P4-35 AI: Auto-Generated Campaign Copy

LLM generates email subjects, body copy, and nudge text — tuned to your brand voice.

---

Scoring Card

Dimension	Score	Rationale
Pain	2/5	Copy writing is tedious but not blocking — teams ship with mediocre copy
Revenue	2/5	Nice-to-have for upsell, not a primary purchase driver
Build	3/5	LLM API integration is straightforward; brand voice tuning adds complexity
Moat	2/5	LLM wrappers are low-moat; value is in the integration context
Total	9/20

---

Classification

Vitamin AI Layer

Vitamin

Auto-generated copy saves hours of writing time per campaign and unlocks A/B testing at scale (because generating 5 variants costs no extra effort). It is a convenience multiplier, not a must-have. The moat comes from context — GrowthOS knows the contact, the campaign history, and the brand voice.

---

The Pain It Kills

Writing campaign copy is the most time-consuming part of launching a growth campaign. Most indie teams fall into one of two traps:

• Reuse the same templates forever — “Hey {name}, check out our new feature!” gets stale fast. Open rates decay as contacts tune out repetitive copy.
• Skip A/B testing entirely — testing 3 subject line variants means writing 3 subject lines. Most teams test zero.
• Hire a copywriter they can’t afford — freelance SaaS copywriters charge $100–$500 per email. At 4 campaigns/month, that is $400–$2,000/mo for copy alone.
• Use ChatGPT manually — copy-paste between ChatGPT and the email tool, losing context and consistency every time.

---

What It Does

1. Learn brand voice — analyze past campaigns, website copy, and tone preferences to build a brand voice profile per tenant.
2. Generate email subjects — 3–5 subject line variants per campaign, optimized for open rates.
3. Generate body copy — full email body with personalization tokens, matching brand voice and campaign intent.
4. Generate nudge text — short-form copy for in-app nudges, push notifications, and WhatsApp messages.
5. Auto-generate A/B variants — produce multiple copy variants for automated A/B testing with zero extra effort.
6. Human approval workflow — all generated copy goes through a review-and-approve step before sending. No fully autonomous sends.

---

Competition & What We Replace

Tool	Pricing	Limitation
Jasper	$49+/mo	General-purpose, no email platform integration, no brand voice from past campaigns
Copy.ai	$36+/mo	Generic templates, no contact-aware personalization
ChatGPT	$20/mo	Manual copy-paste workflow, no campaign context, no A/B variant generation
Braze AI Copywriter	$60K+/yr	Enterprise-only, locked behind Braze platform

GrowthOS auto-copy is context-aware — it knows the contact’s lifecycle stage, past interactions, and campaign history. Generic LLM tools generate copy in a vacuum.

---

Moat & Defensibility

Integration context is the moat (2/5).

• The LLM wrapper itself is low-moat — anyone can call GPT-4 or Claude. The value is in the context passed to the LLM: contact data, campaign history, brand voice profile, past open/click rates by copy style.
• Over time, the system learns which copy styles perform best for each tenant’s audience — a feedback loop that generic AI writing tools cannot replicate.
• Tight integration with A/B Testing closes the loop: generate variants, test them, learn what works, generate better variants next time.

---

Interoperability Advantage

The feedback loop between auto-generated copy, A/B testing, and analytics means the system gets smarter with every campaign sent. This is impossible with standalone AI writing tools.

---

What Ships

• Email subject generation — 3–5 variants per campaign with predicted open-rate ranking
• Body copy generation — full email body with personalization tokens and brand voice
• Nudge text generation — short-form copy for in-app nudges, push, and WhatsApp
• Brand voice learning — automatic tone/style extraction from past campaigns
• A/B variant auto-generation — generate N variants for testing with one click
• Human approval workflow — review, edit, approve before any generated copy sends

---

What Does NOT Ship

• Image generation — no AI-generated images, banners, or visual assets
• Landing page copy — generation is scoped to campaign messages (emails, nudges, push), not full pages
• Social media posts — no Twitter/LinkedIn/Instagram copy generation
• Fully autonomous sending — all generated copy requires human approval before delivery; no “set and forget” AI sends

---

Build vs Buy

BUILD (thin wrapper).

The core LLM capability is bought (OpenAI GPT-4 / Anthropic Claude API). What we build is the integration layer: brand voice extraction, contact-aware prompt engineering, A/B variant generation, approval workflow, and the feedback loop from analytics back to the model.

Estimated effort: 4–5 weeks.

---

Dependencies

Dependency	Why
LLM API (OpenAI / Anthropic)	Core text generation capability
Lifecycle Emails (P1-03)	Primary destination for generated copy
In-App Nudges (P2-14)	Destination for generated nudge text
A/B Testing (P3-21)	Variant testing and feedback loop
Past campaign data	Brand voice extraction and copy style learning