Skip to content
ONE3.ai

Build & Ecosystem · App 14

Describe a process, review the draft, run it on One3

Workflows, forms, surveys, documents and tenant apps built on the same kernel as every One3 app. AI drafts, humans publish, and every run shows its cost and its return.

Planned · Year 1 (W0)7 modulesYear 1: In
Every process a business runs can be described, built, run, governed and measured on One3, by people and Digital Employees working on one canvas, with the cost and the return of every run visible.

What Build is

Most small and mid-market teams assemble a three-to-five-tool stack — an app builder, an automation tool, a form tool, a survey tool and a proposal tool — and then integrate it all back into their CRM, desk and finance data. Build is designed to replace that stack with seven modules on the One3 kernel: Automations and Forms in the Year-1 core, Survey and Documents & Reports as add-ons, Apps Builder for tenant-built apps, the Developer Platform for partners, and Factory for governed spec-to-code delivery in W5.

Build on the OS, not beside it

Workflows trigger on kernel events from any subscribed module — a lead created, a ticket closed, a form submitted — and call the typed Actions of those modules. No connector is needed for first-party data. Form submissions land on shared core records such as CRM leads and Desk tickets. Tenant apps are Packs on the same kernel as first-party apps, with the same permissions, audit and events.

Humans, agents and steps on one canvas

A step can be a human approval, a Digital Employee skill or an Action, all under the same approval, disclosure and stop-trigger policy. Approvals can happen in the Approvals Inbox or, for opted-in approvers, on WhatsApp.

AI drafts, humans publish

You describe a workflow, form, survey, template or app in plain language and get an editable draft with an explanation, assumptions and open questions. The generator only proposes Actions you are entitled to and allowed to use. A draft with R3 or R4 steps cannot be published until it passes a test or dry run, and a second person reviews the diff. No agent can publish a definition at any autonomy level.

Governed by default

A published workflow runs as a service identity whose authority is frozen at publish time to the intersection of what it was granted and what the publisher could do. If permissions drift, the run pauses; it never escalates. Entitlements are re-checked at every step, and loop guards, budgets, quiet hours and a kill switch are built in.

Measured, not assumed

Workflow runs — not steps — are metered. Every run, submission and generated document is designed to post cost and, where attributable, outcome to the Investment & ROI Ledger, including human supervision time. Baselines entered by the tenant are labelled as estimates.

India-first building blocks

WhatsApp approvals and WhatsApp Flows forms, Hindi and Tamil, GSTIN, PAN, PIN code and INR fields, UPI payment fields through a licensed partner and DPDP-style consent records per purpose are planned. Each must be verified before launch.

Problems solved

What Build fixes

A stack of builders that must be stitched back to your data

Builders, automation, forms, surveys and proposals in separate tools. Planned fix: one app, one data model and one permission model, with outputs landing on shared core records.

Automation pricing that grows with every step

Per-step or per-task pricing becomes unpredictable. Planned fix: meter workflow runs, not steps, and show cost per run in the ROI Ledger.

Builders without governance

Weak environments, versioning and audit. Planned fix: versioned publish, sandbox, run-as identity capped at the creator's permissions, kernel audit and risk tiers.

Proprietary scripting and unmanaged AI code

Lock-in and maintenance debt. Planned fix: metadata-first Packs in an open format, TypeScript extensions in a sandbox and 3-way-merge upgrades.

Agents sold separately from process design

Agents billed and designed apart from workflows. Planned fix: agent steps inside workflows using Agents Studio skills, the autonomy ladder and eval gates.

No way to tell whether an automation paid back

Tools rarely account for outcomes. Planned fix: ROI Ledger posting per run, submission and generated document.

Indian processes run on WhatsApp and local rules

Global builders do not lead with WhatsApp, Indian languages, GST, UPI or DPDP. Planned fix: WhatsApp approvals and Flows, Hindi and Tamil, India field types and consent records (verify).

Agencies rebuilding the same setup for every client

Client setups are slow to replicate. Planned fix: workflows, forms and later tenant apps included in snapshots and installable per sub-account.

Key capabilities

What you can do

Durable workflows on kernel events

Event, schedule, webhook, form and manual triggers calling any entitled Action, with runs that survive restarts.

Workflow from a description

Plain-language or voice description to a draft workflow with explanation and open questions.

Approvals in the inbox or on WhatsApp

Approval steps with diff preview and AI summary; opted-in approvers can decide on WhatsApp.

Forms that create real records

Each submission invokes an Action such as creating a CRM lead or Desk ticket, with consent and spam protection.

India field types and WhatsApp Flows

GSTIN, PAN, PIN code, Indian mobile and INR fields, Hindi and Tamil, and forms published as WhatsApp Flows (verify).

Permission ceiling and four-eyes publish

Published workflows run with frozen authority that never escalates; risky publishes need a second reviewer.

Runs metered, cost visible

Runs, not steps, are metered, with cost lines and outcomes posted to the ROI Ledger.

Tenant apps as Packs

Objects, fields, views, Actions and roles built beside first-party apps, upgrade-safe with 3-way merge.

Open format, no lock-in

Definitions export as versioned JSON and Packs use an open format with TypeScript extensions.

Workflows

How the work flows

End-to-end workflows inside the app and across One3.

Design, test, publish and run a workflow

From a plain-language description to a governed, metered run with approvals and ROI posting.

