SAS Viya IFRS 17 · Governed reporting engine

IFRS 17
AI Studio

{{ bootN1 }}Measurement engine

{{ bootN2 }}Semantic layer

{{ bootN3 }}Use cases

{{ bootN4 }}Insights

studio.ifrs17.ai · Production demonstration

SAS^® Viya^®

Welcome to SAS Viya

Let us help you get started with some demos and hands-on experiences.

Productivity and Speed

From data to discovery to deployment, the intelligent automation of SAS Viya supports you through the full analytics life cycle.

Experience SAS Viya Copilot

Show me how to use Copilot Quick start a Viya workflow Help me access enablement content

Data & AI Life Cycle Quick Start

Solutions

You have signed out.

For increased security, close your browser.

IFRS 17 AI Studio Focus mode · Esc to exit

IFRS 17 AI Studio

{{ bc.label }} /

IFRS 17 AI Studio

V2 Operational Intelligence

Needs attentioni

Workspaces

Analysis

Accounting

Control & data

Vasudevan Chandrashekran

Group Actuarial

{{ period }} · FY26 close

Explain this number

›

What moved it

Where the number comes from

→

CFO pack · board-ready movement story

Three slides, written from the {{ cfoFy }} numbers. Charts export as editable PowerPoint shapes, with commentary in the speaker notes.

Insurance revenue composition

CSM movers by product line

Decisions for the board

SAS IFRS17 AI Studio · {{ cfoScope }}{{ cfoSlide.num }} / 3

Auditor pack

One file of evidence for one contract group

Lineage maps, hop evidence, reconciliation, assumption approvals and a sign-off sheet, sealed with a SHA-256 fingerprint.

Loading the traceability index…

Contract group

Format

Sections

Pre-flight checks

Run against the extract before the pack is sealed.

Welcome

Generative and agentic AI across your IFRS 17 results, disclosures and governance. Pick up where you left off, or jump into a module below.

Pinned insights

Annual report insight

Regulator watch

NEW {{ r.item }}

Audit findings & remediation

High: {{ auditHighDetail }}

Avg. days to close: {{ auditAvgDays }} (down from {{ auditAvgDaysPrior }})

Historical performance — 6-year trendi

{{ s.pct }}%

What changed this period

Full audit trail

Every assumption change, sign-off and AI-drafted narrative is logged with reviewer and timestamp.

FY2017 close is in progress

CSM closed at {{ nf.NT001 }}, down {{ nf.NT002 }} since opening. 2 assumption changes worth {{ nf.NT003 }} are pending sign-off and 2 AI-drafted narratives are awaiting review in Governance & Audit Trail.

Ask the Copilot

Ask a natural-language question over your IFRS17 results and get a sourced, confidence-scored answer.

The sections below read live from Data Hub tables in the "Home Intelligence Feed" library. That library is fed by a "Home Intelligence Feed" source & collection — the same annual reports, audit reports and regulator letters, viewable as uploaded documents in Retrieval Agent Manager — plus a daily automated web scrape of regulator publication pages. View in Retrieval Agent Manager → · Open Data Hub →

Since go-live — 6 years under IFRS 17

IFRS 17 explainers

{{ v.duration }} · {{ v.watchLabel }}

No explainer videos match "{{ homeVideoSearch }}".

EXECUTIVE COMMAND CENTER

Scope · {{ execLensScope }} Demo data

Workbook · proxy

Drill-downs · {{ execKpiLabel }}

Demo data

{{ c }}
{{ cell.text }} {{ cell.text }}

Source · {{ p.sourceLabel }}

Lineage · {{ p.lineage }}

Workbook · proxySection figures computed from ICG_PROXY_MEASURES and EXPERIENCE_VARIANCE_ICG

What changedi

Why it changedi

Material itemsi

Ranked by materiality score · click to open the underlying insight

{{ mi.link }} →

Forward viewi

Model {{ execForwardModel }} · uncertainty {{ execForwardBand }} · WW_INSURANCE book, MYR K

Reported Illustrative forecast

Decisions requiredi

{{ d.owner }} · by {{ d.by }}

Recorded: {{ d.chosen }} — logged to the audit trail.

