--- title: Client Sentiment Analysis Tool type: article created: '2026-03-03' updated: '2026-03-03' source_docs: - raw/2026-03-03-weekly-call-w-sebastian-126806076.md tags: - ai-tools - client-management - operations - sentiment-analysis - clickup - slack layer: 2 client_source: null industry_context: null transferable: true --- # Client Sentiment Analysis Tool ## Overview The client sentiment analysis tool provides a unified, AI-powered view of client health by aggregating data across all major communication and project channels. It surfaces issues before they escalate and gives account managers a high-level picture of every account without manually reviewing each channel. As Mark described in the [[meetings/2026-03-03-weekly-call-w-sebastian|2026-03-03 weekly call]]: > "I think you guys are going to find that sentiment analysis — you could run the whole freaking business from that. It basically tells you, here's what they're saying and here's what's happening." --- ## Data Sources The tool ingests and cross-references data from four channels: | Source | What It Watches | |---|---| | **Email** | All emails sent or received by anyone to/from a client — including tone and sentiment, not just content | | **Slack** | Client-related channel activity | | **ClickUp** | Client project folders — task status, due dates, overdue items | | **Google Drive** | New documents added; reads content to understand what's being discussed | --- ## Key Capabilities ### Client Health Scores & Sentiment Analysis Each client receives a sentiment rating (positive, neutral, declining) based on aggregated signals. The tool flags patterns like frustrated email tone, repeated rescheduling, or stalled deliverables — even if no single signal would raise an alarm on its own. ### 7-Day and 30-Day Summaries From the client page, a summary can be generated for the last 7 or 30 days. This produces an executive summary covering: - Recent email activity and tone - ClickUp task status (overdue, backlogged, completed) - Slack activity - Any new documents or deliverables in Google Drive ### Search Across All Communications A search function allows querying across all ingested data. Filters can be applied by client, channel, or date range. Useful for quickly locating a specific email, decision, or commitment without digging through individual tools. ### Private AI Query Layer An AI interface allows natural-language questions against the ingested data only — it does not query the open internet or hallucinate from general LLM knowledge. Filters can be scoped to a single client or channel. --- ## Automation: Weekly Slack Report Every **Sunday morning**, the tool automatically runs the sentiment analysis across all clients and posts the report to Slack. The team can review it Monday morning to identify any accounts that need attention before the week begins. > "What I'm doing now is I'm writing this thing up where every Sunday it's going to take this report and run it and put it in Slack. So you can come in Monday morning and see what it says about the last seven days." --- ## Interpreting Results The tool is intentionally direct and can read as alarmist. The team noted it tends to heavily weight negative signals and may flag issues that have already been resolved (especially on a 30-day lookback). Calibration notes: - **Cross-check against known context.** A declining sentiment score for a client may reflect a single frustrated email, not a systemic problem. - **Overdue ClickUp tasks are a reliable signal.** The tool consistently surfaces these, and they often represent real delivery gaps. - **Tone analysis on email is useful but imprecise.** A client with a blunt communication style (e.g., [[clients/cord-wainer/_index|Cord Wainer]]) may read as more negative than they are. - **Don't dismiss it.** Even when the language is hyperbolic, the underlying data is usually accurate. > "I don't really want to be stroked here. I want it to tell us how it is." --- ## Issues Surfaced in First Review (2026-03-03) The first team-wide review of the tool during the [[meetings/2026-03-03-weekly-call-w-sebastian|2026-03-03 weekly call]] surfaced several actionable issues: | Client | Issue Flagged | Status | |---|---|---| | [[clients/advanced-health/_index\|Advanced Health & Safety]] | Website errors found by client; trust deficit | Resolved; white-glove approach adopted | | [[clients/crazy-lennys/_index\|Crazy Lenny's]] | Google Ads account suspended; $2,500 balance | Under investigation | | [[clients/citrus-america/_index\|Citrus America]] | Blog content approval bottleneck; repeated rejections | AI training workspace proposed | | [[clients/didion/_index\|Didion]] | Overdue ClickUp tasks; scope creep on website guidance | Time tracking initiated | | [[clients/reynolds/_index\|Reynolds]] | Blog tasks backlogged since December | Needs attention during AM transition | | [[clients/exterior-renovations/_index\|Exterior Renovations]] | Blog tasks backlogged; empty meta descriptions | Meta descriptions queued for fix | | [[clients/overhead-door/_index\|Overhead Door]] | Neutral/declining; Google Ads blocked; multiple reschedulings | New proposals being prepared | | [[clients/cord-wainer/_index\|Cord Wainer]] | Flagged "unreliable commitment" — likely false positive | Flyer finalized; situation stable | --- ## Related Tools & Integrations - **[[knowledge/tools/website-visitor-identification|Website Visitor Identification Tool]]** — companion tool being tested on the Asymmetric site; identifies specific visitors and companies, feeds hot lead alerts into Slack and HubSpot - **ClickUp** — primary source for task/delivery signals - **HubSpot** — downstream integration for lead and contact data - **Fathom** — meeting recordings feed qualitative context that complements quantitative sentiment signals --- ## Recommended Usage 1. **Review the Sunday Slack report** at the start of each week. 2. **Run a 7-day summary** before any client call to get a fast briefing. 3. **Use the search function** to locate specific commitments or communications before escalations. 4. **Flag overdue ClickUp tasks immediately** — the tool reliably surfaces these and they represent real risk. 5. **Use the AI query layer** to investigate a specific client situation without switching between tools.