Skip to content
Papa Corporation

Case studies

Six engagements. All shipping in production.

Each one started with a problem that hurt enough to fix, ran for four to twelve weeks, and ended with a result the client could point at without me in the room. Names omitted. Numbers real.

The permit agent that paid for itself in a week

The problem
A Miami real-estate developer was bleeding hours every week chasing reviewer comments across half a dozen Miami-Dade permits. Manual logins, copy-paste into spreadsheets, drafting responses by hand, chasing consultants. Comments slipped through. Deadlines slipped after them.

What I built
An agent now watches the permit system every business hour. It catches every change, drafts the applicant response, and sends three emails: an instant alert when a comment lands, a morning briefing, and an evening digest. The dashboard rebuilds itself after every check. Multi-permit on one screen.

Inside the build

  • Live dashboard rebuilds after every check
  • Multi-permit support, all on one screen
  • Duplicate emails are structurally impossible
  • Runs entirely unattended

Outcomes

  • Project closed weeks ahead of schedule, saving the developer over $100,000.
  • Zero reviewer comments missed since deploy.
  • Response cycle compressed from days to minutes.
AI Agent·Permit Automation·Real Estate

Invoices in, structured ledger out

The problem
A restaurant operator was buried under PDF invoices from a long tail of vendors. No visibility into reorder cadence, vendor concentration, or the prices that quietly crept up between quarters. Stockouts happened. Margins eroded.

What I built
Each PDF lands and parses on arrival into structured rows. A reorder engine watches how often each item normally ships. A pricing engine watches what is creeping up. The kitchen manager runs the whole thing from a phone. Mobile-first, easy enough to learn between services.

Inside the build

  • PDF invoices in, structured rows out
  • Reorder alerts before stockouts, not after
  • Price drift visible the day it happens
  • Mobile-first, easy enough for a kitchen manager

Outcomes

  • Reorders surface before stockouts, not after.
  • Price drift visible the day it happens, not at quarter end.
  • Paper-invoice triage replaced with a single screen.
AI Document Parsing·Inventory Automation·Hospitality

An AI sales agent that does not get bored

The problem
The sales team at a rigging-and-cable distributor was building quotes by hand off inbound RFQ emails across nearly 4,000 SKUs. Wire rope, cable, chain, and rigging hardware come in dozens of variations. Customers send abbreviations, typos, Spanish names, and Outlook mobile messages with broken line breaks. The team was losing deals to faster competitors.

What I built
The agent now reads every inbound RFQ, parses every line item, identifies the right product, prices it against the live ERP, checks live inventory across four warehouses, and drafts a complete quote in minutes. It handles the abbreviations and the typos. The alias table is self-learning and AI-poisoning resistant. Margin alerts fire on lines under 30%. Customer-cycle reports run weekly. Sales managers can approve and reject from inside the email.

Inside the build

  • Over 90% of items resolved without an AI call
  • Self-learning alias table, AI-poisoning resistant
  • Over 10,000 learned product aliases, growing nightly
  • Daily margin-leak alerts on lines under 30%
  • Weekly pricing drift reports and rep scorecards
  • Daily lapsed and due-soon customer report at 8 AM
  • One-click approve and reject from inside the email
  • Customer-facing tracked-link quote portal

Outcomes

  • Response time dropped from three days to four hours.
  • Three times the volume with the same headcount.
  • Win rates up 22%.
  • Used daily by the entire sales team.
AI Sales Agent·RFQ Automation·ERP Integration·Distribution

What a financial firm was missing across 200 mailboxes

The problem
A financial services firm had no real visibility into what was happening inside their 200-plus mailboxes. Forwarding rules, the kind that quietly mail every CFO message to a personal Gmail, are invisible to native security tooling. Nobody had ever looked.

What I built
InboxWatch ran a metadata-only sweep across every mailbox: a hundred checks per inbox, ten threat categories, no message bodies touched. The forty-seven forwarding rules surfaced in the first pass. One of them had been quietly mailing invoices and signed contracts to a personal address outside the firm for the better part of a year. We closed it the same hour and kept the evidence trail the auditor would want.

Inside the build

  • 100+ checks across 10 threat categories
  • Hidden forwarding rules surfaced instantly
  • Risky third-party app grants flagged
  • Weekly automated scans on autopilot

Outcomes

  • 47 hidden forwarding rules and 12 risky app grants caught in the first scan.
  • All findings remediated within 72 hours.
  • Zero forwarding-based incidents since.
InboxWatch·Email Security·Identity Hardening

A Sunday afternoon at a law firm

The problem
A law firm partner's email started behaving strangely on a Sunday. Sent items the partner did not recognize, MFA prompts at 3 AM, mailbox rules that were not there the week before. Client data and active matter correspondence sat in that inbox.

What I built
Live on a call within 30 minutes. Contained inside two hours: every active session revoked, persistent app tokens the attacker had been using to bypass MFA killed, credentials rotated with phishing-resistant authentication. Forensics traced lateral movement to two more accounts and contained both the same hour. Timeline ready for the insurer, the bar, and counsel by Monday.

Inside the build

  • 30-minute response, 2-hour containment
  • 90 days of sign-in logs forensically mapped
  • Phishing-resistant multi-factor rolled out firm-wide
  • Real-time anomalous mailbox monitoring

Outcomes

  • Breach contained in under two hours.
  • Lateral movement traced and remediated across three accounts the same day.
  • Operations restored next business day, with a timeline that survived the insurer, the bar, and counsel.
Incident Response·Email Compromise·Forensics

CMMC Level 1 for an ITAR aerospace shop

The problem
An ITAR-registered aerospace manufacturer in Florida had Federal Contract Information flowing through email, customer portals, an ERP, and Dropbox. Their next contract opportunity required a CMMC Level 1 attestation. Seventeen controls across six domains. Zero documentation.

What I built
Complete Level 1 toolkit: gap assessment mapping current state against all 17 controls, five plain-English policies (acceptable use, access control, incident response, media protection, physical security), a system security plan ready to file, a remediation playbook with evidence tracking, and an SPRS self-assessment worksheet. Each policy was written to reflect how the team actually works, not how an auditor wishes they worked. Thirty days of post-delivery support to walk through SPRS submission and answer questions during early implementation.

Inside the build

  • All 17 CMMC Level 1 controls covered across 6 domains
  • 5 plain-English policies, not legal boilerplate
  • System security plan ready to file
  • Remediation playbook with evidence tracking
  • SPRS self-assessment worksheet, fillable
  • 30 days of post-delivery support

Outcomes

  • Level 1 documentation package delivered in weeks, not quarters.
  • The team can read and follow every policy without a consultant in the room.
  • Level 2 scope already mapped for the next contract tier.
CMMC·ITAR·Aerospace & Defense

The common thread

How every project above was actually run.

Same four principles across all six. They are how the work survives the hand-off and keeps producing outcomes after I am gone.

  • Same operator from kickoff through hand-off. No analyst stack.
  • Free first week. Written scope and flat fee by Friday.
  • Working software every Friday. No status decks, no status calls.
  • Outcomes measured in money, time, or risk. Documented before kickoff.

Yours could be next

Tell me what you are working with. I'll tell you if I can help.

A paragraph is plenty. I'll reply by Friday with a real read on whether this is a fit, what the work looks like, and a number you can plan against.