property management · AI & process automation

Past-due work order escalation engine

A multi-step automation that watches past-due maintenance work orders in Rentvine, picks the right vendor with AI, follows up on its own and logs every decision in HubSpot.

client
Property management company
duration
4–6 weeks
built with
Zapier, Claude Sonnet 4, Rentvine API, HubSpot API
runs
Every day, on its own
One daily runexample
  1. 1
    Daily triggerThe run starts on schedule, no one presses anything
    done
  2. 2
    Fetch past-due work ordersRentvine API
    done
  3. 3
    Classify each orderClaude Sonnet 4, with self-correction
    done
  4. 4
    Score vendors on 9 factors“BestVendor” tags first, exclusions respected
    done
  5. 5
    Send the right reminderExisting orders: 3 steps · new assignments: 2 steps
    sent
  6. 6
    Log the decisionHubSpot work order object, 25+ fields
    done
AI classification
94.7%

accuracy with Claude Sonnet 4

vendor selection
9

weighted factors in every score

audit trail
25+

fields logged per work order

follow-up
−90%

manual follow-up work

the challenge

Work orders were slipping through

Past-due work orders in Rentvine were tracked by hand. Follow-ups took time, happened inconsistently and delayed maintenance, which hurt tenant and owner satisfaction. The team needed an escalation process that could reassign vendors, keep a full audit trail and give management a clear view, without anyone chasing orders manually.

what I built

Three systems, one engine

  1. 01AI vendor selectionClaude classifies each order; a 9-factor score with tag-based priority and fallbacks picks the vendor.
  2. 02Follow-up automationTwo paths: reminders for existing past-due orders and checks on newly assigned vendors.
  3. 03HubSpot loggingCustom work order objects record AI reasoning, vendor responses and manual interventions.
architecture

How it works

Daily check, data fetch, AI classification, vendor scoring, follow-up, logging. Every decision is logged in HubSpot.

  1. scheduleDaily triggerautomated run
  2. data fetchRentvine APIpast-due orders
  3. AI engineClaudeclassification
  4. selection9-factor scorebest vendor
  5. follow-upDual workflowsmart reminders
  6. loggingHubSpotfull audit trail
under the hood

Technical highlights

AI vendor selection

  • Claude Sonnet 4 classification: 94.7% accuracy, with a self-correction step.
  • Tag-based priority search: “BestVendor” contractors are checked first.
  • 9-factor scoring: performance 25%, location 15%, specialization 15% and six more.
  • API resilience: retries and fallback logic for temporary outages.

HubSpot logging and transparency

  • Custom work order objects: 25+ fields track every automated decision.
  • AI reasoning capture: why each vendor was chosen is stored.
  • Multi-step logging: six logging steps, including manual interventions.
  • Dashboard-ready data: throughput, AI performance and exceptions.

Dual-path follow-ups

  • Existing orders: a 3-step sequence with contextual reminders and escalation.
  • New assignments: a 2-step sequence watching the new vendor's first response.
  • Email templates: wording depends on status, vendor response and escalation level.

Integration and processing

  • Metro area matching: vendors are matched by proximity.
  • Dynamic tags: exclusion rules such as DontUse and DontReassign apply in real time.
  • Performance history: completion rates and response times feed the score.
ZapierClaude Sonnet 4Rentvine APIHubSpot APIPython & JavaScript
keep reading

Other ways to handle work orders

Need an escalation system like this?

AI vendor selection, multi-system workflows or audit trails: tell me what your team chases by hand and I'll reply within a business day.

more work

Related case studies

This site uses cookies for analytics. Necessary ones are always on. Read the privacy policy