One taxonomy. Every domain. A question in plain language.

Guardiant’s proprietary taxonomy is one model of how a risk-bearing organization’s data fits together — members, providers, claims, encounters, authorizations, care, pharmacy, the federal return files, public reference data, and the rules that govern all of it.

Your sources map onto it once: the administrator’s claims and eligibility, the credentialing and authorization systems, the care-management platform, each payer’s files if you hold several. From then on, “active member” means one thing across every domain. That sentence sounds small and is most of the value.

We do the mapping, we integrate the pipeline with your systems, and we take ownership of operating it reliably.

On top of the taxonomy sits the encoded domain: the regulations, the specifications, the calculation rules, the traps that make two accurate reports disagree. This is what a vendor sells you access to. Here, it is authored once against the taxonomy, versioned, self-tested, and applied to your data — and it is the part that compounds every time a rule changes, because it changes once for everyone.

Dashboards for the numbers people need every week: fixed, fast, deterministic, precomputed. The figure does not move because someone phrased a question differently.

Conversation for everything else. A question in plain language becomes real analysis — queries against your warehouse, a working spreadsheet, a map, a chart, a written answer with its denominators stated. English or Spanish. Every panel on every dashboard hands its context to the conversation, so “why is that number like that” is one click, not a ticket.

Most analytics ends at a screen, and most of the actual work happens in a spreadsheet emailed to a physician group or a compliance file handed to an auditor. Alberto produces the spreadsheet, the print-ready report, the submission-format table and the interactive map as first-class outputs — evidence attached.

Some of the platform answers questions; some of it does jobs. The voice line is the first agent: it holds a conversation, looks up the record, speaks the answer and knows when to transfer. Agents run on the same taxonomy and the same rule — the model handles the interaction, never the fact — and they are priced as what they replace.

Production is a single-tenant deployment in your cloud boundary: an application layer, your data warehouse, and no copy of your data anywhere else. A frontier model chosen for truthfulness and quality, under an enterprise agreement and a BAA. Bilingual interface. Role-based access with prompts and permissions scoped by job.