Service Marketing Automation
The content pipeline that ships our daily blog.
We run an automated pipeline that publishes content daily across our properties. We will build yours, with research, drafting, fact-checking, human review, and publishing wired end to end.
Projects are scope-dependent. Free discovery call.
pipeline.example.com/runs
pipeline/publish.ts ts
// pipeline/publish.ts
import { createWorkflow } from '@/runtime';
export const dailyPublishPipeline = createWorkflow('daily-publish', async (ctx) => {
const ideas = await ctx.run(pullFromIdeaQueue, { limit: 3 });
for (const idea of ideas) {
const research = await ctx.run(researchTopic, idea);
const draft = await ctx.run(generateDraft, { idea, research });
const factChecked = await ctx.run(factCheck, draft);
const reviewed = await ctx.run(humanReview, factChecked, { timeout: '4h' });
if (reviewed.approved) {
await ctx.run(publishToWordPress, reviewed.content);
await ctx.run(notifySlack, '#content-published', reviewed.content);
}
}
});
Why this matters
Most "AI content pipelines" produce content no one would publish.
The first version is exciting. The tenth is generic. By the hundredth, your blog reads like every other AI-generated blog. We build pipelines that keep your editorial voice, ground every claim in your real sources, and put humans in the loop where judgment matters. The result is content you actually want to publish, at a cadence your team could not hit by hand.
In flight today
Six stages. Articles moving through.
The pipeline view our editors check first thing each morning. Every article passes through draft, edit, image, audit, ready, then live. Each stage has SLA. Watch one article promote forward when you scroll into the section.
Headless WP, year 2: what we changed
v0.1 · 920 wWhy we switched from JetEngine to ACF
v1.2 · 2,140 w moving →Cloudflare Workers Assets at scale
v1.0 · 2,560 wLMS migration from Teachable to LearnDash
og + 4 inlineBuddyPress activity throttling patterns
SEO 92 · A11y 98Setting up Stripe Connect for marketplaces
SEO 88 · A11y 100Astro 6 + WP REST: what to know
scheduled · 09:00 ISTModeration queue patterns we ship
scheduled · MonHow we rebuilt our store on Astro
3.4k views · 18 refPre-release smoke testing for WP plugins
1.8k views · 6 ref- House style baked into prompts · same voice across every article
- Fact-check step uses retrieval over your real sources
- Human gate at audit · publishing never bypasses a person
What we build
A pipeline your editors will actually use.
Research grounded in your real sources. Drafts in your house style. Fact-check before review. Human approval at the right step. Publishing wired to your CMS. Analytics closing the loop.
01
Research that uses your real sources
The pipeline pulls from your existing knowledge base, brand voice docs, customer research, and approved external sources. The AI does not invent context, it composes from the context you already have.
Drafts feel on-brand from version one.
02
Drafts in your house style
Style guide encoded as a Claude skill. Tone, sentence rhythm, banned phrases, paragraph structure all enforced. Editors edit voice nuances, not basic style.
Editor revision time drops 60 to 80 percent.
03
Fact-check before human review
Every claim cross-checked against the source documents. Hallucinated stats flagged before they hit your editor. Sources cited inline so verification is one click.
Fact-check failures caught before publish, not after.
04
Human approval where it matters
No content publishes without an editor approval step. The pipeline does not replace editorial judgment, it just removes the time you spend on draft generation and formatting.
Editorial team retains full control of voice.
05
WordPress, Ghost, MDX, all supported
Publishing adapter ships content to wherever you publish. WordPress via REST, Ghost via API, static MDX via PR. Featured images, internal links, schema, all wired.
No manual copy-paste between tools.
06
Analytics loop closes the cycle
Published post performance tracked in GA4 and Search Console. Underperformers flagged. Best-performing patterns reused in future drafts. The pipeline gets smarter monthly.
Top-quartile post rate climbs measurably each quarter.
$0.50-2
typical AI cost per published post across our pipelines, fully accounted
Includes research, draft, fact-check, image generation, and metadata.
Proof we run this ourselves
The pipeline ships our blog every day.
We use this exact pipeline to publish on wbcomdesigns.com, vapvarun.com, attowp.com, and three other properties. Every workflow we ship to a customer is one we already run for ourselves.
pipeline.example.com/runs
Daily publish pipeline runs
Process
How a content pipeline project runs.
01
Discovery
Two weeks. We capture your style guide as a Claude skill, ingest your source documents, design the workflow stages, and lock the editorial approval rules.
Fixed scope, fixed price.
02
Build
Four to six weeks. Pipeline ships in stages: research first, then draft, then fact-check, then publish. Editors review real output by week three. Iterate weekly.
First real published post by week six.
03
Launch + tune
Two weeks. Production rollout, weekly tuning sessions with your editorial lead, analytics dashboards live. Handoff includes the maintenance runbook.
Your team owns the pipeline at the end.
Common questions
Frequently asked
-
Will this replace our content team?
No, and we will warn you if you ask us to. The pipeline replaces the work editors should not be doing (formatting, image processing, internal link insertion, schema, scheduling). Editors keep doing what only editors can do (voice, judgment, the part the AI cannot fake). Done right, your team ships more without burning out.
-
How long does a content pipeline take to build?
Six to ten weeks for a full pipeline (idea queue, research, draft, fact-check, human review, publish, analytics). Two weeks discovery to capture your style guide and source documents, four to six weeks build, two weeks launch and tuning.
-
Can it ship to multiple sites?
Yes. We have built pipelines that ship to 5+ properties with per-site style guides, per-site source documents, and per-site approval rules. The pipeline scales horizontally without rebuilding the workflow.
-
What about SEO?
SEO is built into the pipeline. Keyword research informs the brief. Internal linking is mandatory and validated against your sitemap (no hallucinated URLs). Schema markup is generated. Featured images include alt text. Yoast or Rank Math fields are populated.
-
How much do AI costs run?
Production cost typically lands at $0.50 to $2 per published post, all in. Prompt caching, model routing, and source document caching keep it well below what manual generation through ChatGPT would cost.
-
What does it cost?
Content pipeline projects are scope-dependent for a single-site pipeline with 4 to 6 stages. Multi-site pipelines with custom analytics and full editorial integration are scoped after discovery. Discovery call is free.
Ready to ship more without breaking your editorial standards?
Tell us what you want to build.
Discovery call is free. Fixed-price quote within 48 hours. NDA on request.