the second product

FieldAssist —
your data, in a chat box.

multi-tenant · single deploy
powered by Azure OpenAI

Think ChatGPT, but connected to your operational data. Crews and ops folks ask questions in plain language, trigger workflows by keyword, drop in an image, and let background agents do the heavy lifting — analyze, email, escalate. There's a web app and a mobile app. One deployment serves every customer.

The big idea

human input Chat (text) natural language Q&A Keyword trigger "start workflow X" Image upload photo → workflow Mobile app in-the-field UX FieldAssist Azure OpenAI · multi-tenant connected to customer DB Shared DLT DB datastores · features · inferences automated output Workflows run multi-step actions Email / notify crews · ops · mgmt Next-step nudge agentic follow-ups the data plane

Four core capabilities

Each is a different "mode" of interacting with the same product.

01

Conversational chat

ChatGPT-style UI that already knows your operational data. Ops folks ask natural-language questions, get answers backed by live DLT data. No SQL required.

RAG over DLT no-SQL access
02

Keyword-triggered workflows

A phrase in chat (or mobile) kicks off a multi-step workflow — pull data, run analysis, send email, update a record. Tight, predictable, scriptable.

workflows deterministic
03

Image workflows

Drop in a photo from the field — equipment, gauge, asset tag — and FieldAssist runs an image-driven workflow against it. Inspection logs, gauge readings, defect detection.

vision field capture
04

Background agents

Agents run on their own schedule: scan data, detect anomalies, draft an email, escalate to the right person, or queue the next step. The "always-on" layer.

agentic async analysis + notify

Where users meet it

Web app

Browser-based chat + workflow UI. The desktop surface for ops & analysts.

Mobile app

Same product, in-the-field UX. Built for one-handed input + image capture.

Background (no UI)

Agents that run on schedule and reach out — email, SMS, Slack — without a user even opening the app.

How it's built

Single deployment, multi-tenant, Azure OpenAI underneath.

customers (multi-tenant) XRI live future customer future customer future customer FieldAssist (one app) single deployment tenant-aware routing Azure OpenAI the model layer per-tenant data (DLT DB) XRI · datastores + features workspace 1 future tenant data future tenant data future tenant data one codebase · one deploy · N customers

onboarding a new customer = pointing at their workspace · no new infra to stand up

Why it matters

Operational leverage

Ops staff get a thinking assistant sitting on top of their data 24/7. The same questions that used to take a Slack thread and a SQL person now resolve in chat, in seconds.

Built for scale

Multi-tenant + single deployment means onboarding customer #5 doesn't cost the same as customer #1. Margins improve with every tenant added.

Composes with DLT

Because FieldAssist reads from the same shared DB, it benefits automatically from any DLT featurestore, metric, or ML model we build. No re-plumbing.

Real product moat

The combination — chat + workflows + image + background agents — over operational data is rare. Most competitors do one of these, not all four.

Status & team

Where it's live

  • XRI — primary customer · heavy use of app + AI
  • Multi-tenant architecture ready for next customer drop-in

next customer to onboard is a config change, not a deploy

Who's building it

  • App dev: Abhishek lead
  • AI / agents: Abhishek · Parth
  • Dashboards: Milind · Lakshay
The short pitch: FieldAssist is what happens when you put a chat interface, a workflow engine, and a fleet of background agents on top of DLT. It's the second product, and the one with the steepest scaling curve.
← back to DLT   ·   see ecosystem view →