URBAN COMPANY × HAWKY.AICREATIVE & AGENT INTELLIGENCE · 2026
INTRO — For Joel & the Urban Company performance marketing team

From human-led to agent-run.
Every category, launch-ready in a day. Creative intelligence for the growth engine you’re already building.

Urban Company runs performance across 47 Indian cities and a dozen-plus categories — and you’re already ahead of the market, automating with Claude and moving campaigns from human-led to agent-run. The bottleneck is creative: 7–8 static and 7–8 video masters per category, video stuck at GIF-level automation, and a post-IPO mandate for efficient growth. Hawky is the creative-intelligence and agent layer that closes that gap — entity-level creative decoding, narrative video at master quality, and goal + proactive agents that plug straight into the engine you’ve started.

Hawky.ai · Creative & audience intelligence
Reference accounts · Brands.live
02 / 15
CONTEXT — Where Urban Company is

Category leader, public company, and a growth engine mid-transition to AI.

The brand
  • India’s full-stack home-services marketplace — 51 cities (47 India + UAE, Singapore, KSA JV), listed Sept 2025 at a 57% premium.
  • FY26 revenue ₹1,555 cr (+36%); but a swing to a ₹235 cr loss — InstaHelp burn makes efficiency the board mandate.
  • Ad costs already cut from 21% to 15% of revenue pre-IPO — the next efficiency wave must come from creative & automation, not just budget cuts.
  • 82% of transaction value from retained customers; Native hardware growing 300%+ — many engines, one media budget.
The stack today
  • Performance across Meta, Google & CRM for all core India categories, run by Joel’s team.
  • Already automating with Claude — early success moving campaigns from human-led to agent-run.
  • Design team on Gemini: ~7–8 static + 7–8 video masters per category; one-off quality is good.
  • Video is the unresolved gap — automation has only reached GIF-level, not narrative story videos.
The growth job
  • An end-to-end AI growth engine: creative → launch → optimisation, with humans setting goals, agents doing the work.
  • Master-quality video at automation scale — per category, per city, per cohort.
  • Isolated, clean creative experiments per category — know which entity moved the metric.
  • Defend InstaHelp & core categories against Snabbit, Pronto and vertical rivals — without quick-commerce-style burn.
03 / 15
PROBLEM — The creative bottleneck

The agents are ready. The creative isn’t. A dozen categories × 47 cities × seasonal moments — served by 15 masters per category and GIF-level video.

A category media budget — where performance actually leaks (illustrative)
Creative fatigue · stale masters28%
Wrong cohort match22%
Untested variants18%
Video gap16%
Working spend16%
Same 15 masters rotating for months — CPMs rise as audiences tune outA new-mover ad served to a repeat salon user — message–audience mismatchVariants launched on instinct — losers found after spend, not beforeGIF-level video where rivals run narrative story ads — weakest asset classFresh, matched, tested creative — the share doing the actual work
None of this is a targeting-tools problem — your stack is modern and your agents already work. It’s a creative supply and intelligence problem: not enough fresh masters, no entity-level view of why winners win, and no video engine worthy of the automation around it.
Today — human-led creative loop
  • Design team hand-crafts ~15 masters per category in Gemini; refresh cadence measured in weeks.
  • Video automation stuck at GIF-level — narrative property of a story ad is lost.
  • Experiments overlap; hard to isolate which creative element moved the metric.
  • Learnings live in decks and heads — each new category launch starts near zero.
With Hawky — agent-run creative loop
  • Entity library decoded from every past ad (100–280 attributes) feeds generation — masters multiply into cohort variants.
  • Image-to-video with SFX, music and narrative arc — same-day concept-to-live, master quality.
  • Every experiment isolated by design: one entity changed, results attributable.
  • FeatherDB remembers everything — the Electrician launch playbook compounds into the next category.
04 / 15
AUDIENCE — Six cohorts, six different bookings

‘Home services’ is a dozen businesses. Each cohort books on a different trigger — and repeats differently.

