← Insights

Five products, one way to buy

Why we build narrow agents for specific industries, sell them through Google Cloud Marketplace, and publish one set of policies for all of them.

20 October 2026 · Carnegie Intelligent Technology · 2 min read

Carnegie Intelligent Technology builds AI agents for industries that run on documents: finance close, freight disputes, equipment after-sales, industrial tenders, sales practice. Each product does one job end to end, cites every source and leaves the decision with a person. This note explains the three choices behind that.

Narrow, not general

A general assistant that can discuss anything is useful to individuals and hard to buy for a department. A product that reconciles intercompany balances, or disputes demurrage invoices, is a line in a budget with an owner. We build the second kind. Each product has its own site, its own case library describing exactly what it handles, and its own pricing in the unit of work that industry already counts in.

Marketplace, not a new vendor relationship

Our customers have Google Cloud invoices and procurement processes already. Listing on Google Cloud Marketplace means the charge lands on that invoice, counts toward committed spend, and goes through a vendor they have already approved. For us it means one billing integration, built once for Sparring and reused for every product since.

One legal framework

Privacy policy, terms, data processing addendum, acceptable use and the sub-processor list are published once, here, and apply to every product. A security questionnaire answered for one product is answered for all of them.

What early access means

Four of the five products are in early access: being built with design partners using their real documents. Their sites describe designed behaviour and say so. We do not publish customer names, testimonials or usage numbers we do not have.

Related

Talk to us.

Sales, private offers, security questionnaires, partnerships.