Status: Available
elan1 Decide Decision Studio
Build typed decisions: Boolean, Choice and Score questions, each with a fixed shape.
Decision Studio in detailDecision Intelligence Platform
elan1 Decide: calibrated AI decisions, routed by risk
elan1 Decide turns business questions into typed, calibrated decisions. Each answer, a Boolean, Choice or Score, carries a tested confidence and is routed by risk: automatic, human review, a System-2 second pass, or hold. Every answer is kept in the Decision Ledger, and any write that follows still goes through elan1 Govern.
Status: Available · governed actions in Mode B
- Question types
- Boolean · Choice · Score
- Routes
- Auto · review · System-2 · hold
- Mode
- B · evaluated and recorded
01What it does
Rules cannot read the document. LLM agents cannot afford the volume.
Each case becomes a typed question, Boolean, Choice or Score, and gets one calibrated answer. A router then sends it by risk: auto, review, a deeper System-2 pass, or hold. Only a resolved route produces an action intent, and that intent goes to elan1 Govern.
Decide answers, scores and routes. It never allows a write: that verb belongs to elan1 Govern.
Illustrative specimen — not customer dataDecision Ledger
claims.fnol_triage · Choice case clm_specimen_0412
- question
- Which handling path fits this first notice of loss?
- shape
- Choice · fast_track | standard | investigate
- answer
- standard
- confidence
- 0.71 · threshold 0.85
- route
- review — below the threshold, a person decides
- reviewer
- usr_specimen_reviewer (named)
- action_intent
- none until the route resolves
- ledger_digest
- sha256:9a1f…c2e0 · bound into the proof record of the write that follows
Routed to review · no record changed
Decide answers and routes. Only elan1 Govern allows a write.Mode B
Who it’s for
Teams with volume and a duty of care.
- Operations owners in insurance claims, including first notice of loss.
- BFSI teams working AML alerts, KYC and disputes.
- Teams with a million or more structured cases a month, arriving as unstructured input.
02How it works
One case, one typed question, one route.
Most cases take a single calibrated call. The unsure ones go to a person or a deeper pass. Only a resolved route can ask for a write.
- Your caseArrives unstructuredA claim form, an alert, a document: input that rules cannot read.
- Typed schemaBecomes a typed questionBoolean, Choice or Score, each with a fixed shape.
- System-1Gets one calibrated answerOne call per case, with a stated confidence.
- RouterIs routed by riskAuto, review, a deeper System-2 pass, or hold, chosen by risk tier.
- A named personDecides when unsureBelow the threshold, the Review Console shows the case and the score side by side.
- elan1 GovernAllows or refuses the writeA resolved route produces an action intent. Govern checks it and records the outcome.
How it connects to elan1 Govern
Decide decides what the answer probably is. Govern decides whether any record changes. If they disagree, Govern wins.
Decide never builds its own gate, approval store, seal, meter or verifier. A decision is not a governed action. Only the writes that follow a decision pass through Govern.
03Key capabilities
Build, calibrate, route, review, record.
Each module carries its status. Nothing here is shipped.
- 01
Typed schemas
Boolean, Choice and Score questions, each with a fixed shape.
Available - 02
Calibrated System-1 call
One call per case, with a stated confidence.
Available - 03
Router
Auto, review, System-2 or hold, chosen by risk tier.
Available - 04
Review Console
Unsure cases go to a person, who sees the case and the score side by side.
Available - 05
Decision Ledger
Every answer, with its route. Its digest is bound into the governed write's proof record.
Available - 06
Decision Studio, Calibration Lab and Decision Packs
Build, calibrate and package decisions. Packs compile to Govern's seven-block declarations.
Available
04Modules
Six modules. One meter.
Every module is included in every plan. Each layer charges for one thing it can count, and nobody pays twice for the same action.
- Status: Available
elan1 Decide Router
Routes each answer by risk tier: auto, review, System-2 or hold.
Router in detail - Status: Available
elan1 Decide Calibration Lab
Calibrates and evaluates each decision for your tenant. No calibration, no automation.
Calibration Lab in detail - Status: Available
elan1 Decide Review Console
Unsure cases go to a person, who sees the case and the score side by side.
Review Console in detail - Status: Available
elan1 Decide Decision Ledger
Every answer, with its route. Its digest is bound into the governed write's proof record.
Decision Ledger in detail - Status: Available
elan1 Decide Decision Packs
Package a decision for one kind of case, such as claims triage. Packs compile to Govern's seven-block declarations.
Decision Packs in detail
No module writes a record. Writes go to elan1 Govern as intents.
Decide pricing by module05Inside the product
The Decide console, screen by screen.
From the command centre to one decision's thread: the catalogue, the Studio, calibration, the router's policy, the Review Console and the ledger. Choose a screen; each lists what to look for.
01 / 14 · Overview
Command Center
- Every headline figure carries its basis: Measured, not estimated.
- Route mix shows how traffic split between auto, review, System-2 and hold each hour.
- The live pulse lists individual decisions with their minimum confidence and route.
