Beacon: AI-Driven Marketing Without the SaaS Tax

Marketing platforms are expensive. HubSpot, Marketo, and their competitors charge thousands per month for features that boil down to sending emails, scheduling social posts, and tracking which buttons people click. We built Beacon because we wanted all of that capability for the AccelaStudy(R) product line without handing a SaaS vendor a percentage of our revenue forever.

Beacon is an AI-driven marketing platform that handles content generation, campaign orchestration, multi-channel publishing, ad management, A/B experiments, and analytics. It runs on the same FastAPI and React stack as the rest of our internal tools, connects to our shared PostgreSQL and Valkey infrastructure, and exposes 69 MCP tools so our AI assistants can operate it directly.

The Dual LLM Strategy

Not every piece of marketing copy needs the same level of scrutiny. A social media post promoting a blog article has different quality requirements than a landing page headline that will be tested against three variants for the next two weeks. Beacon uses two different language models depending on the task at hand.

Mercury 2, from Inception Labs, handles bulk content generation. It produces blog drafts, social threads, email copy, and ad creative at high speed and low cost. When we need to repurpose a single blog post into a Twitter thread, a LinkedIn article, an email campaign, and three ad variations, Mercury 2 does the heavy lifting.

Claude Sonnet 4.6 handles the quality-critical work. Brand voice calibration, nuanced competitive positioning, and content that requires deep understanding of our product's technical differentiation all route to the stronger model. The cost difference is significant, but so is the quality gap on tasks that require genuine reasoning.

Key Features

Campaign Orchestration

Beacon supports multi-step campaigns with sequencing, configurable delays, and conditional branching. A campaign can start with an email, wait three days, send a follow-up to people who opened but did not click, then branch into different paths based on engagement. The lifecycle is fully managed: create, start, pause, resume, and track. APScheduler handles the background execution of scheduled social posts, campaign steps, report generation, and data aggregation.

Content Generation and Brand Voice

Every piece of content Beacon generates adheres to a brand voice profile. These profiles define tone descriptors, vocabulary to include or exclude, and example snippets that anchor the LLM's output to the desired style. The content repurposing engine takes a single piece of source content and produces variants for every channel with one click. All user inputs pass through a prompt injection filter before reaching the LLM.

Landing Page Builder with A/B Testing

Beacon includes a landing page builder with built-in variant testing and traffic allocation. Create a page, define two or three variants with different headlines or layouts, split traffic between them, and let the statistical significance calculator tell you when you have a winner. The experiment framework supports multivariate testing for subject lines, send times, content variations, and page layouts.

Multi-Channel Publishing

Social posts can be scheduled across Twitter, Meta, and LinkedIn with optimal timing and preview. Email campaigns run through AWS SES with step sequencing. Google Ads integration provides campaign management alongside AI-generated ad creative. SEO tools cover keyword research with competitive analysis, meta tag generation, heading structure audits, and keyword density analysis.

Analytics and Attribution

The marketing dashboard shows real-time metrics: active campaigns, content pipeline status, engagement trends, conversion rates, and budget utilization. Channel attribution supports four models (last touch, first touch, linear, and position-based). Conversion funnel tracking provides configurable stages from first touch to purchase with drop-off visualization. Custom reports can combine data from content, campaigns, social, ads, and SEO, with scheduled automated delivery.

Integration Points

Beacon connects to the broader Renkara infrastructure at several points. Authentication flows through the shared auth-service using RS256 JWTs. Social OAuth tokens are encrypted at rest with Fernet symmetric encryption. Webhook signature verification covers all four inbound providers: Stripe, SendGrid, Meta, and Google Ads. The MCP server exposes all 69 tools to AI assistants, covering content generation, campaigns, analytics, SEO, social, ads, segments, experiments, and landing pages.

Why Not HubSpot?

The typical marketing SaaS charges per contact, per seat, and per feature tier. For a company running multiple product lines across multiple domains, those costs compound fast. More importantly, SaaS platforms are black boxes. When the A/B testing framework does not support the specific experiment type you need, you file a feature request and wait. When the analytics dashboard does not break down data the way you think about your business, you export to a spreadsheet and do it yourself.

Beacon costs us compute and storage. We control every feature, every integration, and every data pipeline. When we needed landing page variant testing with conversion funnel visualization, we built it in a day. When we wanted AI-generated ad creative that understood our product's technical positioning, we pointed it at our brand voice profiles and it worked on the first try.

Specs