DroneBlocks

Support Review — Last 90 Days

Everything hitting support@droneblocks.io, categorized — plus a cross-check against recent meeting notes.

prepared 2026-06-22for Henrywindow Mar 24 – Jun 22

// headline

What the inbox is telling us

Counts are approximate — categorized from subjects/senders. A precise tally needs triage labels (recommendation #1).

1

support@ is an un-triaged dumping ground.~200 threads in 90 days, but the large majority is automation/noise — Teachable signup bots, ShipStation/UPS tracking, McMaster order confirmations, bid portals, newsletters, even a tax notice. Real requests are buried. No labels, no routing, no SLA.

2

The biggest human category isn't support — it's sales.~Half of genuine inbound is pricing / quote / PO / enterprise-membership requests (South Lewis, MG, Strikewerx, Macquarie, Code Easy UK, White Plains, Saudi, Peru). Real leads sitting in a support inbox — a speed-to-lead and attribution risk.

3

The repetitive, automatable requests are "access / trial / passcode.""New educator, want to try the sim," "can I get the password," "what's this month's Tello passcode?" (one user asked 3× in a thread). Prime AI-support-agent workload — and a signal the self-serve trial + passcode UX confuses people.

4

Real technical issues cluster on Simulator v2 UX + reliability.Missing Launch button, the "all simulators" page gone, a server scaling outage, and a Mac/Apple-Silicon app that Gatekeeper blocks. All fix-at-the-source — each erases a whole ticket category.

5

A hardware-quality signal.A DroneMinds contact reports a propeller failing on every drone they test, plus replacement-parts requests. Systemic, not a one-off → incoming-QC signal.

6

The team responds well — but 100% by hand.Replies often land same business day (one customer answered in ~12 min; another got a custom Apple-CPU build next day). Responsiveness isn't the problem — it's all manual, unlogged, unmeasured, with no "no inbound disappears" guarantee.

// breakdown

Genuine human inbound

Approximate share, with examples.

CategoryShareExamples
Sales / pricing / PO / enterprise~halfSouth Lewis, MG, Strikewerx, Macquarie, Code Easy, Saudi, Peru
Trial / access / passcode~quarterLarry, Wesley, Shane, Mike, km@scu.dk
Technical / how-to / product~fifthSusan Bock (sim UX), Luigi (Mac app), UTTC (server + STL), DroneMinds (props)
Demo / loaner logisticsa fewKerrigan return, Syosset
Billing / churna fewFailed Teachable billing (Cecilio Pérez)
Non-customer noise (shouldn't be here)the bulkMcMaster, LiteBee, ZeeMaps, tax, newsletters, signup bots

// offload

What the AI support agent should own

droneblocks-ai-support — clear the repetitive half; humans keep sales + real technical.

// fix-at-source

Cut tickets before they happen

Sales leakage: half of human support@ volume is leads. They're caught by diligent humans today, but with no first-response SLA, no source attribution, and no pipeline visibility until someone forwards them. Route support@ sales inbound into HubSpot automatically, tagged source = support-inbox.

// cross-check

Fathom completeness re-check

Re-scanned all 17 recaps. The findings doc holds up; three things the email data adds.

// decisions

What to decide

1

Stand up triage on support@ (labels/rules) so we can measure the categories above.

2

Scope the AI agent to the trial/access/passcode/how-to half + sales routing.

3

Pick 3 source-fixes this quarter: self-serve trial, passcode UX, sim v2 nav.

4

Decide Flare Star vs. LiteBee for light-show drones.

5

Get supplier/ops mail off support@ — it's doubling as our procurement inbox.