Publications
Featured on Vercel: 100,000 media pitches a month, three engineers and results that need clarification
Featured processes more than 100,000 media pitches a month, according to Vercel, with three engineers. For a Quebec SMB, the value of the case lies in simplifying AI architecture—not in proof of financial returns or media results.

More than 100,000 media pitches a month, 17 AI models and three engineers: Featured’s numbers attract attention. But for a Quebec SMB, the useful question lies elsewhere: which functions can a small team entrust to a managed platform, and how can it measure whether that choice is worthwhile?
The case study published by Vercel describes a gradual consolidation of deployments, model access and background processing. However, it is a commercial account from the vendor, not an independent audit. The figures below are those reported by Vercel.
A migration that began with 374 sites
Featured is presented as a public relations copilot for experts and communications teams. Its agents identify opportunities among journalist requests, podcasts and awards, among other sources.
According to Vercel’s account, the initial environment relied on AWS Elastic Beanstalk, custom integrations with AI providers and a separate system for orchestrating long-running tasks.
The trigger was not the choice of a new AI model, but the deployment of 374 Sanity sites. The manual process had become cumbersome; Vercel describes launching the sites on its platform as a “one-click” deployment. No migration time or hours saved are provided.
The team then expanded the migration to the application and agents. The case study does not establish that Featured abandoned all AWS services or replaced Sanity.
Three functions brought together
The architecture described by Vercel brings together three functions that were previously more dispersed:
- Deployments, moved to Vercel after the experience with the Sanity sites.
- Model access, standardized through AI SDK and AI Gateway. Featured reportedly uses 17 models, though their names and the distribution of requests are not specified.
- Long-running processing, moved to Workflow SDK after the AI layer migration: media monitoring, opportunity qualification and combining signals.
The article also states that chat is becoming the main interface, with eve, described as an open-source agent framework from Vercel. Its exact place in the migration timeline is not established in the available excerpts.
The potential lesson is a reduction in the number of components that need to be operated separately. The case suggests that a managed platform can help a small team focus its efforts on the product. However, it does not quantify this reduction in workload.
What the volumes prove—and do not prove
| Metric reported by Vercel | Limitation to keep in mind | |---|---| | Three engineers | This is not Featured’s total headcount. | | Three brands and more than 100,000 users | Active and paying users are not distinguished. | | More than 100,000 media pitches a month | This volume does not correspond to the number of media placements secured. | | One connection every six seconds | The definition and calculation method are not specified. | | More than 100 million HARO emails over the previous year | These are emails, not media results; the exact period is not established. |
These metrics are not interchangeable. They cannot be used to calculate either a conversion rate into media coverage or productivity per engineer. The case study also provides no costs before and after migration, hours saved or ongoing incidents.
Two perspectives for a Quebec SMB
If you are developing an AI product
The approach to evaluate is consolidation, not replicating the use of 17 models. A common interface could make testing and switching model providers easier. On the other hand, bringing deployments, AI access and orchestration together could increase dependence on the platform.
A comparison should therefore consider total cost, operational effort and ease of exit. The case study does not demonstrate that Vercel is less expensive or better suited to every workload.
If you are seeking media visibility
Featured could be an option to test for identifying opportunities and preparing pitches. However, its actual coverage of Quebec and French-language media remains to be verified.
For a trial, the recommended metrics would be the relevance of opportunities, responses received, media placements secured and associated costs—rather than the number of pitches sent alone. Human validation of expertise and messaging would also be a precaution to plan for; the documents provided do not confirm that Featured does this.
Before deciding: verify what is missing
A limited trial should help answer four questions:
- Costs: how much do the platform, models, emails and operations cost?
- Quality: what useful results are achieved beyond the volume processed?
- Controls: how are duplicates, inaccurate messages and irrelevant outreach prevented?
- Data and compliance: which models receive which data, where is it processed, how long is it retained and which requirements apply to the project?
The Featured case offers a credible starting point for considering technical simplification. It is not proof of financial returns or media results. For an SMB, the reasonable next step would be a narrowly scoped trial, with cost, quality and compliance criteria established before launch.