// PRICING
What this costs.
Most firms in this category will not publish a number. We think that is the tell. Here is the actual model: a one-year retainer, priced by how many projects are running at once, with AI spend passed through at cost.
One active project
$5,000
per month, one-year retainer
One project running at a time. We scope it, build it, ship it to production on your infrastructure, then move to the next one.
Most companies start here. It is enough to get one real system running without the scope sprawling.
Two to three active projects
up to $10,000
per month, one-year retainer
Multiple workstreams in parallel, typically once the data foundation is in place and the backlog is obvious.
Companies with several siloed systems that all need to reach the same warehouse.
Why a one-year retainer?
Because this work does not end at a launch. Once a system is in production it gets better in the months after, and the next project is always visible from the last one. A one-year term matches how the work actually behaves.
What is not included?
AI and token spend. That is billed at cost, not marked up. We are explicit about this because it is real money and it moves: we have watched an agentic job run away and put $10,000 on a bill in two days. Every pipeline we build now caches before it spends and stops itself at a cap.
Why price by active projects instead of hours?
Because the constraint is not our time, it is your focus. Once you see what the technology can actually do, the temptation is to start everything at once. Limiting how many projects run in parallel is the mechanism that keeps an engagement from turning into a list of half-finished systems.
What determines where you land in the range?
How many systems have usable APIs, how much of the data needs a workaround to reach a warehouse, and how many projects you want running at once. We tell you which of these apply after the first conversation, before you commit to anything.
The first conversation is free, and it is often a no.
One hour, no pitch deck. We will tell you whether this is the right fit, and we have told clients not to build AI at all until their data was trustworthy. If you want to know where you stand before talking to anyone, the assessment takes two minutes and asks for no email.
Whatever we build runs on your cloud, in your accounts. See how the work actually happens.