CASE STUDY · 01 AI MARKETING · 2024–PRESENT

Tec-Creative — AIGC creative platform for global performance marketing.

A creative system that connects briefs, channel constraints, brand memory, and performance feedback across multi-market, multi-language, multi-channel workflows.

Role: AI Product Director
Org: Tec-Do
Surface: SaaS web app + agent runtime
Audience: Global growth teams

§01 Problem

Creative production for performance marketing is fragmented across briefs, asset libraries, design tools, channel platforms, and analytics. Every campaign starts from scratch even when the same brand, audience, and channel rules apply.

Generic image and video generators don't fix this — they produce assets, not campaigns. The brand voice is not preserved. Channel specs are not respected. Performance feedback never reaches the next iteration.

§02 Approach

Treat creative production as a workflow loop, not a generation tool. Each brand maintains structured memory: tone, motifs, claims, banned words, regional nuances. Each channel maintains structured rules: aspect ratio, copy length, safety, format. The system composes these with the brief at generation time.

brief

goal audience market channels

memory

brand kit tone motifs claims past winners

constraints

channel specs policy localization QA

generate

image video copy variants

launch

creative pkg A/B sets handoff

feedback

CTR / CVR retention → memory

§03 Product surfaces

// brief composer

structured intent capture (goal, KPI, audience)
auto-pull from past winners
localization and market routing

// brand memory

tone of voice + motif library
claim policy and approved phrasing
banned visual + verbal terms

// channel runtime

aspect, length, safety, format per channel
auto-resize and re-cut
copy length adaptation

// performance loop

winner detection by CTR / CVR / hold
creative DNA labeling
memory writeback for next campaign

§04 Generation grammar

Generation is invoked through a structured grammar that the runtime composes from brief + memory + constraints. The model is a substitutable component; the grammar is the product.

$ tec generate \
  --brief="weekend sale, gen-z, US"
  --brand=acme
  --channels=tt:9x16,meta:1x1,gads:16x9
  --variants=8
  --memory=winners:30d
  --policy=acme/v3

§05 Outcome shape

What changes for the user, qualitatively. Numbers vary by advertiser and are not published here.

Campaign cycle

days → hours

Brief to launchable creative pack.

Variants per brief

10×

Without losing brand or channel fit.

Knowledge retention

writeback

Winning DNA returns to the next brief.

§06 What I did

// product

Defined the workflow-loop product thesis
Owned generation grammar and runtime
Brand-memory schema and policy layer
Channel-runtime spec across TikTok / Meta / Google

// org

Cross-functional alignment with creative, ML, GTM
Customer co-design with global growth teams
Pricing, packaging, ICP definition

← back to davelee.ai