CrunchCode.

Your AI budget should have a CTO.

If you pay for AI somewhere — a forecasting pilot, an LLM bill, a vendor's "AI feature" — can you say what each one costs, what it returns, and what breaks if it stops? Our AI CTO answers those three questions first, then puts the ones that pay off into production.

Book an AI auditBook 30 minutes with our CTO

Where AI money goes missing

01The bill nobody owns.
LLM calls from staging environments, vendor "AI features" that turn out to be rules underneath, staging traffic nobody switched off. At one mid-market 3PL the LLM bill was $42,500 a month; $14,000 of it was switched off in the first week (C10).
02Models that live in a notebook.
A forecast changes nothing until it reaches the screen where the planner works. At a distributor with 82,000 SKUs the forecasts now arrive in the ERP as draft purchase orders, approved in one click (C12).
03Models that ignore how demand behaves.
Forty percent of that distributor's catalog had intermittent demand, where standard forecasting fails — so it got a separate model, and supplier lead times went into safety stock (C12).

What the AI audit does

Two weeks, fixed scope, $7,500 — our diagnostic, pointed at AI.

01An inventory of every model, AI vendor and LLM bill: cost per month, owner, return.
02A verdict on each one: scale it, fix it, or switch it off.
03One use case taken to a production plan — deployment, monitoring, a named business owner.
04The rule we leave behind: no AI initiative starts without success metrics, an ROI estimate and a named owner.
What we ask you for

your AI vendor contracts and invoices for the last 90 days · LLM and API billing exports · a list of models and pilots with their owners · read access to the data each model uses.

The findings document is yours. If something is worth building, the team that builds it already knows EDI, TMS and ERP data.

What we build

Demand forecasting and replenishment · carrier-invoice classification · computer vision for yards and terminals · AI deployment, monitoring and cost control.

What changed

C10 · US mid-market 3PL · 3 months

LLM spend $42,500 → $16,000 a month (−62%). First use case, carrier-invoice classification, in production nine weeks after the audit.

C12 · US B2B distributor · 82,000 SKUs · 7 months

$2.3M in working capital released in the first year. Stockouts 12% → about 4%. Planners' time on data 70% → about 15%.

C11 · US container terminal · 139 cameras · 12 months

340+ documentation discrepancies a month caught before customs clearance.

When you don't need us

If you want a chatbot on your website, a vendor will sell you one faster. If your data is not yet in one place, the first project is data, not AI — and we will say so in the first week.

Start with the audit.

Book an AI auditSend us a requirement
C10C10 · US mid-market 3PL · 3 months
LLM spend $42,500 → $16,000 a month (−62%). First use case, carrier-invoice classification, in production nine weeks after the audit.
C12C12 · US B2B distributor · 82,000 SKUs · 7 months
$2.3M in working capital released in the first year. Stockouts 12% → about 4%. Planners' time on data 70% → about 15%.
C11C11 · US container terminal · 139 cameras · 12 months
340+ documentation discrepancies a month caught before customs clearance.