Book a walkthrough
Nonte’s AI position

An exciting data strategy in the age of AI.

Most AI products in research focus on the model. We focus on the substrate underneath. The configuration that researchers describe in plain language flows into the same structured store as the data the platform captures, and out the other side comes data that AI tools can actually defend, line by line.

100% provenance turns research into a 100% defensible asset, including when an AI agent is the thing doing the reading.
CONFIGURATION IN THE NONTE SUBSTRATE SUBSTRATE OUT " Natural-language rules "Flag specimens past stability window" Custom forms & fields Described in plain language, versioned Alert routing & thresholds Per study, cohort, site, or sponsor Cohort-specific rules IBD, dermatology, oncology, anything NONTE Provenance Store Structured Versioned Machine-readable All changes audited Field-level lineage Before/after, user, time, study context Versioned protocols AI claims tied to rules live at the time Identifiers that travel Stable, join-friendly across systems Machine-readable custody Structured events agents can replay
Configuration in through plain language. Substrate out ready for AI interpretation. Both governed by the same audited provenance store.

Nonte’s AI position · Part 1: Configuration in

Configured by conversation, not configuration trees.

QA Companion is built to be set up the way a researcher would describe it. Tell the system what matters in your study and it builds the rules. No tickets to IT. No three-week scoping engagement. No buried admin tree to learn.

You

Flag any specimen that has not moved in fourteen days and is within thirty days of its stability window.

QA Companion

Rule created. I will mark these as amber on the dashboard, exclude specimens already flagged for disposal, and route a daily digest to the programme lead. Sound right?

You

Yes. Add the same rule for the dermatology cohort with a seven day threshold.

QA Companion

Done. Two rules live, scoped per study. Version logged.

What you can configure this way

  • Signal thresholds per study, cohort or site. Stability windows, custody gaps, dwell times.
  • Alert routing tied to programme leads, QA officers, or sponsors. Daily digest, instant, or on-trigger.
  • Custom forms and fields described in plain language. Versioned automatically.
  • Cohort-specific rules for IBD, dermatology, oncology, or anything you're running.

Every configuration change is itself audited. The rule, the user, the time, the rationale: all of it lands in the same provenance store as the data the rule governs.

Nonte’s AI position · Part 2: Substrate out

Data that AI can actually trust.

AI agents are increasingly used to interpret research data: literature synthesis, hypothesis generation, regulatory drafting, decision support. They are only as defensible as the data they reason over. Nonte produces that data: structured, versioned, provenance-stamped, machine-interpretable by design.
01

Every field, every change, every actor.

Field-level lineage with before and after values, the user, the timestamp, and the study context. An AI agent asking “why did this volume change?” gets a structured, queryable answer.

02

Versioned forms and protocols.

When a protocol evolves mid-study, the old version is still queryable against the data captured under it. AI interpretations stay tied to the rules that were live at the time.
03

Identifiers that travel.

Specimen IDs, subject IDs, tray locations, study codes: stable, persistent, and join-friendly. AI agents working across your data plus external systems can ground every claim to an object.

04

Machine-readable chain of custody.

Every specimen move, every export, every reactivation. Not as free text in a log file, as structured events your agents can replay and audit.

In the age of AI, the moat is not the model. It is the substrate underneath. The labs that win the next decade will be the ones whose data their AI tools can defend, line by line.