Intelligence & AI · App 15
From the records you keep to answers you can check
Intelligence is planned to start from metrics your One3 apps already define and to end at a tracked result, not a chart. Knowledge and Dashboards come first, in Year 1.
Intelligence turns everything a business records in One3 into answers people can trust, decisions they can track and results they can measure. Every answer cites its evidence. Every metric has one definition. Every recommendation runs through governed Actions, and its outcome is recorded against what was expected. The same tools are given to Digital Employees, whose permissions never exceed those of their human owner.
What Intelligence is
Intelligence is where a business asks questions of its own documents and numbers, and gets answers it can check. It is planned as 15 modules. Knowledge and Dashboards are the core and the Year-1 modules. Research, Context Graph, BI Studio, Data Fabric, Decision Engine, Master Data Management, Data Catalog & Governance, Forecast & Simulation, Data Science Workbench, Synthetic Data Studio, Business Twin and Data Clean Rooms are planned add-ons from W3 onwards. Data Marketplace is a planned partner service.
Four commitments
- Evidence before eloquence. Knowledge answers cite the passage they rely on, or say "I don't know". Metric answers show the definition, filters and freshness. Root-cause findings carry an evidence level: observed correlation, quasi-experimental or experimentally confirmed.
- One definition of the business. CRM, Marketing, Service Desk, Billing, People and Operations share one kernel schema, so a metric is defined once, next to the objects that produce it. Dashboards, Copilot, Digital Employees and the Decision Engine read the same figure.
- Insight has to reach action and a measured return. A recommendation becomes a decision record with a pre-registered expectation, runs only through the owning app's Actions with risk tiers and approvals, and is closed with a measured outcome. Where agents act, the result is posted to the Investment & ROI Ledger.
- Intelligence is the AI Workforce's toolset. Digital Employees get knowledge scopes, metric queries, graph context and research as governed tools, never with more access than their owner.
How it works with the rest of One3
Knowledge powers Copilot, Service Desk suggested replies, the Help Center and agent answers. A Knowledge Starter is included in the Platform so every tenant has cited answers. Dashboards read metrics that each subscribed app publishes, so a new CRM or Marketing tenant gets role home pages without modelling data. Alerts write signals that Strategy & Governance cockpits and the Decision Engine pick up. Approvals go through the Approvals Inbox. Summaries and alerts can be delivered by email and WhatsApp (template approval and opt-in to be verified).
Built for India first, usable anywhere
India-pinned tenants keep documents, embeddings, indexes, caches and inference in the India region. Number formats support lakh and crore, the fiscal year starts in April by default, and Hindi and Tamil questions are planned at Knowledge GA. DPDP erasure is designed to propagate to chunks, embeddings, answers and derived datasets (verify with counsel).
What it is not
Intelligence is not a data warehouse or lakehouse. Businesses that already run Snowflake, BigQuery, Databricks or Microsoft Fabric are planned to connect them as Data Fabric sources and read in place, while Intelligence adds the semantic layer, governance, decisions and delivery on top.
Problems solved
What Intelligence fixes
The AI tools we tried make things up
SOPs and policies live in Drive folders, WhatsApp forwards and people's heads. Knowledge is designed to answer with citations and a groundedness check, prefer "I don't know", and turn unanswered questions into draft articles.
Monday MIS takes hours and still does not match
Someone stitches CRM exports, Tally and spreadsheets by hand. Pack dashboards and role home pages are planned to ship on day one from shared-schema metrics, with freshness and quality badges on every tile.
BI tools need a data team before anyone gets an answer
Dashboards read a metric registry the apps already publish, and BI Studio is planned to generate semantic models from app manifests instead of hand-modelling raw tables.
Dashboards say something went wrong, not why or what to do
The Decision Engine is planned to go from signal to root cause with an evidence level, to options including "do nothing", to approved Actions and a measured outcome.
One customer appears five times across systems
Shared party records, MDM-lite from W1, Context Graph entity resolution and full Master Data Management are planned to give reports and agents one record.
We cannot send our data abroad
India region by default, with vector indexes, caches and inference kept in-region for pinned tenants, and erasure that reaches derived data (DPDP readiness to be verified with counsel).
Research takes days and we cannot tell which claims are sourced
Research is planned to run multi-step investigations over permitted sources with claim-level citation checks and human review before publishing.
Agencies rebuild the same client report every month
Dashboard snapshots per sub-account, a cross-client portfolio view and white-label client reports in the portal are planned for W2.
Key capabilities
What you can do
Cited answers or "I don't know"
Every claim maps to a document, version, page and quote span. Answers below the groundedness threshold are not shown.
Permission-aware retrieval
Users and agents only retrieve what they are allowed to see. A leak test runs in CI and per connector.
Metrics the apps already define
Each subscribed app publishes its metrics and Pack dashboards, so role home pages work from day one.
Freshness and quality on every tile
Tiles and answers show how fresh their data is, and warn when quality falls below the tenant's threshold.
