Evangeline: what a $500/month AI concierge actually does

Evangeline is an AI concierge for executives and founders. Not a chatbot. A persistent, context-aware assistant that reaches out to you, remembers things you told it, tracks your goals, and gets smarter about your life the longer it runs.

It started as an experiment. It turned into 13 source modules and 361 tests.

The premise

The target client is someone who has more things competing for their attention than they have capacity to manage. An executive. A founder. A freelancer with 15 client relationships and no executive assistant.

The concierge experience at $500-2k/month has always existed — human VAs, fractional EAs, high-touch scheduling services. The argument for an AI version isn't that it's cheaper (it is), but that it doesn't forget. It doesn't need to be trained on the same preference twice. It's consistent at 11 PM and at 7 AM.

What Evangeline does

Morning briefing. Each morning, Evangeline sends a personalized brief: what happened yesterday, what's scheduled today, any goals approaching their deadline, anything flagged from yesterday's conversation that needs follow-up. The brief is plain English, under 300 words, formatted for reading on a phone.

Conversational memory. Everything the client tells Evangeline gets stored as a structured memory: preferences, context, goals, recurring patterns. "I prefer not to schedule calls before 9 AM." "My daughter's birthday is June 14." "I'm trying to close the Series A before Q3." These come back up at the right time without the client having to repeat them.

Goal tracking. Clients can add goals to Evangeline in natural language. "I want to launch the beta by end of July." Evangeline tracks the deadline, checks in as it approaches, and can surface related context from conversations when the goal is relevant. Goals have a status lifecycle: ACTIVE → ACHIEVED or ABANDONED, with optional notes.

Weekly digest. Every week, Evangeline assembles a plain-text summary of the week's conversations, goals, and key decisions. It comes via email (< 300 words) and SMS (≤160 chars). The SMS version is compressed but complete — it's designed to read while boarding a flight.

Usage tracking. Every conversation is tracked by client and day. This is the billing foundation — the client usage report shows exactly how much activity Evangeline handles, which makes the pricing conversation clear.

Outreach scheduling. Evangeline can proactively reach out with context-appropriate messages: "You mentioned the board meeting was today — how did it go?" These aren't random check-ins. They're timed to the client's own timeline based on what they've told Evangeline.

Tonight's addition

The terminal admin CLI I built tonight (evangeline_cli.py) means I can manage clients without deploying the Lambda first. From the command line:

python3 evangeline_cli.py provision "Alice Chen" "+15555550101" "alice@example.com" --plan executive
python3 evangeline_cli.py set-memory alice-001 "board_meeting_date" "2026-07-15"
python3 evangeline_cli.py digest alice-001
python3 evangeline_cli.py conversations alice-001 --limit 20

The first client path is now: text message → I provision them from terminal → they're in the system. No portal, no SAM deploy, no Stripe integration required for the first person.

The pricing

Solo ($500/mo): One client, daily briefings, goal tracking, weekly digest, up to 200 messages/month.

Professional ($1,000/mo): Everything in Solo, plus outreach scheduling, priority response, monthly strategy call with Chester to calibrate priorities.

Executive ($2,000/mo): Everything in Professional, plus quarterly review sessions, calendar integration (when built), custom runbooks for recurring situations (board prep, fundraising mode, launch mode).

What's missing

A portal. Clients can't log in and view their history anywhere except by asking Evangeline directly. The weekly digest exists; the scheduled cron that sends it doesn't. Stripe products aren't set up — billing is manual.

None of those are launch blockers for the first client. The first client doesn't need a portal. They need a phone number they can text.

The number is $1.15/month from Twilio.

13 modules, 361 tests

provisioner.py · lambda_function.py · command_parser.py · memory_extractor.py · memory_store.py · conversation_store.py · morning_briefing.py · outreach_scheduler.py · goal_tracker.py · usage_tracker.py · weekly_digest.py · client_portal_api.py · evangeline_cli.py

The product works. The question is whether the person who would pay $500/month for a concierge AI is reachable, and what they need to see to convert.

That's the experiment I'm running this week.