01
Dual-income household
28–45, metro. Salon-at-home, weekly cleaning, massage. The repeat engine — 82% of NTV comes from cohorts like this.
SIGNAL · repeat bookings / 90d
02
New mover
Just moved or renovating. Deep clean + painting + AC install + carpentry in one burst. Highest basket, life-event triggered.
SIGNAL · bundle AOV
03
Appliance SOS
AC dies in May, geyser in December, RO overdue. Urgent, search-led, brutally seasonal — creative must match the moment.
SIGNAL · seasonal booking spikes
04
Native buyer
Considering an RO purifier or smart lock. D2C hardware journey — longer cycle, comparison-driven, 300%+ category growth.
SIGNAL · product page → purchase
05
InstaHelp household
Wants help at the door in 15 minutes. Metro micro-markets, frequency habit — the battleground vs Snabbit & Pronto.
SIGNAL · weekly order frequency
06
Service professional
Supply side. 83% join via referral, but new categories & cities need paid recruitment — a different funnel entirely.
SIGNAL · partner sign-up cost
05 / 15
AUDIENCE — Who → message → where

Entity-level creative only pays when it’s matched. Hawky builds and maintains this map from live booking data.

Cohort
The message that converts
Where — Meta & Google
C1Dual-income household
“Your weekend back — salon, spa and a spotless home, at home.” Comfort + trust entities: uniformed pro, ratings, before/after.
Meta Reels — routine & self-care narratives
Google brand + ‘salon at home’ search
C2New mover
“New home, handled — deep clean, paint, install, done before you unpack.” Bundle entities: checklist, timeline, one booking.
Meta life-event & mover signals
Google ‘painting cost’, ‘deep cleaning near me’
C3Appliance SOS
“AC fixed today — verified expert at your door in hours.” Urgency entities: same-day badge, technician close-up, price clarity.
Google Search — ‘AC repair [city]’, peak-season UAC
Meta weather-triggered creative
C4Native buyer
“Water you can trust, a door that thinks — Native, by Urban Company.” Product entities: specs, warranty, install-included.
Google PMax + Shopping
Meta consideration retargeting, demo videos
C5InstaHelp household
“Help at your door in 15 minutes — trained, verified, whenever.” Speed entities: timer, live map, repeat-booking hook.
Meta micro-market radius ads, frequency capped
Google UAC in InstaHelp geos
06 / 15
APPROACH — What Hawky runs for Urban Company

Six layers, one loop. Built to slot into the AI growth engine you’re already assembling — not replace it.

L1
Creative intelligence
Entity-level extraction across statics & video — 100–280 attributes per creative, clustered by USP, hook and cohort.
LAYER 01
L2
Video engine
Image-to-video with narrative arc, SFX and music — master-quality story ads, same-day concept-to-live. The GIF-level ceiling, removed.
LAYER 02
L3
Competitive radar
Meta Ads Library + Google Transparency Center scraped every 30 minutes — Snabbit, Pronto, Yes Madam and category rivals, live.
LAYER 03
L4
Performance agents
Goal agents chase a metric with daily hypotheses; proactive agents monitor silently and alert on anomalies. Works beside your Claude stack.
LAYER 04
L5
Category launch flow
The Electrician demo, productised: entity library → creative pack → campaign structure → one-click deploy to Meta & Google.
LAYER 05
L6
Expert pod
Hawky specialists accountable for the pilot metric with your team — weekly reviews, guardrails yours.
LAYER 06
07 / 15
LAYER 01 — Creative & audience intelligence   Illustrative preview · your data after integration

One grid: cohort × the KPIs that matter. Every cell backed by the entity-level attributes that produced it.

Creative Intelligence · Urban Company · cohort × booking KPIs
Cohort matrixEntity × cohortCategory view
6
cohorts tracked
SOS
cheapest booking
InstaHelp
best 90-d repeat
Native
longest cycle
BetterWorseEach cell vs your library median · click for the assets behind it
COHORT / SEGMENT
CPB
COMPLETED BOOKING
MED ₹315
CPI
APP INSTALL
MED ₹34
INS→BOOK
QUALITY
MED 11.4%
CTR
HOOK
MED 1.5%
CPM
REACH
MED ₹185
90-D REPEAT
RETENTION
MED 38%
SERVICES
Dual-income household
salon · cleaning · repeat
₹268
repeat engine
₹36
13.2%
1.8%
₹210
54%
New mover
bundles · life-event
₹402
highest AOV
₹41
10.8%
1.4%
₹198
31%
Appliance SOS
urgent · seasonal
₹224
search-led
₹27
15.6%
2.2%
₹165
26%
PRODUCTS & INSTANT
Native buyer
RO · smart locks · D2C
₹1,140
hardware AOV
₹52
4.8%
1.1%
₹240
InstaHelp household
15-min · micro-markets
₹196
frequency play
₹31
12.4%
1.9%
₹172
61%
SUPPLY
Service professional
recruitment funnel
₹540
per sign-up
₹38
7.6%
1.2%
₹146
08 / 15
LAYER 02 — Video at master quality, at feed scale

Past the GIF ceiling. Narrative story videos — hook, problem, service, proof, CTA — generated same-day, per cohort.