- Review backlog shows breached and due-soon items, so the queue is never a surprise.
02 / 14 · Decisions
Decision catalogue
- A decision moves through five stages, from Draft through Shadow and Canary to Live, one gate at a time.
- Certification needs a named decision owner and risk owner, neither of them the author.
- Versions are pinned, so callers never get “latest” by accident.
03 / 14 · Decisions
Decision detail
- Decision owner and risk owner are named people, shown on the record.
- Thresholds v4 shows who signed them, when, and from which cost matrix.
- Rollback triggers watch QA accuracy, calibration error, latency and drift.
- Auto-action can be switched off for the auto band without redeploying.
04 / 14 · Design
Decision Studio
- Seven steps from Author to Compile, each with a visible status.
- Every answer is Boolean, Choice or Score, so it can be calibrated and audited.
- The same lint rules (DS-01 to DS-10) run in CI with no platform present.
05 / 14 · Prove
Calibration and Evaluation Lab
- Golden sets are frozen by hash and split 60/20/20.
- Agreement (kappa), calibration error (ECE) and accuracy are shown against their floors.
- Cut-over readiness lists each gate, through non-inferiority and shadow.
06 / 14 · Route
Router and policy
- Projections are labelled Modelled, separate from measured figures.
- New thresholds take effect only when the risk owner signs them.
- Tier bands set defaults; a signed cost matrix replaces them per decision.
- A kill switch sits alongside the bands, not buried in settings.
07 / 14 · Design
Simulation
- Replays draw a stratified sample from 30 days of recorded decisions.
- Automation, auto-band error, System-2 cost and review load are compared side by side.
- Runs are deterministic given inputs and versions, so they can be replayed in an audit.
08 / 14 · Operate
Review Console
- Personal data is redacted before inference; tokens such as PERSON_1 stand in its place.
- The reason a case is here is stated: in the review band, or no single answer.
- Confirm, Override, Escalate or Defer, with SLA timers on every item.
- Cases are claimed on a 10-minute lease so two people never work the same one.
09 / 14 · Record
Decision Ledger
- Append-only: rows are added, never edited in place.
- Saved views for typed holds, System-2 escalations, review bands and QA samples.
- Each row shows route, minimum confidence, model, latency, cost and caller.
10 / 14 · Record
Decision thread
- Each step is timestamped, from request received to ledger acknowledgement.
- Pinned versions list everything that produced the route; “latest” is rejected.
- A determinism replay re-runs the route from the recorded answers and versions.
11 / 14 · Govern
Governance
- A model never produces an allow; policy and named people do.
- Refusals are never retried. elan1 Govern meters each one at 40% of a governed action; Decide counts the decision once.
- The claims boundary states plainly what the record can and cannot be said to prove.
- Each write intent records which enforcement mode applied to it.
12 / 14 · Learn
Outcomes and learning
- Nine steps around the Decision Ledger, from Decide to Train heads.
- Label mix shows gold, silver, outcome and weak labels, so you know what learning rests on.
- The learning plane never sits on the live request path.
13 / 14 · Models
Model fabric
- Champions are pinned per schema version; challengers run in canary.
- Residency filters show which engines run in-region (ap-south-1).
- Latency, licence and region are listed for every engine.
14 / 14 · Operate
Operations
- Nine service levels, each measured on production traffic before it is quoted.
- An at-risk SLA is flagged, not averaged away.
- Fail direction and paging are set per dependency, with a kill switch to hand.
Illustrative specimen. Not customer data. Screens come from the elan1 Decide console running on fictional sandbox data.
06Assurance level and mode
Mode B: evaluated and recorded.
L1 · Guarded
Available now.
Mode B, attested advisory. Decide's writes to Govern are evaluated and recorded. Scaling in regulated BFSI depends on L2 evidence.
Mode B · Attested advisory
How Decide connects to Govern
Every write that follows a decision goes to elan1 Govern over KernelPort, where it is evaluated against your rules and recorded with the decision's ledger digest.
- We sayEvaluated and recorded.
- We never sayEnforced, or decided by the model.
Governed actions from Decide run at L1 Guarded, with L2 Attested available.
07Meter and pricing
Paid per decision, never per write twice.
The unit
One decision
Writes that follow are metered by Govern, once.
Govern meters the writes that follow a decision. Decide never counts them a second time.
