summary for each agent and team: resolution time, CSAT, ticket volume, FCR, escalation rate, and specialization accuracy. Surfaced to team leads with coaching recommendations for underperformers.

On every `supportservices/ticket/create` event, extract the ticket payload, look up the order and customer context, predict the customer's CLV and Churn risk, fetch the credit-policy and brand voice for the customer's tier from the knowledge bases, classify the ticket via a fine-tuned SLM, draft a brand-voice-correct apology email with a goodwill credit computed STRICTLY as `order_value * credit_percentage / 100`, send the email via the notification agent, and finally produce an admin-facing report capturing every decision and the email delivery status.

Tracks CSAT scores at the team, agent, and issue-type level continuously. A 0.3-point drop triggers a drill-down analysis: is it a specific agent, issue type, product category, or systemic routing problem? Returns a ranked list of probable causes with supporting evidence.

Monitors escalation rates by product, issue type, and team in real-time. When a spike is detected, diagnoses the root cause – product defect, fulfillment failure, or policy gap – and surfaces it to the right team before it compounds into a larger CX failure.

A customer contacting 3+ times in a month is a churn signal. Flags them automatically and triggers proactive outreach: a personalized apology and resolution offer delivered by the right channel – before they decide to leave.

Scans the entire open ticket queue every 30 minutes, predicts which tickets will breach SLA in the next 2 hours, and takes action: re-routing to available agents, escalating to senior staff, or proactively notifying customers.

Forecasts today's expected support ticket volume – by channel, issue type, and time-of-day distribution – and translates it into a staffing recommendation. Accounts for active promotions, recent order spikes, known product issues, and day-of-week patterns.

Compares current category sales trajectory against historical seasonal norms and flags any category deviating more than 15% from its expected seasonal curve – in either direction – before it becomes an over/under-stock situation.

Compares your product catalog against competitor catalogs to identify category-level gaps – products or sub-categories that competitors carry and you don't. Prioritizes gaps by estimated demand and revenue opportunity.

For every product in its markdown lifecycle, calculates the optimal time for the next price reduction, the depth of that reduction, and projected sell-through at each price point – using stochastic optimization against real demand curves.

Surfaces top trending-up and trending-down products by sales velocity, cross-referenced with current inventory. Gives merchandisers a weekly pulse on what's accelerating and decelerating – with stockout risk flags so urgency is always visible before the week starts.

Scrapes competitor product pages daily for matched SKUs, detects price changes, models the likely demand impact, and provides pricing response recommendations – all grounded in your own pricing rules and channel policies.

A daily morning briefing for category managers: revenue, units, conversion, AOV, and return rate by category – with DoD, WoW, and YoY comparisons. The AI narrative calls out the most significant movers and connects metric changes to known causes.

On every OMS order-create event, first delegate payload extraction to payload_extractor. If the extracted routingkey is 'order.return' or "ordertype":"Return", predict the SKU's current return rate via the fine-tuned prediction model (using the extracted skuId), compare it against the category-specific threshold from the platform KBs, classify the return reasons via the categorisation SLM, and generate a structured report enriched with the extracted generic order metadata (order_id, user_id, dates, amount, currency). The report uses an ALERT tone when the threshold is crossed and an INFORMATIONAL tone otherwise. When routingkey is anything else, route directly to narrative_writer with a 'not a return order' message and finish.

Analyzes PO receipt data against expected delivery dates and quantities. Tracks which suppliers are consistently late, short-shipping, or showing declining fill rates – and surfaces the financial impact of those failures on stockouts and lost sales.

Identifies suppliers whose actual lead times are systematically longer than their stated lead times – quantifying the drift and its impact on stockout exposure and safety stock requirements. Enables data-driven vendor reviews and safety stock recalibration.

Analyzes actual shipping spend by carrier, service level, zone, and package type against alternative rates – identifying where different carrier routing choices would achieve equal or better service at lower cost.

