Day ’n’ Nite io Book a buildout callBook a call

Portfolio · screenshots from the real automations, names blurred

Three signature builds.
Yours awaits.

Each one runs for home improvement clients right now. The explainer video shows how they fit together; the screenshots below are the real thing.

These come from client operations where I worked alongside other people and software. I describe my part, show the evidence for it, and say where that evidence ends.

  1. 01 Lead Scoring
  2. 02 GHL Managed Agents
  3. 03 AI Edge Case Reviewer

The signature builds

HighLeveln8nMeta Conversions API

Lead Scoring

Every new lead, from a Meta instant form, a landing-page survey or a website survey, is valued and scored the same way.

  • Check and value. Address verified, home valued (Zillow, then Redfin, then Realtor.com).
  • Score. Compared to that client's own threshold, then tagged high quality or not.
  • Report. Sent to Meta as a value-tiered conversion event, deduplicated per source.
  • Alert. The point person gets a text and app push with the home value and a satellite photo, held overnight during quiet hours.
  • Close the loop. When the CRM reaches Estimate Sent or a signed job, the real quote or contract dollars go back to Meta.

Built for home improvement client operations.

HighLevel Agent StudioAI agents

GHL Managed Agents

Three managed agents configured for a client location, each limited to the permissions its job needs.

  • Appointment detector. Can move the CRM stage when a lead agrees to a visit; its configured scope excludes stages that carry dollar values.
  • Name cleanup. Fixes contact names typed into ad forms.
  • Nurture safety net. Checks whether a repeat lead is eligible to re-enter a nurture sequence and records the decision.
  • Audit trail. Configured workflows can leave a verbatim note, and the skill list is locked against agent edits or deletion.
  • Permission audits. I review agent permissions and available actions through the API.

Managed separately for each client location.

Server-side ClaudeMulti-model reviewn8n

AI Edge Case Reviewer

A renter, a lot or land parcel, a missing home value, a commercial address. This reviewer researches cases the standard rules cannot resolve and records its decision.

  • Rulebook. A server-side Claude agent researches the lead and applies a written rulebook.
  • Agreement. The configured workflow requires agreement at or above its 90% confidence threshold before a CRM write.
  • Escalation. A split goes to a two-model council, then a third-model tiebreaker.
  • Safe default. Anything still tied is configured to remain a normal lead for review. A daily digest lists recorded decisions.

Built to review and organize uncertain leads, keeping your CRM updated

[ 02 ] Next

Want one of these in your business?

You get a clear map of how the work runs, a tighter process, and automation that holds up: built in your tools, people on the judgment calls, tested before it's yours.

Book a buildout call