Executive overview, financial, risk and forecast lenses are computed from the SAS WW_INSURANCE sample workbook as at 31 Dec 2017 (amounts relabelled MYR). CSM, loss component, revenue and service result are proxies pending SAS IFRS17 engine output. Close operations and governance lenses remain illustrative demo values.

Close readinessi

{{ closeBand }} FY2017 close review population

{{ closeScore }} / 100

Control gatesi

Work queuei

Ranked by materiality

Item	Value	Materiality	Owner	Team	Status	Age
{{ x.item }}	{{ x.value }}	{{ x.score }}	{{ x.owner }}	{{ x.team }}	{{ x.status }}	{{ x.age }}	Investigate →

No items match these filters.

Illustrative demo data. Figures and generated answers are not production-calculated.

Close calendar

FY2017 year-end · working days · CP = critical path · select a phase for its tasks

Phase

Phase detail

{{ ctSel.window }} · {{ ctSel.team }} · {{ ctSel.owner }}

Blockers to sign-off

Ranked by exposure · each links to where it is resolved

{{ b.owner }} · due {{ b.due }}

Team lanes

Illustrative close calendar. Late-rerun exposure is calculated live from the Scenario Lab cash-flow engine; other exposures come from the My Close queue.

Loading subledger balances, journals and accounting events…

Year ended 31 Dec

Scope

Threshold · {{ fxAbsLabel }} And at least · {{ fxPctLabel }}

Commentary agent

Flux by account

Account{{ fxPrevLbl }}{{ fxCurLbl }}Change%Status

Line detail

{{ fxSel.gl }} · {{ fxSel.cat }}

Change

{{ fxSel.d }} · {{ fxSel.pct }}

Commentary{{ fxSel.textNote }}

No commentary yet. Draft this line, or run the agent across every line above threshold.

What moved it · {{ fxSel.postedLbl }}

{{ x.n }} journal lines · {{ x.m }}

By portfolio

{{ g.g }} · {{ g.v }}

Journal lines · largest {{ fxJrN }}

DateTypeD/CMYRMeasure · group

{{ r.dt }}{{ r.et }}{{ r.dc }}{{ r.amt }}{{ r.sm }} · {{ r.icg }}

Balances and journals are the SAS IFRS17 subledger sample (SL_BALANCE, SL_JOURNAL) converted at 4.95 to MYR. The subledger carries P&L accounts cumulatively, so P&L lines compare year activity; balance-sheet lines compare closing balances. Commentary is drafted from the posted journals and needs controller review.

Loading subledger balances, journals and accounting events…

Financial year

Month · {{ mpMoLabel }}

View

Income statement · {{ mpTitle }}

MYR · income positive · green is favourable

LineActualBudgetVariancePrior yearFY forecast

Budget assumptions · live what-if

Revenue growth on prior year · {{ mpGLabel }} Claims inflation · {{ mpCiLabel }} Acquisition expense inflation · {{ mpEiLabel }}

CommentaryRedrafts as you move the assumptions

Insurance service result by month

Bars actual · marker budget · faded months not yet closed

By product line · {{ mpTitle }}

Product lineRevenueService resultvs budgetCSM

The subledger posts IFRS 17 results annually, so monthly actuals phase each account's year activity (revenue on coverage units, claims on the seasonal profile). Budget is prior-year activity flexed by the assumptions above and phased evenly. Product lines are display names for the 16 sample contract groups.

Loading subledger balances, journals and accounting events…

FY2017 external audit · today WD+3, Thu 4 Jan 2018

Ageing of open requests

{{ a.l }}{{ a.n }}

Evidence agent

Requests

Auditor request · raised {{ pbSel.raised }}

Matched evidence

Not matched yet. Run the evidence agent or match this request on its own.

