Weekly summary — Small Appliances area. Conversations analysed: 847 (chat 312, email 198, reviews 189, social 148). Recurring issue: compressor noise on the Fridge X200 (38 mentions, 91% negative). Product request: touchscreen version of the B50 kettle (27 mentions). Positive sentiment: new induction cookware collection.
AGENT · VOICE-OF-CUSTOMER AGGREGATOR
The customer's voice reaches the category manager structured by product area.
Voice-of-Customer Aggregator combines cross-channel customer conversations weekly (web chat, customer service email, online reviews, social mentions) and produces structured summaries by product area. The retail category manager finds the picture ready for assortment decisions.
Voice-of-Customer Aggregator at work.
Thanks. I'll bring the Fridge X200 issue to the quality team this week. I'll flag the touchscreen kettle to the product team as the next variant.
Decisions recorded. Next summary: next Monday.
Why it exists.
Mid-to-large retailers generate customer voice from multiple channels: customer service chat, email, review platforms, social mentions. Manual synthesis by the category manager is often impractical at volume; signals get lost.
How it aggregates each week.
Voice-of-Customer Aggregator reads all configured customer-channel conversations weekly, identifies recurring themes by product area, and flags emerging patterns (a quality issue on a specific SKU, a repeated request for a new feature, positive sentiment around a new collection). Structured output goes to the category manager with detail by product line.
The decision stays with the team.
Assortment decisions, supplier interventions, and product campaigns stay with the category manager. The agent aggregates, structures, and surfaces.
The retail functions that change the way they read their customers.
Category manager
Receives a structured summary by product area instead of having to sift manually through hundreds of conversations. Assortment decisions are based on real customer signals, not impressions.
Cross-channel · 7d 1,420 mentions
Site chat NPS 7.2
Reviews top theme: shipping
Social · IG sentiment +0.4
Brief to the brand owner
Head of customer service
Sees recurring inbound themes before they become abnormal volume spikes. Proactive escalation activates on identified issues before the request volume becomes critical.
Claim type Auto liability
Severity Medium
Adjuster Dr. Ferrari · Northeast
Initial reserve € 4,200
Routed ✓
Marketing manager
Monitors sentiment by campaign or product launch. Signals from social channels and review platforms arrive structured — not as a raw stream to interpret manually.
Brief 07:00 12 active segments
Gold segment +8% engagement
Silver segment redemption -3%
Anomaly · campaign 14 discarded
Monday action map
A weekly summary that transforms the way assortment decisions are made.
3,000 conversations a week ordered by product area.
For the category manager at a mid-to-large retailer with 6 product areas and around 3,000 customer conversations per week across channels, the weekly brief is the starting point for assortment decisions. The agent has aggregated the week's conversations across all configured channels (website chat, customer service email, reviews, social mentions). For each product area it identified recurring themes and ranked them by volume and sentiment (positive / neutral / negative).
Quality issue, product request, positive sentiment.
Three signals stand out in this week's summary. A recurring quality issue on a specific SKU (abnormal noise reported by many customers): the category manager brings it to the quality team and to the supplier. A repeated product request (a variant with an additional feature that customers are asking for with growing frequency): the signal goes to the product team as input for planning. Positive sentiment around a recently launched collection: confirmation that the launch communication worked.
Supplier callback, product team proposal, plan confirmation.
Intervention decisions (supplier escalation, product team proposal, communication plan confirmation) come from the category manager, with the data the agent has structured. Next week, the cycle starts again.
Declarative rules from the team, integrations configured during delivery.
Rules are declarative. The category management and retail marketing team defines the product areas, the relevance thresholds for each theme, the summary format, and the aggregation frequency. Integration with customer service systems (HubSpot natively for customers who use it) and with social listening tools is delivered during the project for non-HubSpot sources. Online review platforms and social channels require dedicated adapters configured during delivery.
- Language
- TypeScript (Node.js)
- LLM model
- customer's choice: Anthropic, OpenAI, Mistral, open source models hosted internally, AWS Bedrock for a private model
- Built-in controls used
- pii-detector, tool-rate-limit
- Native delivery channels
- Slack, Telegram, WhatsApp, OpenAI-compatible HTTP
- CRM and ticketing integration
- HubSpot native; other customer service systems via adapter during delivery
- Social listening and review integration
- adapter during delivery (Google Reviews, Bazaarvoice, social platforms per customer configuration)
- Scheduling
- configurable per instance (typical: weekly summary)
- Memory
- persistent per instance
- Registry
- immutable, inspectable audit
Frequently asked questions about the agent.
HubSpot is available natively for customer service chat and ticketing. For other channels (review platforms, social listening, non-HubSpot email), the integration is delivered with dedicated adapters during the project by the Exelab team.
The typical pattern for Voice-of-Customer Aggregator is 8-12 weeks. Discovery 1-2 weeks, product area configuration and theme rules 2-3 weeks, channel integrations 3-4 weeks, hand-off to the category management team 1-2 weeks. The main factor is the number of channels to integrate and the structure of the customer's product areas.
The customer's category management or marketing team, via the graphical admin panel or directly in TypeScript. When a new product area is added or a theme relevance threshold is changed, the declarative rule is updated in the repository. For those who want to delegate operational management too, Exelab provides managed services: details are at /managed.
Value is measured on the time the category manager reclaims from manual reading and synthesis of customer conversations, the speed of quality-issue identification versus the pre-agent baseline, and the quality of assortment decisions made from structured data rather than impressions. Voice-of-Customer Aggregator is one of the agents the customer can choose to build for their retail team.
From a 30-minute conversation to the squad in production.
A 30-45 minute conversation to understand how Voice-of-Customer Aggregator would be configured for the customer's case. Channels, product areas, frequency, summary format.