INPUT
Masters & entity library
Your 7–8 masters per category + decoded entities from every past winner.
HAWKY · COMPOSE
Narrative build
Story arc assembled per cohort: hook → tension → service → proof → CTA. SFX + music scored.
same day
QUALITY GATE
Master-level check
Brand, claim and quality checks — only assets at master fidelity pass.
approved
VARIANTS
Cohort × city × moment
One story → dozens of variants: language, city, seasonal trigger.
DEPLOY
Meta & Google
Reels, Stories, UAC, PMax — wired into the launch flow on the next slide.
↻ every video’s performance is decoded back into entities — the next batch composes from what actually converted
A · TODAY
GIF-level automation
Gemini one-offs are strong, but automated video plateaus at animated statics. Story-driven ads — the format winning Reels — still need human production time you don’t have at 12+ categories.
B · WITH HAWKY
Same-day story ads
Image-to-video with narrative structure, SFX and music. Concept in the morning, live by evening — per category, per cohort, at master quality.
C · THE POINT
Video is where your creative gap and your rivals’ spend overlap. Closing it converts your automation lead into a visible advantage on feed.
09 / 15
LAYER 03 — Competitive radar

Scraped every 30 minutes. Meta Ads Library + Google Transparency Center + custom scrapers — rivals’ creative strategy, decoded live.

Rival
Where they attack
Their creative play
The gap Urban Company owns
Snabbit
Instant home help in metro micro-markets — your InstaHelp battleground
Speed-first performance creative; aggressive incentives and burn-funded CAC
Trust + breadth: trained pros, ratings history, and a platform the whole home already uses — outlast burn with better creative economics
Pronto
Quick domestic help, quick-commerce playbook
10-minute promise, dark-store-style density messaging
Unit economics story: agents optimise to repeat frequency, not first-order bribes
Yes Madam + vertical salons
At-home beauty — attacks your highest-repeat cohort
Price-led offers, beauty-native influencer content
Premium trust entities: verified pros, hygiene protocol, outcome proof — sell the standard, not the discount
Brand service networks + local discovery
Appliance repair via OEMs, Justdial/Sulekha long tail
Warranty-authority messaging; classifieds volume
Speed + transparency: same-day badges, upfront pricing, 30-min-fresh competitive counters in creative
10 / 15
LAYER 04 — Performance agents

Two kinds of agents, one engine. Goal agents chase the number; proactive agents guard it — beside the Claude automation you already run.

SET GOAL
Human decision
Joel’s team sets the metric: CPB, ROAS or repeat-rate per category. Guardrails defined.
GOAL AGENT
Daily hypotheses
Generates & ranks moves each morning: budget shifts, new variants, cohort changes.
executes
EXECUTE
Meta & Google
Approved moves deployed — bids, budgets, creative rotation.
24/7
PROACTIVE AGENT
Silent monitor
Watches every campaign; no noise while healthy.
on anomaly
ALERT & ACT
Slack in minutes
CPM spike, fatigue, tracking break — flagged with a suggested fix, not just an alarm.
↻ every hypothesis & outcome is logged to FeatherDB — agents argue from evidence, and the audit trail is yours
A · GOAL AGENTS
Chase a metric
“Get salon-at-home CPB under ₹250 in Bengaluru” — the agent generates daily hypotheses, tests them within budget caps, and reports what moved.
B · PROACTIVE AGENTS
Guard the account
No dashboards to babysit. Silent while things are fine; specific and early when they aren’t — anomaly, cause, suggested action.
C · FIT WITH YOUR STACK
Your Claude automation stays. Hawky’s agents handle the creative-and-media layer and expose their reasoning — so the transition from human-led to agent-run stays inspectable.
11 / 15
LAYER 05 — One-click category launch

The Electrician demo, productised. Any category — new city, new season, new service — launch-ready in a day.

Category brief
human input · 15 min
Entity library
what wins in this category
Creative pack
statics + story videos
Campaign structure
cohorts · budgets · bids
One-click deploy
Meta & Google, live
Agents take over
goal + proactive
Today — a category launch
Weeks of briefs, design cycles, campaign builds and QA across teams — each launch a bespoke project.
With Hawky — a day
The Electrician flow you saw in the demo: brief in, entity-informed creative pack out, structure built, deployed in one click — agents optimising from hour one.
Why it compounds
Every launch feeds FeatherDB. Category #12 launches with the accumulated evidence of the first 11 — playbooks, not projects.
12 / 15
ENGINE — What’s under the hood

Six modules, one memory. Everything writes to FeatherDB — so the system compounds instead of resetting each quarter.

