Skip to content

Risk Improvement Recommendation Chase

Example prompt: "When an insurer's surveyor sends through risk-improvement recommendations on a policy, log every recommendation, chase the client to address them, and remind me before the insurer's deadline so we don't end up with a cover restriction at renewal."

The Problem

The insurer's surveyor visits the client's warehouse, files a report two weeks later with eight recommendations — three immediate (fix the fire-door magnetic releases, replace the missing extinguisher in the loading bay, document the hot-work permit procedure), three to be addressed in 90 days, two to be addressed by renewal — and emails it through with a covering line that says "please confirm completion of items 1, 2 and 3 within 30 days as a condition of cover". The account handler forwards it to the client and at renewal we discover that only two of the three immediate items were actually done. The insurer either restricts cover, imposes a warranty, or loads the premium, and the client is unhappy because nobody told them clearly what they had to do or when.

How GloriaMundo Solves It

We build a workflow that fires when a surveyor's report arrives in the broker mailbox or is logged on the Surveys register. An LLM step reads the report and extracts each recommendation as a discrete item — rec_number, description, severity (immediate / 30-day / 90-day / by-renewal), and the insurer's stated deadline. An integration step writes one row per recommendation to a Recommendations Tracker, an LLM step composes a single client email summarising what each rec requires and when, and an integration step saves it as a Gmail draft. A scheduled step runs every weekday, finds rows still in 'Open' or 'In Progress' status, bands them by days_to_deadline (14, 7, 3, 0), and drafts the appropriate chase email or alert. Glass Box preview shows the tracker rows being created, the client email, and the chase ladder before any of it goes out.

Example Workflow Steps

  1. Trigger (integration): A surveyor's report PDF arrives in the broker mailbox or is filed in the client's /Surveys folder on Drive and logged on the Surveys register.
  2. Step 1 (llm): Read the report and extract each recommendation — rec_number, description, severity, the insurer's stated deadline_date, and any conditions-of-cover language (warranty, subjectivity, deferred-cover restriction).
  3. Step 2 (integration + code): For each recommendation, write a row to the 'Recommendations Tracker' tab with policy_id, rec_number, description, severity, deadline_date, status 'Open', and a tracker_row_id allocated as 'REC-[policy_id]-[rec_number]'. Dedupe on (policy_id, rec_number) so a re-uploaded report does not duplicate rows.
  4. Step 3 (llm): Compose a single client email listing every recommendation grouped by severity, with the insurer's deadline against each one and a clear note on which items are conditions of cover.
  5. Step 4 (integration): Save the client email as a Gmail draft to the policy contact, with the surveyor's report attached. Persist gmail_intro_draft_link to each Recommendations Tracker row created in this run.
  6. Step 5 (scheduled, weekday): Every weekday at 9am, read the Recommendations Tracker filtered to status 'Open' or 'In Progress'. For each row, compute days_to_deadline and bucket into a band — 14, 7, 3, or 0.
  7. Step 6 (code, gate): Dedupe on (tracker_row_id, band) against the 'Recommendations Chase Log' tab; a row already chased at this band halts; otherwise write a provisional row before any drafting step.
  8. Step 7 (llm + integration + conditional): For band 14, draft a gentle reminder to the client. For band 7, draft a clearer ask citing the insurer's deadline. For band 3 on an immediate-severity rec, additionally red-alert in #compliance on Slack tagging the producer. For band 0 on a condition-of-cover rec, escalate to the producer with a phone-call recommendation because the cover position is at stake.
  9. Step 8 (integration): Post the daily risk-improvement digest in #compliance — counts by severity and band, the band-3 and band-0 names called out for the day.
  10. Step 9 (integration): When the client confirms completion (via reply or via the account handler marking the row 'Closed'), flip the Tracker row to status 'Closed', record closure_date, and stop the chase ladder for that row.

Integrations Used

  • Gmail — the surveyor's report trigger, the client reminder drafts, the chase drafts
  • Google Sheets — the Surveys register, the Recommendations Tracker, the Recommendations Chase Log
  • Google Drive — the per-policy /Surveys folder where the report and any closeout evidence is filed
  • Slack — the daily #compliance digest, band-3 red alerts, band-0 escalations to the producer

Who This Is For

Commercial brokers placing property, fleet, manufacturing, hospitality, or construction risks where insurer-led risk surveys are routine and the surveyor's recommendations are conditions of cover or carry warranty implications. The use-case applies equally to in-house risk managers running their own follow-up programme against landlord, insurer, or audit recommendations.

Time & Cost Saved

An account handler running risk-improvement follow-up across thirty live surveys typically loses two to four hours a week to scrolling through reports to work out who has what outstanding, and the cases that bite are the ones that surface at renewal with a missed condition. This workflow turns it into a fifteen-minute review of pre-drafted client emails and a daily digest of where each survey sits, and the band-3 / band-0 alerts surface the cases where a phone call is going to do more than another email. The cover-restriction-at-renewal cases that come from "nobody chased" disappear from the renewal book.