Drafted response

Loading subledger balances, journals and accounting events…

Agent steps

Generated SQL

{{ alSql }}

Result · {{ alN }} rows

{{ b.l }}{{ b.v }}

Source rows

ASOFDATEAccount / measureContract groupD/CMYRType

Recent

Step {{ tourStepLabel }}

Contractual Service Margin

Risk Adjustment

Loss Component

Insurance Revenue

{{ dashDrillMetricLabel }} — {{ dashDrillLevelLabel }}

Most granular level — the movement drivers behind this balance. Click a bar to drill into the next level of aggregation.

CSM by year-end, with run-off

{{ csmChartLevelLabel }} — most granular level

Generative highlights

Proxy measures

•{{ h.text }}

Open review items

Contract group	Product line	Issue	Severity	Status
{{ r.id }}	{{ r.lob }}	{{ r.issue }}	{{ r.sev }}	{{ r.status }}

Use cases

No use case matches.

Insights

No insights match that filter.

{{ libView }} · Owner {{ libOwner }} · {{ libNum }}

Drill deeper

Driversi

Materialityi

Absolute · {{ libMat.absolute }}

Relative · {{ libMat.relative }}

Close impact · {{ libMat.closeImpact }}

Evidencei

Definition · {{ libEv.definition }}

Source · {{ libEv.table }} · {{ libEv.rows }}

Filters · {{ libEv.filters }}

Freshness · {{ libEv.freshness }}

Lineage · {{ libEv.lineage }}

Source detail

The reconciled line items behind this insight

{{ c }}
{{ cell }}

Outlooki

Next quarter

Uncertainty

Model

Trigger · {{ libForecast.trigger }}. {{ libForecast.reversal }}

Action

{{ libAction.priority }} priority {{ libAction.status }}

✓{{ x.text }}

Owner

Target date

Track

Illustrative demo data. Figures and generated answers are not production-calculated.

Opened from Intelligence Library · use case and insight follow that session.

1. Use case

2. Insight

Grounded to · {{ copilotFocusSummary }}

Ask a question

Locating the governed insight that answers this question...

Your question

Answered from {{ ipUseCase }} · {{ ipTitle }} · {{ ipNum }}

Grounded to · {{ ipScope }}

Reconciled visual

1 · Executive answer

2 · Root causes

3 · Materiality

Absolute · {{ ipMat.absolute }}

Relative · {{ ipMat.relative }}

Close impact · {{ ipMat.closeImpact }}

4 · Evidence availability

Definition · {{ ipEv.definition }}

Source · {{ ipEv.table }} · {{ ipEv.rows }}

Filters · {{ ipEv.filters }}

Freshness · {{ ipEv.freshness }}

Lineage · {{ ipEv.lineage }}

5 · Forward view

Next quarter

Uncertainty

Model

Trigger · {{ ipForecast.trigger }}. {{ ipForecast.reversal }}

6 · Recommended action

{{ ipAction.priority }} priority {{ ipAction.status }}

✓{{ x.text }}

Owner

Target date

Track

Illustrative demo data. Figures and generated answers are not production-calculated.

COPILOT

{{ c }}
{{ cell }}

Disclosure notes

No note matches that filter.

Compare against

{{ ncBaseLabel }} → {{ ncDraftLabel }}

Removed Added

Figure tie-out

No figures cited in this narrative.

{{ selectedNote.name }} ({{ selectedNoteUnit }})

{{ c }}
{{ cell }}

Drafted narrative

Confidence: {{ selectedNote.confidence }}%

Technical Plain-language

✎ Edited by you — captured in the Governance & Audit Trail log.

Tracked changes since last draft

✎{{ tc }}

No edits since the last draft.

Scope · {{ actScopeLabel }} · {{ actScopeShareLabel }} · assumption set {{ actSetId }}

Calculated with

{{ k.label }}i

Assumption change register, re-measured on the workbook

The workbook carries no assumption changes. Each move in ACT_ASSUMPTION_CHANGE is applied to the contract-group cash flows (CALC-020). CSM, OCI and finance expense in MYR K.