Tracks orders processed vs. capacity at each warehouse in near-real-time. Identifies bottlenecks (receiving, picking, packing, shipping) as they emerge and flags risk to same-day or next-day SLAs – giving operations managers time to act before promises are broken.

Delivers a daily carrier performance briefing: on-time delivery rates by carrier, region, and service level – with trend analysis, exception rates, and cost vs. performance comparison. Surfaces carriers degrading before SLA breaches impact customers at scale.

Tracks the returns processing queue: volume awaiting processing, aging by return type and product, and SLA compliance for refund issuance. Flags returns approaching refund SLA deadlines and identifies product return spikes signaling quality or listing issues.

For every Buy Online Pickup In Store order, assesses whether the store can fulfill by the promised ready-by time. If at risk, triggers a staff alert immediately – reducing the #1 BOPIS complaint: arriving before the order is ready.

Monitors demand forecasts 8 weeks ahead of projected peak periods and generates a preparation checklist: staffing needs by function, carrier capacity reservation timing, inventory pre-positioning, and warehouse throughput readiness.

Monitors real-time carrier exception data. When a carrier or geography experiences an unusual surge in delays or lost packages, identifies affected orders, assesses customer impact, and recommends whether to reroute new orders to an alternative carrier.

No mission description available.

No mission description available.

No mission description available.

No mission description available.

No mission description available.

No mission description available.

Delivers a weekly executive-level view of inventory health expressed in financial terms: total inventory value, aging profile, stockout exposure (revenue at risk), and overstock exposure (capital tied up in slow-moving inventory).

Delivers a weekly competitive intelligence summary: pricing moves by key competitors on matched SKUs, new product launches, estimated market share shifts, and strategic signals – synthesized into an executive-readable briefing.

Audits all active ML models: accuracy trends, drift detection status, prediction vs. actuals validation, and business ROI per model. Gives executives visibility into the performance and value of the AI investment, and flags models needing retraining or retirement.

The anti-dashboard: only fires when something is materially off-plan. Monitors all key business metrics continuously and alerts executives only when a threshold deviation occurs – eliminating noise while ensuring nothing critical is missed.

Provides a monthly view of overall customer base health: churn risk distribution, LTV trends, NPS movement, segment migration, and reactivation opportunity sizing. Translates complex ML signals into an executive-readable assessment of where the customer base is heading.

The CEO/GM's Monday morning briefing: revenue, orders, AOV, conversion, and NPS – with YoY comparison and AI-generated narrative. Three agents collaborate to produce it: one pulls the data, one identifies the most significant signals, one writes the narrative.

Monitors all open quotes for response status. When a quote passes its optimal follow-up window without a response, drafts a personalized follow-up for the rep – or sends it directly on their behalf (L3). Prioritizes effort by deal value x close probability.

Tracks account health across the B2B customer base: order frequency trends, purchase volume trajectory, engagement signals, and churn risk scores at the individual account level. Surfaces accounts deteriorating before they become churned accounts – when a single account can represent $50K-500K+ in annual revenue.

Monitors each B2B account's consumption patterns and predicts when they'll need to reorder – before they initiate the purchase themselves. When an account approaches their predicted reorder window, sends a proactive outreach with a personalized quote already prepared at their account-specific pricing.

Identifies B2B accounts whose product penetration is low relative to peer accounts with similar size, industry, and tenure. Surfaces specific cross-sell categories the account has not purchased from – ranked by fit and revenue opportunity – and equips the account manager with a data-backed expansion pitch.

Generates a 90-day view of upcoming contract renewals: renewal probability, estimated contract value, risk assessment, and recommended actions for each account. Enables sales leadership to allocate renewal effort where it matters most – before contracts lapse.

This workflow runs in TWO phases. Phase A (collect) gathers data and ends with the FINAL dead-inventory SKU list. Phase B (decide + act) is a MANDATORY human-review gate where the user chooses what to do with that list. Phase A is NOT the deliverable — finishing after Phase A is a BUG. The workflow's real output is the human-approved action (email and/or revival plan).