Email and WhatsApp delivery
Scheduled summaries, threshold alerts and narratives that only use numbers copied from query results.
Closed decision loop
Signal, root cause, options, approved Actions, pre-registered expectation, measured outcome and calibration.
Gap detection
Unanswered questions cluster weekly into gaps and become draft articles for human review.
Tools for Digital Employees
knowledge.ask, metric.query and graph tools over MCP and API, scoped to the agent and its owner.
India-pinned data
Documents, embeddings, indexes, caches and inference stay in the India region for pinned tenants.
Modules
15 modules, one data model
Core modules come with the app. Add-ons switch on when you need them. Some modules can be bought on their own.
Knowledge
Answers from your own documents, with the source, or an honest "I don't know"
ExploreDashboards
Role home pages built from metrics your apps already define
ExploreResearch
Multi-step research where every claim links to evidence you can open
ExploreContext Graph
How your customers, deals, tickets and suppliers relate, across every app
ExploreBI Studio
A semantic layer and certified metrics that every team, tool and agent reads
ExploreData Fabric
Bring ERP, spreadsheet and warehouse data in, with quality checks and lineage
ExploreDecision Engine
From signal to root cause to approved action to a measured, calibrated outcome
ExploreMaster Data Management
One golden record for each customer, supplier, product and employee
ExploreData Catalog & Governance
Know what data you hold, who owns it, where personal data sits and who used it
ExploreForecast & Simulation
Forecasts with ranges, backtests and stated assumptions, not a spreadsheet nobody trusts
ExploreData Science Workbench
Governed notebooks and models, deployed as skills only after evaluation and approval
ExploreSynthetic Data Studio
Realistic test and training data that does not expose real people
ExploreBusiness Twin
A validated "what happens next" projection of your business, built from its own event log
ExploreData Clean Rooms
Measure overlap with a partner without exchanging raw customer data
ExploreData Marketplace
Licensed data products with lineage, quality scores and enforced usage terms
ExploreWorkflows
How the work flows
End-to-end workflows inside the app and across One3.
Question to cited answer
How a question from Copilot, Desk, an agent, the portal or the API becomes a cited answer, or an honest "I don't know".
- 1.A question arrives from Copilot, Service Desk, an agent, the portal or the API
- 2.Entitlement and permissions are checked and the actor's allowed ACL keys are loaded
- 3.Hybrid retrieval combines vector and full-text candidates, then reranks them
- 4.An answer is drafted with citations and passes a groundedness check
- 5.Below threshold, the user gets "I don't know" plus search results and the query is logged as no-answer
- + 1 more steps →
Knowledge gap to published article
Questions the company cannot answer yet become reviewed content.
- 1.No-answer and refused queries are clustered weekly
- 2.Each gap is stored with example queries (PII removed) and counts
- 3.A draft article is created in Content Hub for review
- 4.A person edits and publishes it
- 5.The gap closes only when the new source is indexed and the example queries now answer
Scheduled summaries and threshold alerts
Numbers reach people where they read, each rendered with that recipient's permissions.
- 1.An author schedules a dashboard or sets a threshold rule on a metric
- 2.Each schedule renders once per recipient with that recipient's permissions
- 3.Narrative text is checked so it contains only numbers from query results
- 4.Delivery goes to in-app, email, WhatsApp (opt-in and approved template) or Slack and Teams
- 5.Threshold breaches write a deduplicated signal and notify the owner
Signal to calibrated outcome
The planned Decision Engine loop from an alert to a measured, calibrated result.
- 1.A signal arrives from an alert, anomaly, connector health or ROI threshold
- 2.Root cause is drafted using metric trees, segment drill-down and graph traversal, each hypothesis labelled by evidence level
- 3.Options are generated, always including "do nothing", with ranges for impact, cost and reversibility
- 4.A critic review flags missing options and optimistic estimates
- 5.The owner approves, chooses "test first" or dismisses with a reason, and registers an expectation
- + 2 more steps →
External source to certified metric
Bringing ERP, spreadsheet or warehouse data into governed use.
- 1.A data engineer connects a source through Integration Hub
- 2.Incremental pipelines load data with checkpoints and quality rules
- 3.Datasets with personal data need purpose tags before they can be published
- 4.Datasets join to core entities through cross-references in BI Studio
- 5.A metric owner certifies the metric after review
- + 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.
Cited answer generation A4
Drafts answers with citations, a groundedness check, confidence and "I don't know" below threshold. Read-only.
Gap clustering and draft articles A2
Clusters unanswered queries weekly and drafts articles in Content Hub; publishing stays with a person.
Narrative "what changed" summaries A3
Writes tile-linked sentences where every number is copied from query results; suppressed on a numeric mismatch.
Auto-dashboards per role A3
Creates private-by-default draft dashboards from subscribed modules; sharing is a human step.
NL to metric query A4
Turns a question into an editable metric query over certified metrics only, never raw SQL to tables.
Root cause and option drafting A2
Proposes hypotheses with evidence levels and options including "do nothing"; always reviewed by a person.