Change	Assumption	Prior	Current	Move	CSM	OCI	Fin. exp.	Study	Approver	Status
{{ c.id }}	{{ c.name }}	{{ c.prior }}	{{ c.now }}	{{ c.move }}	{{ c.csm }}	{{ c.oci }}	{{ c.finexp }}	{{ c.study }}	{{ c.approver }}	{{ c.status }}
Total assumption impact					{{ actChangeCsmTotal }}	{{ actChangeOciTotal }}	{{ actChangeFinTotal }}

Assumption impact tree

{{ actProp.id }} · {{ actProp.scope }}

Coefficient {{ actProp.coefficient }} · {{ actProp.formula }}
Scope {{ actProp.shareLabel }} · study {{ actProp.study }}

→

Cohort impact attribution

Contract group	Cohort	CSM	Loss component	Risk adjustment	Share
{{ r.icg }}	{{ r.cohort }}	{{ r.csm }}	{{ r.lc }}	{{ r.ra }}	{{ r.shareLabel }}

Actuarial interpretation

Owner {{ actProp.owner }} · {{ actProp.decision }}. The proposal reduces future service profit and narrows the CSM buffer for the groups in scope; the loss component and risk adjustment absorb the remainder.

Recommended actions

•{{ a.text }}

Scenario levers

{{ s.coefficient }} · {{ s.formula }}

Outputs calculated with

Base {{ o.base }} · {{ o.move }}

Generated interpretation

Factor contribution and audit basis

Each line re-measures the scope's cash flows with that factor alone (CALC-020). Amounts in MYR K.

Factor	Shock	CSM	Loss component	OCI	Service result	Formula	Calibration study
{{ f.factor }}	{{ f.delta }}	{{ f.csm }}	{{ f.lc }}	{{ f.oci }}	{{ f.profit }}	{{ f.formula }}	{{ f.study }}

Scenario runs on file

Published to ACT_SCENARIO_RUN and ACT_SCENARIO_RESULT. Select a row to load its levers.

Run	Scenario	Scope	CSM impact	Run by	Run at	Posted	Status
{{ r.id }}	{{ r.name }}	{{ r.scope }}	{{ r.csm }}	{{ r.by }}	{{ r.at }}	{{ r.posted }}	{{ r.status }}

Onerous radar

How far each contract group is from turning onerous

Across: claims severity increase that uses up the group's CSM. Up: CSM size. Bubbles left of the red line flip under the shock you set.

Claims severity shock{{ actOnShockLabel }}

{{ actOnShockLabel }} shock

Onerous today Flips under this shock Within 10% headroom Comfortable {{ actOnerousCount }}

Watch list · nearest to the edge

Select a group for its tipping point and drivers

{{ actOnSel.gaugeLabel }}headroom

CSM buffer

Loss component

CSM trend

What would tip it

Share of groups onerous, product line × cohort

Click a cell to scope the radar to it.

Workbook discount curve · {{ actCurveName }}

{{ actCurveNote }} · {{ actCurveAvgShift }}

■ Approved curve ■ Scenario curve

{{ m.k }}
{{ m.v }}

Base and scenario rates

Tenor	Approved	Scenario	Shift
{{ r.tenor }}	{{ r.base }}	{{ r.scen }}	{{ r.shift }}

Reverse stress objectivei

Solve backwards from a management boundary

Target metric Boundary — {{ actBoundary }}% of the metric in scope

Solver configuration

Minimum assumption boundary

Solved live on the workbook cash flows (CALC-020). Achieved movement {{ actSolverAchieved }}.

Factor	Base	Required shock	Boundary level	Formula
{{ r.factor }}	{{ r.base }}	{{ r.shock }}	{{ r.level }}	{{ r.formula }}

{{ m.k }}
{{ m.v }}

Management interpretation

Reverse stress tests on file

Published to ACT_REVERSE_STRESS and carried into the governance audit trail.

Test	Target	Boundary	Scope	Solver	Run by	Run at	Status
{{ t.id }}	{{ t.target }}	{{ t.boundary }}	{{ t.scope }}	{{ t.solver }}	{{ t.by }}	{{ t.at }}	{{ t.status }}

Evidence and lineage

Insights

No insight matches that filter.

