Brief-to-draft agent
Feed a brief, get a publication-ready draft in your brand voice. The agent applies your style guide, keyword targets, and structure template. First draft in under 20 minutes.
Service · ContentOps
ContentOps Autonomy. Editorial teams that ship 3x faster.
AI agents turn a brief into a draft, a person checks every draft, and it all connects to your CMS. Built for marketing agencies, news media, and in-house editorial teams.
Free 30-minute intro call · Fixed price · You own the agent code
Sound familiar?
Your editorial team is not slow. The process is.
The brief-to-published cycle takes 3-5 days for a single article. Writers spend Tuesday morning in Notion, Tuesday afternoon in Google Docs, and Wednesday morning in Slack chasing feedback. The actual writing is 2 hours of a 3-day loop.
QA gates are manual: a senior editor reads every draft for tone, fact accuracy, SEO compliance, and brand voice. At 8-12 articles per week with 3 FTE, that editor spends 40% of their time as a bottleneck, not a strategist. Add multi-locale requirements and the workload multiplies by the number of languages.
The result: an editorial team producing 8-12 articles per week with 3 FTE, when the same team with the right process delivers 30-40 articles per week. The constraint is not talent. It is the tools.
The Reframe
AI agents that draft. Humans that decide. Your CMS that ships.
Checked by a person
Editor approves at two checkpoints: outline and final draft. Every stage is editable. The human decides what ships. No rogue AI content reaches readers.
CMS integration
Publish to Payload, WordPress, Webflow, or any CMS with an API in one click. Draft appears in your CMS, complete with metadata, slug, and featured image slot.
We use Claude Sonnet for drafting and brand-voice fine-tuning. Your data stays in EU. No third-party AI vendor sees your unpublished content.
How it works
From kickoff to steady state in 4 weeks.
- 01 Week 1
Brand voice discovery and CMS review
We document your brand voice rubric, map your existing CMS setup, and scope the agent precisely to your content types and approval chain.
- 02 Week 2
Build the drafting agent and review screen
The agent is built and tested on 10 real briefs from your archive. The review screen goes into your editor's browser. You approve or reject with tracked rationale.
- 03 Week 3
CMS integration and multi-language publishing
We wire the agent output to your CMS. If you publish in multiple languages, each language version is translated and adapted, not translated literally.
- 04 Week 4
Editor training, handover, and 30-day support
Your editorial team runs a full production week on the new process. We monitor output quality, fix edge cases, and hand over written operating instructions. Thirty days of async support included.
Pricing
Transparent pricing. Agreed before we start.
- Getting started
- Free intro call · paid discovery from €900 only if the scope is unclear
- One-time setup
- €2.000 - €5.000 (CMS complexity and locales)
- Monthly retainer
- €4.000 - €8.000 per month (article volume and agent updates)
- Included
- Agent code ownership · brand-voice fine-tuning · monthly performance review · operating instructions
- AI API costs
- Passed through at cost, no markup
Exact scope is set on the intro call. You get a fixed number before you commit to anything.
Live since 2023 Read full case study →
1.4 million legal documents. Answers with numbered sources in 40 to 60 seconds.
Tax-Fin-Lex · Slovenia
Tax-Fin-Lex has published Slovenia's legal and tax information since 2004. Lawyers, accountants and tax advisers use its database to look up laws, court decisions and professional articles. Finding the answer in a long court decision takes time.
TFL AI answers a legal question in plain language and links every claim to numbered sources the reader can open. The first feature, court-decision analysis, went live in December 2023. The question-and-answer chat followed in April 2024. GALOR's founder has worked on the engineering of TFL AI since 2023, together with Tax-Fin-Lex's own team.
Today TFL AI has more than 1,800 users and draws on more than 1.4 million documents, including 280,734 court decisions. In June 2024 the Tax-Fin-Lex team received a silver innovation award for it from the GZS Central Slovenia chamber of commerce.
FAQ
Common questions
-
No. Editors approve every piece before it publishes. The agent handles first-pass writing so your editors can focus on strategy, judgment calls, and quality direction rather than mechanical drafting.
-
Payload, WordPress, Webflow, and any CMS with a REST or GraphQL API. We confirm compatibility before we quote. Custom integrations are part of the setup scope.
-
We build a brand voice rubric from your best-performing content and run it as a system prompt constraint. Editors score drafts at the outline stage, and those scores feed back into the fine-tuning over time.
-
SEO for Google and for AI search answers is built into the agent output. Keyword placement, heading structure, structured data for search engines, and locale-specific search intent are all part of the first draft, not a separate checklist.
-
Yes. Each language version is translated and adapted, applying local idiom, search intent, and cultural context. It is not a literal translation of the EN master.
-
EU-only. Hetzner infrastructure in Germany and Finland. No US data transfers. No third-party AI vendor receives your unpublished drafts.
-
The agent code belongs to you. The CMS integration is a thin adapter layer. Porting to a new CMS is a scoped week of work, not a rebuild.
-
Yes. A short paid discovery (from €900) runs the agent on 5 real briefs from your backlog and delivers a scoped recommendation. You see the output quality before committing to setup.
Tell us what needs building
Fixed price. No lock-in. EU hosting. Code you own.