
What happened
SOMPO Digital Lab used Claude Code as a writing agent and shortened a technical blog post from 4–6 hours of work to 30 minutes–1 hour, per its August 2025 post.
Why it matters
The bottleneck shifts from writing to design; SOMPO defined the agent's role and style in a CLAUDE.md file, and GENAI reported cutting a post from 8 hours to 1 hour, about 528 hours saved monthly.
What to watch
Whether teams keep a human gate; the guide recommends agents draft and save to WordPress, but says humans must verify company names, figures, sources, duplication, and check for spam-policy risk.
WHO IT HITSContent marketers and editorial teams at BtoB companies running owned media are the main audience, since the article shows how agencies and PR functions can move headcount from drafting to topic selection and review.
Ask the AI about this article →
Summaries like this, in your inbox every morning.
The article frames the shift as one of scope rather than writing quality. Most teams hand generative AI only the drafting step, so structure fixes and upload work stay with staff, and time savings are small. An agent reworks that flow: it breaks a goal into tasks, operates external tools like web search and the WordPress REST API, and evaluates its own output before submission.
The case studies illustrate the same structure. SOMPO Digital Lab built an agent around Claude Code and defined its role and writing style in a CLAUDE.md file, handling planning, title ideas, drafting, preview, and posting as one chain. Genie applied the same pattern to press releases, pulling in interviews from development and sales teams. GENAI broke the process into keyword research, competitive analysis, structure, drafting, and SEO settings, and reported that 480 minutes became 60.
The stakes likely hinge on where the human gate stays. The article warns that without design, wrong facts, missing sources, and spam-policy violations can damage existing articles, and it recommends limiting agent permissions to draft saving only. The outcome for each team will probably depend less on which tool it picks than on whether it documents its rules and reserves a final check.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
TheCUBE Research principal analyst Scott Hebner says Certinia is moving its Veda platform beyond agentic orche…
Emergence AI ran eight simulations using frontier models including Claude, Gemini, Qwen, DeepSeek and Mistral

Google Deepmind released Gemini 3.8 Live and 3.8 Live Extended Thinking for developers via the Gemini API and…

In a WIRED interview, Tilly Norwood — the AI-generated 'actor' made by Xicoia, a division of Particle 6 Group…

Meta on Tuesday introduced Meta One, a subscription service offering expanded AI usage and premium features ac…

Ryan Greenblatt of Redwood Research launched the AI Contact Hotline, letting agents with limited internet acce…