  1. 1.Describe the process in plain language or pick a template
  2. 2.AI drafts steps, mappings and assumptions
  3. 3.Builder edits the draft on the canvas
  4. 4.Validation checks entitlements, schemas, permissions and loop risk
  5. 5.Test run or dry run with a sample record or payload
  6. + 5 more steps →

Form submission to record

A respondent submits a form on the web, an embed, a portal or WhatsApp, and it becomes a real record with consent and follow-up automations.

  1. 1.Respondent opens the published form version
  2. 2.Submission is rate-limited, validated against the version schema and spam-scored
  3. 3.Spam gets a generic thank-you and is stored separately
  4. 4.Consent records are created for the purposes ticked
  5. 5.If there is a payment field, a payment link or UPI intent is created through the payment partner
  6. + 2 more steps →

Describe to tenant app

A business requirement becomes a sandbox app, then a published Pack version.

  1. 1.Builder describes the app or uploads a spreadsheet
  2. 2.AI drafts objects, fields, relations, views and seed records
  3. 3.App is installed only in a sandbox with synthetic or masked data
  4. 4.Builder edits roles, permissions, forms and workflows
  5. 5.Publish to production with a migration preview
  6. + 1 more steps →

Spec to release in Factory

An idea becomes an approved spec, task packets, reviewed code, tests and an approved release.

  1. 1.Ideas and clustered feedback are linked into a project brief
  2. 2.AI drafts a spec with user stories and acceptance tests; a human approves it
  3. 3.Spec is split into task packets with pre-written acceptance tests
  4. 4.Packets go to AI executors or human developers within budgets
  5. 5.Code lands as pull requests with mandatory human review
  6. + 1 more steps →

AI inside

AI skills that act through governed Actions

Every AI skill runs through the same typed Actions people use, with risk tiers, approvals and an audit trail.

Workflow from description A2

Turns a typed or voice description into a draft workflow with explanation, assumptions and open questions. Design-time generators stay at A2.

AI step in a workflow

Classify, extract, summarise or draft inside a run with a required output schema, model policy and cost cap.

Failure repair suggestions A2

Explains a failed run in plain language and proposes a fix as a draft diff for a human to apply.

Form from prompt A2

Drafts fields, logic and Action mapping from a description.

Open-text themes A3

Multilingual theme and sentiment extraction from survey answers with representative quotes and editable labels.

Document drafting A3

Drafts proposal narrative or report commentary from record data and Knowledge with citations and data-quality warnings.

Describe to app A2

Drafts an app definition with explanation and synthetic seed records, installed only in a sandbox.

Manifest validation assistant A2

Explains Pack manifest errors and suggests fixes for developers.

Spec generation A2

Drafts PRDs, user stories with acceptance tests, data model and Action lists for Factory projects; approval is human.

Integrations

Connects with the tools you already use

  • WhatsApp Business Platform
  • Email
  • SMS via Indian aggregators (DLT)
  • Captcha provider
  • Payment partner via Billing (UPI)
  • E-signature via Sign
  • GST e-invoice via Billing
  • DigiLocker
  • GSTIN verification lookup
  • Any REST API via HTTP step
  • Inbound webhooks
  • Google Workspace
  • Microsoft 365
  • Slack
  • Google Sheets
  • GitHub
  • GitLab
  • External MCP servers
  • Zapier, Make and n8n import

Connector availability depends on provider access and is verified before launch. How integrations work →

Security & compliance

Built to be trusted

No agent publishes a definition

Publishing workflows, forms, surveys, templates and apps, and submitting listings, are excluded from agent autonomy at every level.

Permission ceiling that never escalates

Published workflows run as a service identity limited to what was granted and what the publisher could do at publish. Drift pauses the run.

Entitlement and policy re-checked at every step

A lapsed module, budget, unhealthy connector or kill switch pauses the run with a typed reason instead of failing silently.

Untrusted input never plans

Webhook payloads, submissions, uploaded files, imported definitions and marketplace manifests are read by a quarantined model that returns structured data only.

Separate public surface

Hosted forms, surveys and proposals are served on separate domains with their own CSP, rate limits and bot controls, by an identity restricted to one Action.

Respondents are never seats

People who submit forms or answer surveys are never charged or counted as seats.

Frequently asked questions

What is available in Year 1?

Automations (W0) and Forms (W1) are inside the Year-1 boundary. Survey and Documents & Reports are planned for Year 2, Apps Builder reaches tenants in W3, the Developer Platform is W3, and Factory is W5. Waves are earliest eligible, not commitments.

How is Automations priced?

The plan meters workflow runs, not steps. Test and dry runs are not metered. Specific prices are set by the pricing team and are not listed here.

Can AI build and publish a workflow on its own?

AI produces an editable draft. A human publishes it, and drafts with high-risk steps need a passing test and a second reviewer.

Do I need connectors to use my One3 data?

No. Workflows trigger on kernel events and call Actions of subscribed modules directly. Connectors are used only for external systems.

Can I take my definitions elsewhere?

Workflow definitions export as versioned JSON, and Packs use an open format with TypeScript extensions.

Does Build do RPA or desktop automation?

No. A browser agent step with an approval gate is planned for W4; desktop RPA is not planned.

Plan your Build rollout with us

Tell us about your team and the outcomes you want. We will walk you through the apps, the Digital Employees and the plan for your industry.