platform / the stack
Built on Vercel primitives. All of it.
Incumbent MAPs carry twenty years of datacenter with them, and you can feel it — in the queue times, in the maintenance windows, in the UI. veloquent is what a MAP looks like when the infrastructure is someone else's core competency.
The parts
Sends, syncs, and journey steps each run as their own tracked job with a queue in front of them. Translated: if a step fails it retries by itself rather than stranding someone halfway through a nurture; if the servers restart mid-journey, everyone's place is remembered; and a journey can wait six months without anything having to stay awake to watch it. Scheduled work runs on a clock we don't maintain, media lives in managed storage, and every AI call goes through one gateway. Capacity follows demand, so a Tuesday-morning send doesn't queue behind someone else's.
A live evergreen journey: waits, event branches, a step that writes copy per person, and a 50/50 test. Each step is its own job, so a week-long wait is still a week and a failed step retries instead of dropping the person.
Why an operator should care
Because infrastructure is where MAP pain hides, and it always shows up as your problem: the campaign queue that backs up at 9am, the smart list that times out at a few million records, the maintenance window the week of your launch, the sync that broke on Friday. Renting infrastructure that already handles far bigger traffic than a marketing database means we're not solving those from scratch — and you're not the one paging someone about them.
Workflows can do anything
In most platforms a journey step can only pick from a fixed menu: send, wait, change a field. Here a step is a piece of code we can extend, so a flow can write a personalized line mid-journey, ask your warehouse a question before deciding where someone goes, or call another system and use the answer. That's also what makes real A/B testing — split the flow, not just the subject line — a roadmap item rather than a rewrite.
Questions operators ask
- What infrastructure does veloquent run on?
- Entirely on Vercel primitives: Fluid Compute for scale-to-demand compute, durable Workflows and Queues for sends and journeys, Crons for scheduled work, Blob for media, and AI Gateway for model routing.
- How does veloquent handle scale?
- By inheriting it. veloquent runs on infrastructure built for web-scale traffic, and its own data access is designed for large workspaces: the campaign tree loads one level at a time instead of everything at once, moving a folder updates one record rather than rewriting everything beneath it, and each journey step is tracked individually so long flows don't accumulate cost.
- Can veloquent workflows include AI-generated content?
- Yes — workflow steps are real compute, so a journey can generate or personalize content through AI Gateway as part of the flow, subject to the same review gate as everything else.
Follow the build.
One email when something ships. Or better: bring us the thing your MAP won't fix.