Pipeline generator A2
Drafts a pipeline, mapping and quality rules from a description and a sample; publishing is human.
Research plan and citation verifier A2
Plans sub-questions and checks each factual sentence against its evidence; unsupported claims are removed or marked unverified.
Forecasting with backtests
Produces P10/P50/P90 forecasts and refuses to use models that do not beat a seasonal naive benchmark.
Digital Employees
Agents that operate Intelligence
Rent them by the month. They work under your team's supervision and their cost and returns appear in the Investment & ROI Ledger.
AI Knowledge Base Curator
Clusters gaps, drafts articles, flags stale and conflicting documents. Publishing always stays with a person. Planned as the rentable Knowledge Curator template.
Learn moreAI Data Analyst
Runs metric queries and creates private draft explorations and dashboards for review.
Learn moreAI Insights Narrator
Drafts weekly "what changed and why" briefs with recommended next steps; takes no Actions.
Learn moreAI Decision Journal Keeper
Creates decision records from approvals above R2 and links outcomes at review dates.
Learn moreAI Data Quality Steward
Detects rule failures and drift and proposes fixes; rule activation needs a person.
Learn moreAI Metrics Steward
Detects conflicting metric definitions and proposes changes; never certifies a metric.
Learn moreAI Strategy Analyst
Runs research projects within an approved plan and budget and drafts reports for review.
Learn moreAI FP&A Analyst
Runs forecasts and drafts variance commentary and scenario inputs for planner review.
Learn moreIntegrations
Connects with the tools you already use
- Google Drive
- Gmail
- Notion
- SharePoint / OneDrive
- Confluence Cloud
- Slack
- Microsoft Teams
- Zoho WorkDrive
- WhatsApp Business Platform
- PostgreSQL
- MySQL
- SQL Server
- Google Sheets
- Tally
- Zoho CRM and Books
- Shopify
- SAP
- Snowflake
- BigQuery
- Databricks
- Microsoft Fabric / OneLake
Connector availability depends on provider access and is verified before launch. How integrations work →
Security & compliance
Built to be trusted
Four scopes, always intersected
Workspace scope, object sharing, source-app row scope and field-level masks apply together in the UI, API, exports, narratives, schedules, MCP tools and agent calls.
Forced tenant isolation
Row-level security is enabled and forced on every Intelligence table; tenant identity comes from the token only and caches are tenant-keyed.
Region pinning without silent fallback
India-pinned tenants keep documents, embeddings, indexes, caches, datasets and inference in-region; out-of-region model fallback is refused.
Small-group suppression
Aggregates over personal or employee data suppress small groups and guard against difference attacks.
Erasure that reaches derived data
DPDP erasure is designed to propagate to chunks, embeddings, answers, feedback and gap examples with an evidence record (verify with counsel).
Secrets and keys handled by the platform
Connector credentials live in the cloud secret manager; BYOK is a planned Enterprise option; prompts are redacted before tracing.
Industry Editions
Industries that use Intelligence
Frequently asked questions
Is Intelligence available today?
No. It is being built in waves. Knowledge is planned from W0 and Dashboards from W1, both inside the Year-1 boundary. The other modules are planned for W3 and later.
Do I need a data team to get dashboards?
Not for the core. Each subscribed app publishes its own metrics and Pack dashboards, so role home pages are designed to work without modelling data. BI Studio adds a semantic layer later for custom metrics.
What happens when Knowledge does not know the answer?
It says "I don't know", shows search results and logs the question. Repeated unanswered questions become a gap and a draft article for review.
Can a Digital Employee see more data than I can?
No. An agent's effective access is its granted scope intersected with its owner's permissions and the tenant's entitlements.
Do I have to move off my warehouse?
No. Snowflake, BigQuery, Databricks and Microsoft Fabric are planned as read-in-place Data Fabric sources (API terms to be verified).
Is there a Knowledge option without buying the app?
Yes. A Knowledge Starter with limited collections and documents is included in the Platform so Copilot and Helpdesk work for every tenant. Limits are set by pricing tests.
Does Intelligence help with DPDP?
It is designed with India region pinning, purpose tags, personal data maps and erasure propagation. These are tools, not legal advice; verify obligations with counsel.
Works with
Related apps
Platform
One governed foundation for every human, AI agent, integration and channel
ExploreCRM & Sales
A WhatsApp-first CRM where sellers and Digital Employees work the same pipeline
ExploreMarketing
Consented, measurable marketing across WhatsApp, email, search, ads and social
ExploreService Desk
Support on every channel, resolved on the same customer, invoice and order records
ExploreOffice
Mail, calendar, files, meetings and signatures attached to the work they are about
ExploreFinance
Bill, collect, pay, account, plan and comply on the records your business already keeps
ExploreStrategy & Governance
Set direction, run the management rhythm and govern the company on one evidence base
ExploreAI Workforce
Rent ready-made AI Digital Employees that work for your team, with a monthly ROI statement
ExplorePlan your Intelligence 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.