Case studies

Real AI systems, running in production.

Three controlled AI systems managing real business workflows — built and operated by Alborz Fallah. Contact-form triage, finance operations, and marketplace sales automation. Not demos. Not prototypes. Production systems with measurable outcomes.

Operational agent

Content Bot

Morning intelligence briefings, contact-form triage (10-min cycles), editorial QA monitoring, dashboard maintenance, and repository operations — across 14 connected systems.

📧 Contact-form detection: manual → 10-min automated cycle
📊 14 systems: Telegram, GitHub, Postgres, Pipedrive, GA4, Xero, SMTP, email, Slack, Railway
🛡️ Fixed routing tables — no hallucinated decisions
⏱️ Every weekday morning briefing at 8:30am Brisbane

Read the full case study

Finance operations

FinanceBot

Monitors finance email, Slack, Xero, Airwallex and Employment Hero. Learns from thousands of real finance messages. Automates bill forwarding, receipt checks, monthly reimbursement reminders and ad hoc reporting.

📚 3,368 learned finance items (email + Slack)
💰 Found $2,547 missing receipt; answered $35,970 Aircall spend query
🏦 Xero, Airwallex, Employment Hero (174 employees), Gmail, Slack
🔒 Salary data restricted to 4 authorised people — documented and enforced

Read the full case study

Sales operations

MarketplaceBot

Scans 450+ Pipedrive deals every weekday morning, finds activity exceptions, monitors marketplace email, and delivers same-day reports for the leadership team.

🔍 Daily 8:10am scan: 450 deals → 74 activity exceptions found automatically
📊 YouTube leads report: 50 deals analysed, CSV delivered same day
📧 10-min email monitoring with auto-reply for verified internal senders
🛡️ Read-only Pipedrive by default — no record changes without approval

Read the full case study

The pattern

Every one of these systems follows the same design.

Start narrow One workflow

Contact-form triage. Bill forwarding. Daily Pipedrive scan. Prove the workflow before expanding access.

Read-only first No write access until proven

Every system starts read-only. Graduates to draft-only, then write-after-approval. Autonomous action only after months of evidence.

Human owns the rules Fixed routing, named approvers

No hallucinated decisions. Fixed templates. Named people approving actions. Every outbound action traceable.

Operate, don't abandon Every system has an owner

Daily monitoring, weekly reviews, monthly governance. AI systems need ongoing care, not a launch date.

Connect real systems Not just chatbots

These systems read email, query CRMs, check accounting, pull analytics, monitor Slack — and route work to the right person.

Know when to stop Escalation rules built in

If confidence is low, a topic is sensitive, or the answer matters — the system escalates to a human instead of guessing.

What these systems prove

AiBorz is not a theory. The patterns we use — read-only first, human approval gates, fixed routing rules, operational monitoring, and graduated access — are the same patterns running these three production systems every day. The approach works because it was built from real implementation, not from a whitepaper.

Every AiBorz client engagement starts with the same question: "What is the one repetitive, measurable workflow where AI could remove admin drag without removing human control?"

Find the workflow worth building first.

Start with the free AI Readiness Scorecard to check your systems, data and readiness — then book a fit check to discuss your first workflow.

Start the AI Readiness Scorecard