I run INSIDEA. 150 person team, 1,500+ businesses served, three other ventures, a household with a one-year-old. The math on email gets ugly fast.
Tuesday wasn't unusual. 217 unread when I sat down at the desk around 5pm Bangkok time (my work day starts when most people's ends). Gmail had 184. Slack had the rest. The screen looked like a wall.
I opened Pranan. It had been working since 6am.
What Pranan caught
The morning briefing was the first surprise. 30 seconds of reading and I knew what mattered. Three things needed me. One investor follow-up where the timing was sensitive. One client escalation Kiran had flagged overnight. One Slack thread in our HubSpot team that needed a call I had to make.
Everything else? Already drafted or already filed.
Pranan classified each thread by relationship tier. Investor emails got the warm-but-direct tone I use with our cap table. Vendor emails got the polite but no-nonsense reply pattern. Internal team comms got my actual rhythm with each person, including Marshall's preference for short answers and Disha's preference for bullet points.
Of the 174 it handled:
- 92 routine threads were auto-filed without a draft (subscriptions, automated reports, calendar updates, things that need no human reply)
- 61 had drafts ready that I sent without editing
- 21 had drafts I tweaked one or two words on before sending
That's the boring part. The interesting part is the 43 it didn't handle.
What Pranan missed
Three categories of miss, and I think they're the actual map of where AI drafting still needs work.
1. Context that lives outside the inbox
One thread was a follow-up from a partner I'd had coffee with three weeks ago. Pranan didn't have the coffee in its memory. The draft was technically correct but missing the warmth that would have come from referencing that meeting. I rewrote it from scratch.
This is fixable. Pranan's relationship intelligence layer can ingest more context over time. It just needs to.
2. Decisions that look routine but aren't
A client email asked what looked like a simple scheduling question. The right answer involved a contract renegotiation conversation that was about to happen. Pranan didn't know that. Its draft would have committed me to a date that conflicted with the bigger conversation. I caught it because I was paying attention. If I'd been on autopilot, that would have been a real problem.
The lesson: AI drafts are tools, not decisions. The day you stop reading what you send is the day you stop being a thoughtful operator.
3. Tone calibration on hard moments
One investor wrote with a question that needed careful framing. Not bad news, just sensitive. Pranan's draft was good. It just wasn't quite mine. The difference was small, maybe ten percent, but it would have read differently to someone who knows me.
I rewrote it. The original draft helped me think through the structure. The final version was mine.
What this taught me
Three things, none of which I expected.
First, the morning briefing is the killer feature. Not the drafts. The drafts are great when they work, but the briefing is what saves my actual day. 30 seconds of "here's what matters" replaces 45 minutes of inbox archeology.
Second, AI drafting works best when the relationship is well-mapped. Threads with investors I email weekly: drafts almost always send-ready. Threads with people I email twice a year: drafts are okay but need work. The relationship signal is the thing.
Third, the goal isn't inbox zero. It never was. The goal is showing up well for the people who matter. Some days that means sending 60 drafts in 20 minutes. Some days it means not sending anything for an hour because the next email needs more thought than a draft can give it.
What I'm doing differently now
I'm using Pranan as a triage layer first, drafting layer second. The briefing tells me what to focus on. The drafts handle the volume so I can focus.
I'm tagging threads where the AI shouldn't draft. Some relationships need to feel hand-written every time. Pranan respects the tag.
I'm reviewing the relationship classifications weekly. The AI gets most of them right, but a few need correcting, especially for people whose relationship to me is shifting (an investor becoming a client, a vendor becoming a partner).
Tuesday at 9am, I had 11 things to actually decide on. Not 217. That's the number that matters.
If you're running multiple things at once and the inbox feels like it's running you, try the free plan. 300 drafts a month, no credit card. See if it sounds like you.