AI agents · real client workflow · live system
From field photos to a client-ready report in one run.
The CompanyCam Assessment Agent is the “beef” of this presentation: a real automation that pulls field data, structures it, generates Docs/Sheets/report outputs, and writes the summary back to the source system.
The bottleneck
Vince already has the data. The problem is everything after the site visit.
CompanyCam is where the work starts: photos, window labels, comments, field notes, and project context. The old workflow is manually turning that scattered field data into a professional assessment package.
St. Mark's Episcopal Church Assessment
412 Oak Street, Chattanooga, TN · stained glass survey
Exact workflow
The agent is an extraction pipeline, not a chatbot.
It starts where the contractor already works, then produces every downstream artifact the business needs.
Input
CompanyCam project
Project metadata, address, photos, comments, tags, and notepad context.
Pull
API fetch
Gets paginated photos, per-photo tags, per-photo comments, and image URLs.
Parse
Panel extraction
Claude-first structured extraction with deterministic fallback parser.
Structure
Assessment data
Window, panel, dimensions, issues, severity, condition, notes.
Generate
Docs + Sheets
Field notes doc, color-coded sheet, Excel workbook, and full report.
Write back
CompanyCam summary
Updates the project notepad with counts, critical windows, and output links.
What it extracts
Raw field notes become structured business records.
Raw CompanyCam comment
Window 1 - East east exposure, bell tower 2 panels 30x82 a. top 30x40 - 2 breaks, sig lead deterioration, cat 2 warping, past repairs failing, poor b. bottom 30x42 - 1 break, moderate lead deterioration, cement eroding, water entry, fair
Structured panel record
{
"window_number": 1,
"exposure": "east",
"location": "bell tower",
"overall_dimensions": "30x82",
"panel_letter": "a",
"panel_dimensions": "30x40",
"num_breaks": 2,
"lead_deterioration": "significant",
"warping_category": "cat 2",
"past_repairs_failing": true,
"condition": "poor"
}
Run workflow
What the audience watches live.
This slide is the calm walkthrough. The final slide opens the real app and runs it against CompanyCam.
CompanyCam Project ID
Presenter line: “The button is not asking AI to improvise. It starts a fixed workflow: fetch → parse → generate → write back.”
What comes out
One run creates the whole assessment package.
Google Doc
Organized field notes
Raw observations regrouped by window so Vince can audit what the agent extracted.
Sheet / Excel
Assessment schedule
Color-coded panel table with condition counts, critical windows, and issue columns.
Report
Client-ready narrative
Reference-style Word/Google Doc report with cover, sections, valuation, summary, and appendices.
CompanyCam
Source-system writeback
Notepad summary gets updated with output links and condition breakdown.
Report proof
The output is not a generic AI write-up.
Appendix 4 · Site Notes
Window-by-window proof with photos and measurements
• 2 glass breaks
• significant lead deterioration
• warping cat 2
• Poor condition
b. Bottom 30x42
• 1 glass break
• moderate lead deterioration
• Fair condition
32
windows present in the polished report artifact
136
panel records represented in the final assessment
32
embedded image files detected inside the Word report
9
reference-style sections: overview, condition, 100 year plan, summary, appendices
Time saved
For this workflow, the savings are almost a full workday per assessment.
Conservative manual baseline
manual work
− live runtime
= time saved
Baseline includes CompanyCam collection, panel normalization, spreadsheet formatting, report assembly, upload/share links, and CompanyCam writeback.
Manual total
10.15h
conservative estimate
Report scale
136
panels processed
Reduction
97%+
for 10–20 min live run
From the other agent inputs
Where this fits inside the ZC agent menu.
The CompanyCam demo is the clearest proof, but the same build pattern maps into four sellable categories for ZC clients.
Audit a client site against competitors and return the ranked backlog: keywords, content gaps, technical fixes, and why they matter.
Find, qualify, enrich, and package B2B prospects so a rep starts with context instead of blank research time.
Draft the repeatable 75%, then flag the few strategic sections that still need Robert’s judgment.
Pull data from wherever it lives and ship the recurring Docs/Sheets/reports that quietly eat the week.
More of the work
A few other agents already in the portfolio.
Lead generation
Find & enrich B2B prospects.
Built for a services client. The agent searches, qualifies, enriches contact routes, and packages each lead with the context a sales rep actually uses.
SEO & competitor analysis
From “where do we start” to a ranked backlog.
Point it at a client site and competitor URLs. It audits keyword overlap, content gaps, technical issues, and priority fixes.
Administrative reporting
The recurring report, on autopilot.
The same pattern as CompanyCam: pull source data, normalize it, build the deliverable, and put it where the team expects it.
Built for ZC
An AI proposal generator for ZC — Robert plus agent.
Robert’s bottleneck
Robert spends thirty-plus hours writing a single proposal. The agent drafts the repeatable 75% in under five minutes, then tells him exactly which sections still need strategic input.
Partnership
How Clawdified plugs into ZC.
Why this matters to ZC
This is the pattern behind useful agents.
A ZC proposal agent is the same shape.
Pull the source material, structure the messy inputs, generate the client-ready package, flag what still needs human strategy, and save the finished output where the team already works.
Live reveal
Now run the actual CompanyCam agent.
Over to you
That is the work. What should we dig into?
We can open the CompanyCam project, inspect the generated report, walk the code, or talk about how the same agent pattern maps to ZC proposals.
Built by
Wes · Clawdified
Main demo agent
CompanyCam Assessment Agent
Local path
/workspace/bots/vince-agent
south exposure, sanctuary
5 panels 29x161
a. top 29x38 - 2 breaks, sig lead deterioration, cement eroding, poor
b. 2nd from top 29x38 - cat 2-3 warping, lead splitting, poor