Every 8 minutes, fetch the platform's inventory-quantity threshold from the knowledge bases, ask the Equinox Inventory Agent for the current quantity of every inventory bin item, build the list of SKUs whose current quantity falls strictly below that threshold (each entry carrying both sku_id AND current_quantity), then route to stock_predictor ONCE with the FULL list of at-risk SKUs (never one call per SKU) so it predicts every SKU's ideal stock in parallel tool calls within a single turn. Aggregate the per-SKU predictions and pass them, the threshold, and the inventory data to the narrative writer, which produces ONE table-first reorder-suggestion report whose primary deliverable is a per-SKU table stating the exact REORDER AMOUNT (ideal_stock_quantity – current_quantity, clamped at 0) for each at-risk SKU.

Every 6 hours, identify best-selling SKUs — those ordered by customers in the last 30 days — that are currently running low on stock (ATP quantity below the platform's low-stock threshold). Produce a stockout-risk report and, when at-risk SKUs are found, send an alert email to the inventory manager so restocking can be actioned before a stockout occurs.

Every 6 hours, fetch the platform-wide stockout / minimum-quantity threshold from the knowledge bases, scan inventory for SKUs whose current quantity is strictly below that threshold (the stockout-risk set), predict the ideal stock quantity per at-risk SKU via the fine-tuned Ideal Stock Quantity Predictor, draft a quantity-only Purchase Order from the predictions, and PAUSE at a HITL interrupt where a reviewer must approve / reject / request_changes. On approve, email the PO to alankar.saraf@InfosysEquinox.com via eq-notification-agent. On reject, discard the PO. On request_changes (the reviewer supplies modifications in the freeform field), re-draft the PO applying those modifications and email the modified PO to the same recipient. Produce a final markdown trace summarising the run.

Every 6 hours, build 12 forward-looking Google Ads campaign_ids (4 regions x 3 months: current / next / next+1), score each campaign via FOUR fine-tuned Ad Spend Optimizer models (Revenue, Margin, Spend, Optimal Spend) using predict_entity with entity_type=`campaign_id` – 48 predict_entity calls per run – pull the documented Google Ads expected margin and the documented Google Ads strategy from the knowledge bases (two separate ask_knowledge_bases calls), and produce ONE brief markdown report comparing per-campaign Spend vs Optimal Spend and Margin vs the documented margin band, with concrete optimisation moves where the gap is material.

Every 6 hours, fetch all orders created in the last 7 days, aggregate per-SKU units sold and revenue, identify the top 5 SKUs by units_sold (ties broken by revenue desc, then name), and produce a markdown report. No predictions, no knowledge lookups, no SLM calls – this workflow uses an agent_ref for the data work and an inline writer for formatting.

Every 6 hours, fetch all orders created in the last 7 days, aggregate the total revenue and per-product revenue, identify the top 3 products by revenue contribution, and produce a markdown report. No predictions, no knowledge lookups, no SLM calls – this workflow uses an agent_ref for the data work and an inline writer for formatting.

Every 6 hours, identify all products created in the past 30 days, look up category-scoped launch targets / performance thresholds / buyer notes / lifecycle benchmarks from the platform knowledge bases, then iterate the new products ONE AT A TIME and score each via four fine-tuned ML models (New Product Success Predictor, New Product Demand Estimator, Product Lifecycle Classifier, Product Return Rate). Aggregate the per-product and per-SKU predictions and hand them, together with the catalog data and the category-keyed KB guidance, to the narrative writer, which produces ONE table-first ALERT-tone report whose primary deliverable is a per-product table of forecast-vs-actual (with under/outperformer flags and deviation magnitude), demand estimate, average return rate (worst SKU called out), lifecycle stage, and confidence – followed by a per-product narrative paragraph and a Notes section.

No mission description available.