Source connector (Claims Data Feed) is degraded — this insight reflects data as of 2 days ago, not the current close.

Dashboard view

AI is analyzing this insight...

{{ c }}
{{ cell.text }}

COPILOT

{{ c }}
{{ cell }}

Insight summary

Confidence: {{ reconSelectedInsight.confidence }}%

•{{ b }}

Recommended actions

✓{{ a }}

Source: {{ reconSelectedInsight.source }}

Three-way auto-reconciliation

Engine to subledger to general ledger

Every accounting event is matched to its journal postings, and every journal position to its GL balance, by contract group. Breaks are ranked by MYR impact and given a likely cause.

As at

Loading ACCOUNTING_EVENT, IFRS17_SUBLEDGER_JOURNAL and IFRS17_SUBLEDGER_BALANCE…

Rows

Value

Break heatmap by contract group

Click a cell to filter the break list

Contract group

Engine → Subledger

Subledger → GL

Impact

Breaks by likely cause

Breaks ranked by MYR impact

Why this broke

Suggested action

{{ c }}
{{ cell }}

No breaks match these filters.

Formula Registry

Every calculation behind the studio in one place. Engine formulas come from the SAS IFRS 17 formula library; application logic lists each calculation the screens perform, with where it is used. Source files: data/formulas.json, data/posting.json, data/formula_registry.json.

Loading formula library…

Could not load the registry files. Serve the project over HTTP (see README.md) so data/*.json can be fetched.

{{ frMatchCount }} of {{ frTotal }} formulas · {{ frPostedCount }} reach the journal

ID	Variable	Approach	Area	Order	Deps	Journal
{{ r.id }}	{{ r.variable }}	{{ r.approach }}	{{ r.area }}	{{ r.order }}	{{ r.depCount }}	{{ r.postedLabel }}

{{ frSelected.variable }}

Approach

Area

Order

Calc type

Ceded

Transition

Expression

{{ frSelected.formula }}

Dependency graph · green = posts to journal

Upstream ({{ frSelected.upCount }})

Input measure

→

{{ frSelected.variable }}

→

Downstream ({{ frSelected.downCount }})

Terminal output

ƒ {{ frAppSelected.id }} {{ frAppSelected.status }}

Expression

{{ frAppSelected.expression }}

Inputs

Output

Implemented in

{{ frAppSelected.function }}

Data Hub tables

Engine formulas

Used on

ID	Name	Category	Status
{{ e.id }}	{{ e.name }}	{{ e.category }}	{{ e.status }}

Walk a single contract group from input measures through to its disclosure note, or review the audit trail below.

Pending Approvals

Assumption Changes (MTD)

Unreviewed AI Narratives

Data Quality Score

{{ k.label }}i

Assumption Change Log

The same ACT_ASSUMPTION_CHANGE rows the Assumptions workbench runs on. The CSM column foots to the MYR {{ govChangeCsmTotal }}K assumption bar in the Executive Update CSM bridge.

Change	Date	Assumption	Prior Value	New Value	CSM	OCI	Study	Changed By	Approved By	Status
{{ c.id }}	{{ c.date }}	{{ c.name }}	{{ c.prior }}	{{ c.now }}	{{ c.csm }}	{{ c.oci }}	{{ c.study }}	{{ c.by }}	{{ c.approver }}	{{ c.status }}
Total assumption impact					{{ govChangeCsmTotal }}

Connector Status & Data Freshness

Ties every insight in this app back to its source system's last successful sync.

