Musica.com.pt — Product, Growth & AI

📣 Context

Portugal has world-class music, festivals and artists, but no modern, opinionated digital destination covering them with the depth and craft international outlets bring to their scenes. Musica.com.pt was built to fill that gap: an independent newspaper about Portuguese and international music, with a journalist's voice, a designer's eye and a product manager's attention to growth and data.

The constraint shaped everything: a small team, and one year to build both an audience and authority.

🏆 Outcomes

🥋 My Role & the Team

A Digital Streetz project that I direct as Director of Brand & Product. I lead the product, the brand, the design system and the editorial direction, and I directed the design and build of Redação, the newsroom's back office and its AI.

🧭 Key Decisions

• SEO and performance as constraints from day one. Not retrofitted later: the information architecture was defined to capture long-tail search across artists, festivals, genres and dates, and every page type carries its own structured data (NewsArticle, MusicGroup, MusicVenue, events).
• Content as a connected system. Artists, venues and events are content types of their own and link to the articles, so an artist page gathers everything written about them and a venue page lists the events held there. Every content type (review, interview, festival guide, news) has its own template.
• Bespoke over off-the-shelf. The design system was built in PenPot and translated directly into a custom WordPress theme with five plugins made for it (ads, concerts agenda, newsletter, related articles, search), with no page builder.
• AI structures, people write. The newsroom's back office has AI in every step, but never to write the article. That line shaped the whole tool.

🤖 Redação: the AI That Does Not Write

Behind the site sits Redação, a private back office the newsroom uses every day. It gathers press releases from email, pages sent from the browser, feeds and notes in a single inbox, understands embargoes and scheduling, and creates the draft or the scheduled post in WordPress, with SEO and promotion in the same place.

AI is built into each step, but only where it helps:

• It structures. It reads the press release and fills the fields (headline, summary, category, tags, and the embargo only if the text says so), with category and tags forced to values that already exist in WordPress.
• It assists on request. Translation, the Instagram caption, hashtags and the newsletter excerpt are explicit requests, never automatic. A separate service suggests SEO terms and a sample paragraph.
• It runs where it should. Each analysis runs in the cloud or on a local model, chosen item by item. Images are cropped to the site's real sizes and upscaled on the local graphics card (Real-ESRGAN and ATD models), so photos never leave the machine.
• It stops at the skeleton. The draft that reaches WordPress is a summary, the facts, the sources and an embargo notice, never text ready to publish. Everything goes through human review and writing, as the newspaper's AI use policy requires.

With everything in one place, an article now takes under an hour instead of several. Read the full Redação case study →

🖼️ The Product

Typography, colour and hierarchy are tuned for reading at editorial depth: long-form articles, rich media and artist profiles sit alongside short news, from phone to desktop. Two themes, light and dark, recall paper and ink, with red as the only accent. Around the core sit the concerts agenda, an Academy of practical guides for artists and promoters, and Viseu ao Centro, a series about the local scene.

Growth experiments covered structured data, internal linking, image optimisation, social metadata, newsletter capture, and editorial partnerships with record labels and festival press offices.

🧠 Takeaways

• Decide where AI stops before deciding what it does. Drawing the line at "structure, never write" keeps the newspaper's voice human and makes the tool easy to trust.
• SEO is architecture, not a plugin. The growth was built on content types, templates and structured data planned from day one.
• Small teams need sharper choices. Unlike the Kindred work, this is full ownership with a small team: every hour goes to what moves readers or saves the newsroom time.