Client Sentiment Analysis Tool — Google Workspace Integration
Overview
An internally built AI tool that ingests all company communications via the Google Workspace API and uses a vector database to provide real-time client sentiment scoring and surface critical operational issues. The tool was demoed internally on 2026-02-18 and access is being rolled out to the broader team.
First discussed and demonstrated in [1].
How It Works
The tool connects to Google Workspace and pulls every communication artifact associated with a client:
- Email (all company inboxes)
- Slack messages
- Call transcripts (via Fathom recordings)
- Google Drive documents
All of this data is embedded into a vector database, enabling both structured filtering and semantic search across the full communication history. The system then runs analysis to produce:
- Sentiment scoring — overall tone and trend across the relationship
- Critical issue flags — operational problems surfaced from within the communication record
- Activity statistics — volume of touchpoints over time, with date-range filtering
- Tone trend monitoring — tracks changes in client responsiveness, professionalism, and engagement patterns
"It's basically rooting through all your stuff and saying, here's what's going on." — Mark Hope
Key Capabilities
Sentiment Analysis
- Reads every document, email, and call transcript associated with a client
- Produces an overall sentiment label (e.g., Positive, Stable, Declining)
- Tracks tone trend over a configurable date range (e.g., 30 days, 90 days)
- Monitors response latency — flags if a client starts taking longer to reply
Critical Issue Detection
- Surfaces operational problems mentioned or implied across communications
- Examples from the 2026-02-18 demo:
- Citrus America: 32 inactive trade show geofencing campaigns still enabled and wasting budget; Google Ads remarketing inefficiency; conversion action misconfiguration
- Adavacare: Task delivery delays; 3 of 10 locations missing pricing data; photo asset verification issues
Search & Query Interface
- Structured filters: filter by client, person, content type (email, Drive doc, Slack), and date range
- Semantic search: natural language queries like "tell me about budget issues" or "project status"
- AI Q&A: ask freeform questions (e.g., "When was the last time we changed the budget?") and receive answers grounded in the actual communication record
Activity Dashboard
- Per-client statistics card showing total email, Slack, Drive, and call volume
- Activity timeline to identify periods of high/low engagement
- Useful for spotting relationship drift before it becomes a churn signal
Important Limitations
- Sentiment reflects stated tone only — if a client is being politely positive while privately dissatisfied, the tool will not detect it
- Not a campaign analytics tool — it does not analyze ad performance numbers, CRM data, or billing figures; it only processes communication content
- Complementary to, not a replacement for, X-Ray — X-Ray handles structured performance data; this tool handles unstructured communication data
- Still in active development — the AI Q&A ("Ask") feature is noted as not yet fully reliable; onboarding flow had minor login issues at launch
Access & Rollout
- Built by Mark Hope
- Access being extended to Melissa Cusumano and others as of 2026-02-18
- Login is via invite; password reset flow had minor issues at initial rollout (Mark is aware and monitoring)
- Contact Mark to request access or report issues
Example Outputs
Citrus America (run: ~2026-02-18)
| Dimension | Result |
|---|---|
| Overall Sentiment | Positive and professional |
| Tone Trend (30-day) | Stable, consistently positive |
| Responsiveness | No declining patterns |
| Critical Issues | 3 flagged (geofencing waste, remarketing inefficiency, conversion misconfiguration) |
Adavacare (run: ~2026-02-18)
| Dimension | Result |
|---|---|
| Overall Sentiment | Positive and stable |
| Tone Trend | No deterioration detected |
| Urgent Issues | 4 flagged (task delays, missing location pricing, photo asset verification, messaging adjustment needed for Fardale) |
Strategic Value
The tool addresses a specific blind spot: clients can appear healthy in calls and emails while operational problems accumulate underneath. The Adavacare example is illustrative — the client's communication tone was positive and appreciative, yet the tool surfaced four urgent delivery issues that required attention.
This makes it particularly useful for:
- Pre-call preparation (know what's broken before the client brings it up)
- Account health monitoring at scale across the full client roster
- Early churn detection via tone trend and responsiveness monitoring
- Internal accountability (task delays become visible in the communication record)
Related
- [1]
- [2]
- [3]
- [4]