// PRICING
What this costs.
An embedded AI engagement starts at $12,500 for a two-week Production Readiness Diagnostic, applied in full as a credit against a build. Build engagements run $18,000 to $24,000 per month on a six-month initial term, then month to month. Enterprise starts at $35,000 per month. AI and token spend is billed at cost, capped.
Most firms in this category will not publish a number. We think that is the tell.
Production Readiness Diagnostic
$12,500
two weeks, fixed scope
A data-trust audit across your named systems, an API and integration feasibility map with the cost of each workaround, a ranked build sequence with cost bands, and a written recommendation, including a recommendation not to build if that is the honest answer.
The way to start. Applied in full as a credit against your first invoice on a build engagement begun within 90 days.
Embedded Build
Most engagements$18,000 to $24,000
per month, six-month initial term, then month to month
One embedded operator building and operating production systems on your cloud. Includes AI spend governance: caching before spend, hard caps, per-batch yield monitoring, and cost attribution per pipeline.
Mid-market companies with operational data spread across several systems. Two active workstreams included; a third by written change order.
Enterprise
$35,000+
per month, six-month minimum
The same work under enterprise conditions: security review, compliance evidence, change control, and coordination with an internal IT function.
Organizations where the governance overhead is the real cost driver, not the engineering. Priced explicitly rather than absorbed.
Pricing questions
Why six months and then month to month?
Because a long contract is a bad way to earn a long relationship. Most professional-services work runs under six months, and the firms charging more than us ask for less commitment than we used to. Six months is not how long a build takes; the build phase runs 8 to 16 weeks. Six months is the build plus a full operating quarter, because the systems that fail are the ones nobody ran in anger before the person who wrote them left. After that you stay because the work is worth it, on thirty days' notice. Annual prepay takes 10% off if you prefer it.
What is not included?
AI and token spend, billed at cost with no markup, under a monthly not-to-exceed cap you agree in advance. Anything above the cap needs your written approval first. We are explicit about this because it is real money that moves. On one of our own systems the cost sentinel fired in anger, auto-pausing 72 scheduled jobs when daily spend hit $130 against a $120 cap, and it was right to. That is why every pipeline we build caches before it spends, monitors yield per batch, and stops itself at a limit rather than sending you a surprise.
Why not price on hours?
Because you are not buying hours, you are buying a system that runs. What the monthly figure buys is a defined scope, accepted in writing, and a limit on how much is in flight at once. We will not have more than two things half-finished in your environment, which is a guarantee about delivery rather than a restriction on your ambition.
What determines where you land in the range?
How many of your systems have usable APIs and how many need a workaround built, how much reconciliation the data needs before anyone can trust it, and how much governance overhead your environment carries. The diagnostic answers all three before you commit to a build, which is most of why it exists.
Is this cheaper than the alternatives?
Not always, and we are not going to pretend otherwise. Six months here is $108,000 to $144,000, which lands inside the band a consultancy would quote for the same build, not obviously under it. What differs is what you hold at the end: the systems run on your infrastructure with your name on them, there is no platform to keep licensing, and the person who scoped the work is the person who builds and operates it. We are not the cheap option. We are the one where nobody hands off.
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.