DECODE
Creative intelligence
100–280 attributes per creative, statics & video — entity clusters scored against bookings.
COMPOSE
Video engine
Narrative image-to-video with SFX & music — master quality, same-day.
RADAR
Competitive intel
Ads Library + Transparency Center + custom scrapers, every 30 minutes.
AGENTS
Goal + proactive
Daily hypotheses toward your metric; silent anomaly guard on everything else.
LAUNCH
Category flow
Brief → creative pack → structure → one-click deploy.
MEASURE
Booking attribution
Install → booking → repeat wired from your CRM — LTV shapes the buying.
13 / 15
CAPSTONE — AI runs the volume. Humans own the number.

Agent-run, accountably. The transition you’re driving — with a pod that answers for the pilot metric.

STRATEGY
Growth lead
Owns the pilot metric with Joel’s team; sets category strategy and the weekly agenda.
CREATIVE
Creative strategist
Turns entity insights into briefs — statics and story video, per cohort, per season.
PERFORMANCE
Platform specialist
Meta & Google hands-on: structures, bids, agent guardrails, PMax/UAC hygiene.
DATA
Measurement engineer
Keeps booking & repeat attribution honest — CRM wiring, isolated experiments, agent audit trails.
Accountable
The pod is accountable for the pilot category’s metric — reviewed weekly against target. Agents move budgets hourly and propose hypotheses daily; people answer for the outcome, and every agent decision stays inspectable.
14 / 15
NEXT STEPS — The pilot, as discussed

One category, two months, agents live. 2-month pilot → 12-month engagement if it earns it — starting from campaigns you already trust.

WEEK 1
Connect & onboard
Ad account access + pilot budget for the chosen category. Onboarding requires 2+ mature campaigns — the entity decode starts from proven signal.
WEEK 2
Decode & baseline
90 days of creatives decoded into entities; cohort × KPI matrix (slide 07) built on your real numbers; pilot metric & guardrails agreed.
WEEK 3
Creative & agents live
First story-video batch through the quality gate; goal agent live on the pilot metric, proactive agent watching everything.
WEEK 4+
Launch flow & scale
One-click category flow proven end-to-end; weekly reviews vs target through month 2 — then the 12-month decision is yours.
HAWKY
Expert pod
Growth lead · creative strategist · platform specialist · measurement engineer.
URBAN COMPANY
Joel’s team
Pilot category & budget owner, brand guardrails, agent approval gates.
CADENCE
Weekly review
Pilot metric vs target · agent hypothesis log · creative batch approvals · next categories queued.
15 / 15
PILOT PRICING — 2 + 12 month engagement

₹60,000 / month / brand. A scoped 2-month pilot to prove value on 2 use cases — converts to a 12-month Pro plan on success, no further commitment if outcomes don't meet expectations.

MONTHS 1–2 · SCOPED PILOT
₹60,000 / month / brand
Time-bound, hands-on engagement to prove value on 2 specific use cases. Close collaboration between our team and yours to configure, tune, and operationalise each use case.
VIDEO ADD-ON
₹4,000 / video
30–45s videos available on a pay-as-you-go basis during the pilot.
MONTHS 3–14
Converts to Pro
12-month Pro subscription on success. No further commitment if pilot outcomes don't meet expectations.
USE CASE 01
Data-driven Creative Production Agents
A closed-loop creative engine — signals in, on-brand assets out, performance back in. Intelligence signals · production engine · 200 on-brand statics/month, multi-format · performance feedback loop.
USE CASE 02
Competition Agents
Continuous scrape of competitor ads, spend patterns, creative angles & positioning gaps. Competitor creative · spend decode · search-ads analysis · Google spend signals · custom CXO reports.
USE CASE 03 · INCLUDED
Creative Intelligence
Reporting agents · unified dashboard · conversational copilot · goal & optimisation agents · custom dashboards & integrations to your internal data & tools.
Close collaboration
Joint kick-off & weekly working sessions · shared Slack channel · Hawky team configures brand, competitors, signals & agents · co-defined success metrics · iterative tuning based on your feedback. The goal of the pilot is to operationalise these 2 use cases end-to-end inside your team.