- A yearly platform plan with a decision allowance included; decisions beyond it are billed per 1,000.Excl. GST
- System-2 passes are passed through at model cost plus 30%, on their own line.Excl. GST
- Every plan includes all six modules; individual Decision Packs from the catalogue are also priced per pack. A write intent that Govern refuses is metered by Govern at 40% of a governed action; the decision itself is counted once, as a decision.Excl. GST
Platform plans
Every plan includes all six modules: Decision Studio, Router, Calibration Lab, Review Console, Decision Ledger and Decision Packs. Individual Decision Packs from the pack catalogue are also priced per pack. About 30% of a plan's typical monthly volume is included; decisions beyond it are billed per 1,000.
| What | Price · excl. GST | Includes |
|---|---|---|
| Growth — self-serve and developer teams | ₹4,80,000 per year | Typical volume 1,00,000 decisions a month |
| Mid-market | ₹36,00,000 per year | Typical volume 15,00,000 decisions a month |
| Enterprise — BFSI and insurers | ₹1,50,00,000 per year | Typical volume 1 crore decisions a month; SSO and SLA |
Usage
| What | Price · excl. GST | Includes |
|---|---|---|
| Decisions beyond the plan allowance | ₹30 per 1,000 decisions | |
| System-2 pass (escalated decision) | Model cost + 30% | Shown separately on the bill; about ₹1.14 per escalated decision at today's model prices |
| A write intent Govern refuses | 40% of a governed action | Metered by elan1 Govern, not by Decide; the decision itself is counted once, as a decision |
Packs and services
| What | Price · excl. GST | Includes |
|---|---|---|
| Decision Packs | Priced per pack | The Decision Packs module is in every plan; each catalogue pack is priced individually |
| Implementation — Mid-market | ₹5,00,000 one-time | |
| Implementation — Enterprise | ₹30,00,000 one-time |
All prices are exclusive of GST. Prices are current list prices; GST is added at the applicable rate on the invoice.
08Where it is weak
What elan1 Decide does not do.
Plain limits, so you can plan around them.
- It is Mode B only.Decide's writes are evaluated and recorded. We never say they are enforced.
- The cost comparison is per deployment.We measure it on your own cases during shadow. We publish no general figure.
- A confidence can drift.It is calibrated on past cases. Below your threshold a person decides, and the Calibration Lab is where drift shows.
- A decision changes no record.A decision is not a governed action. Only the writes that follow pass through Govern.
- Regulated scale needs L2.Scaling in regulated BFSI depends on L2 evidence: signed records, checkable offline.
- Outside India, decisions only.Decision-only pods run until an elan1 residency domain exists in that region.
09Status
Available now. All six modules.
Every elan1 Decide plan includes all six modules.
- ModulesDecision Studio, Router, Calibration Lab, Review Console, Decision Ledger and Decision PacksAvailable
- Mode BGoverned actions through elan1 Govern, evaluated and recordedAvailable
- L2elan1 Govern records signed and checkable offline, for regulated scaleAvailable
Available now. Future improvements are listed, in order and without dates, on What’s next.
Announced 10 Oct 2026: elan1 Govern, Work, Decide and Run are available.
10FAQ
Nine direct answers.
Questions a business answers many times over, typed as a Boolean, a Choice or a Score: insurance claims triage, AML alert triage, KYC review and payment disputes, among others. Each decision is defined once in Decision Studio, with a fixed shape, before any model answers it.
Whenever the answer's calibrated confidence falls below your threshold, or the case's risk tier says a person should look. The Router then sends it to the Review Console, where a reviewer sees the case and the score side by side. No answer changes a record on its own authority.
Calibrated confidence means that when a model says it is 90% sure, it is right about 90% of the time on your own cases. The Calibration Lab tests every decision on your tenant's cases first, and a decision without calibration is never routed to automatic.
Not by itself. A resolved decision can produce a write intent, and that write goes to elan1 Govern over KernelPort in Mode B, where it is evaluated against your rules and recorded with the decision's ledger digest. So the writes that follow a decision are evaluated and recorded, in Mode B.
A yearly platform plan with a decision allowance: Growth at ₹4,80,000, Mid-market at ₹36,00,000 and Enterprise at ₹1,50,00,000 a year, each including all six modules. Decisions beyond the allowance cost ₹30 per 1,000, and System-2 passes are passed through at model cost plus 30%. Prices exclude GST.
System-1 is the single calibrated call that handles most cases. System-2 is a deeper pass for the cases the router escalates. You see its cost as a pass-through.
It depends on your cases. We measure it on your own volume during shadow and report it to you before any route runs on its own. We publish no general figure.
Yes, as decision-only pods, until an elan1 residency domain exists in that region.
No. Decide never builds its own gate, approval store, seal, meter or verifier. Its Decision Ledger records every answer and its route, and that digest is bound into the proof record elan1 Govern writes for the write that follows.
Next steps
Related reading
AI for AML alert triage and KYC reviewCalibrated confidence, risk-based routing and maker-checker review.RBI FREE-AI and India's AI governance guidelinesWhat banks, NBFCs and firms using AI agents should keep as evidence.elan1 GovernWhere the writes that follow a decision are checked.The proof recordWhere a decision's ledger digest is bound.Assurance levelsWhy regulated scale waits for L2.Reference architectureWhere Decide sits above the KernelPort seam.
One write path, four layers
Proposeselan1 WorkProposes what should changeProposeselan1 DecideProposes the answerControl planeelan1 GovernAllows, refuses, recordsRuntimeelan1 RunCarries every call under contract
Work and Decide propose. Govern and Run assure. Only Govern allows.
// elan1 DECIDE
Bring one decision to a lighthouse.
Pick a high-volume decision. We run it in shadow beside your current process before any route goes live.