AI for the City — by someone who actually lives here.
Hi Jessica — I'm PJ. I live three blocks from the Self-Realization Fellowship. I built this page because the City of Encinitas is one of the few clients in my orbit where the work happens to be five miles from my house, and I'd genuinely love to wire it. $100/hr, two-week pilots, no RFP cycle.
You're not Salesforce-with-a-CIO-title — you're the only IT director for ~600 people, ~63K residents, and a budget that doesn't grow.
Most "AI for government" pitches talk about enterprise transformation. That's not what's happening here. What's happening here is: council prep eats the City Manager's calendar, the planning desk gets the same five ADU questions every week, and the homeless services team is drowning in case-note paperwork.
The AI plays that matter for Encinitas aren't enterprise. They're surgical. One play, two-week pilot, real numbers. If the first one lands, we ship the next. If it doesn't, we stop. That's it.
Surgical, sequenced, shippable.
Council-meeting summarization with action-item extraction.
Public-comment letters arrive in batches of 30+ for hot agenda items. Council members read them at speed, miss themes, and the City Manager prep doc takes a full day. The build: every public comment + staff report runs through a Claude pipeline that outputs (1) a 5-bullet recap per item, (2) a thematic cluster across all letters ("12 letters opposed to X for these 3 reasons"), and (3) flagged anomalies for council attention. Council members get the prep packet at half the page count and 2× the signal.
Build: $8K-$12K · Run: $40-$120/mo · Lift: 4-6 hrs/week recovered per council member + 1 day saved on City Manager prep
Municipal-code RAG — "can I run an Airbnb?" gets a real answer.
Encinitas Municipal Code is 1,000+ pages. Residents call the planning desk asking the same five questions about ADUs, short-term rentals, business licenses, coastal-commission rules. A public RAG layer over the code (+ FAQs + planning department guidance docs) lets a resident type the question and get a plain-language answer with the actual ordinance citation linked. Front desk volume drops; planners do real work.
Build: $10K-$18K · Run: $80-$200/mo · Lift: 30-50% reduction in tier-1 planning desk calls
Smart routing for permit / license / service requests.
Current flow: form fills → email to general intake → manual lookup → forwarded to right department → resident waits 5 days for "we received it." AI replacement: form classifies by type → routes directly to the specific staff member → auto-acknowledges with timeline + next step + what info is still needed. Cuts the 5-day "where's my permit?" loop to <24 hours.
Build: $6K-$15K · Run: $50-$150/mo · Lift: same-day acknowledgment on 90%+ of intake
Internal staff search across SharePoint / docs / institutional memory.
Where is the latest emergency response template? Who's the contact at SANDAG for the bike-lane study? What was the council's position on the 2019 housing-element update? Right now: ask the senior staff member who's been there longest. With AI: ask the search bar. New staff onboarding goes from 6 weeks to 2; institutional knowledge stops being a person.
Build: $5K-$12K · Run: $60-$180/mo · Lift: ~3 hrs/week saved per knowledge worker
Citizen FAQ assistant — "when is trash pickup on Christmas week?"
A small, smart assistant on encinitasca.gov that handles the 80% of resident questions that aren't actually code or planning ones: hours, holidays, park reservations, beach rules, recycling, council-meeting attendance, public records request status. Routes the 20% that need a human directly to the right department. Reduces front-desk call volume; respects residents' time.
Build: $4K-$10K · Run: $40-$120/mo · Lift: -25% inbound general-line calls
Encinitas isn't a Workday/Salesforce situation — that's why it could ship.
- Coastal-town quirks matter. ADU rules vs Coastal Commission, the Olivenhain–Cardiff–Leucadia–Old Encinitas community identity, parking on 101 — none of which a generic GovTech stack accounts for.
- ~63K residents is small enough to ship a focused tool without a 14-month procurement cycle. The pilot is two weeks, not two years.
- No "phase 1 visioning workshop" — I write the spec on a napkin, ship the v1, you and your team kick it. If it works, we keep it. If not, we delete it.
- You own everything I build — the prompts, the eval set, the deploy. No vendor lock-in. Total transparency. If a future CIO wants to take it in-house, the handoff is one Github repo + a README.
- Same town, same time zone, same surf check. If something breaks, I'm a five-minute drive away.
The Real ROI Math
Conservative numbers for one play (Council Summarization · Play #01):
Council prep before: 7 council members × 4-6 hrs/week reading public comments + staff reports = 28-42 hrs/week of council time + ~8 hrs of City Manager prep.
After: 30-min skim of an AI-generated thematic packet + targeted deep-reads on flagged items. Council time: 10-14 hrs/week. CM prep: 2 hrs.
Fully-loaded council compensation + CM time at conservative $80/hr public-sector rate: recovered 18-32 hrs/week × $80 × 50 weeks = $72K–$128K/year. Build cost: $10K. Year-1 ROI: 7-12×. The other 4 plays compound on top.
What I'd love to ship first.
Council-meeting summarization (Play #01). It's the most visible win — council members feel it the first agenda packet that arrives. Two weeks from "yes, let's try" to working v1. $8K-$12K depending on which agencies/feeds we pipe in. If it works, we ship Play #02 next quarter. If not, we stop — and you've got a $10K experiment that's still cheaper than one consulting workshop.
Not a procurement portal. Not a 200-page SOW. Just two messages back and forth, an hour over coffee on 101 (you pick the spot — I'm partial to Pannikin or Java Hut), and a written one-page scope before any work starts.
Jessica — I built this page on a Friday morning at my kitchen table in Cardiff because the City of Encinitas is the rarest kind of client: local, real, and small enough to ship something that actually matters. If any of these five plays speaks to a problem you're already wrestling with, two messages and we'll know if it's worth a coffee. — PJ
Coffee on 101?
Pannikin, Java Hut, Better Buzz — your pick. 30 minutes, no deliverables, no procurement. If we like the conversation we draft a one-pager. If not, you got a coffee on the gold-dome side of town.
This page is built to evolve. Pick a mode, I'll text back.
Hand-crafted for the City situation, but not finished. Want me to swap the lead play, drop a different department in (Public Works? Library?), or scope a totally different idea you've been chewing on? Three modes, three texts, no decks.
Send this to the City Manager — or whoever should see it.
If a colleague at the City would also want this in their inbox (CM, Deputy CM, Council aide, Comms lead), one click puts the link + your note in your email — you hit send.
No tracking, no list. Your email app does the work.