Data Source	Status	As-of Date	Freshness
SAS IFRS17 Results (CAS Tables)	● Connected	Jun 30, 2026	Fresh
General Ledger / Sub-ledger Feed	● Connected	Jun 30, 2026	Fresh
Actuarial Cash Flow Engine	● Connected	Jun 30, 2026	Fresh
Assumption Governance Log	● Connected	Jun 30, 2026	Fresh
Claims Data Feed	● Degraded	Jun 28, 2026	Stale — 2 days
Market Data (Discount Curve)	● Connected	Aug 5, 2026	Fresh

Approval / Sign-off Workflow

Item	Stage	Owner	Status	Due
FY2017 Assumption Package	Actuarial Sign-off	R. Mehta	Approved	Aug 10
FY2017 CSM Roll-forward	Finance Review	S. Tan	Approved	Aug 12
Motor-2023 Experience Variance	Actuarial Sign-off	A. Lim	Pending	Aug 14
Board Pack — FY2017	Audit Committee Review	Audit Committee	Pending	Aug 18
Disclosure Note 12 Draft	CFO Approval	CFO Office	Not Started	Aug 20
{{ a.item }}	{{ a.stage }}	{{ a.owner }}	{{ a.status }}	{{ a.due }}

Data Lineage

Source	Transformation	Output
SAS IFRS17 Results (CAS Tables)	GenAI & Agentic AI Layer	Insights, Narratives & Disclosures
General Ledger / Sub-ledger	Reconciliation Engine	Reconciliation Status & Breaks
Actuarial Cash Flow Engine	CSM / RA Roll-forward Models	CSM, RA & Loss Component Balances
Assumption Governance Log	Assumption Impact Models	Assumption Analysis & Stress Testing
{{ l.source }}	{{ l.transform }}	{{ l.output }}

Assumption & Scenario Audit Trail

Every action taken in the Assumptions workbench — saved scenarios, applied curve scenarios and reverse stress tests — with the object it touched. Actions taken in this session are tagged.

Timestamp	User	Role	Action	Object	Record
{{ a.at }}	{{ a.user }}	{{ a.role }}	{{ a.action }}	{{ a.ref }}	{{ a.tagLabel }}

AI Narrative Audit Trail

Narrative	Module	Generated On	Status
Disclosure Note 12 draft	Disclosures	Aug 5, 09:14	Pending Review
Executive KPI Narrative	Use Cases	Aug 5, 08:02	Reviewed — S. Tan
CSM Movement Explanation	Use Cases	Aug 4, 16:40	Reviewed — R. Mehta
Reconciliation Executive Narrative	Reconciliation	Aug 5, 07:55	Pending Review
{{ el }}	Disclosures	Just now	Manually Edited

Access & Activity Log

User	Role	Action	Timestamp
R. Mehta	Group Actuarial	Approved FY2017 assumption package	Aug 5, 09:20
S. Tan	Finance	Reviewed CSM roll-forward	Aug 5, 08:45
A. Lim	Actuarial Analyst	Updated morbidity assumption	Jun 14, 14:02
Audit Committee	Auditor	Viewed Board Pack draft	Aug 4, 11:15
CFO Office	Executive	Viewed Executive KPI Dashboard	Aug 5, 07:30

Signed run Run 04 · approved WD+1, 2 Jan 2018 Reported basis for the FY2017 close

Late reruns received

Overall materiality

Net effect on CSM less loss component: {{ lrNet }}{{ lrNetPct }}

Clearly trivial {{ lrTriv }} Materiality {{ lrThr }}

{{ k.label }} {{ k.base }} → {{ k.now }} {{ k.delta }}

Close timeline

What the rerun touches

Impact by product line

CSM less loss component · MYR K · green is favourable

No product line moves by more than MYR 0.1K.

Contract groups moving most

Contract group	Portfolio	CSM Run 04 → rerun	Net effect
{{ g.disp }}Turns onerous {{ g.icg }}	{{ g.pf }}	{{ g.csm }}	{{ g.d }}

Preset Scope

Leversi

{{ L.label }}i {{ L.valueLabel }}

Cash-flow driven

Cash flows: real CSM / LC: estimate

{{ h.label }}i {{ h.value }} {{ h.delta }}

Levers & impact · as at {{ scnPeriod }}i

{{ r.label }} {{ r.deltaLabel }} · {{ r.pctLabel }}

Base {{ r.baseLabel }} → scenario {{ r.nowLabel }}

Impact by metrici

MYR K · green is favourable

Root-cause attributioni

Each lever re-run on its own · MYR K

All levers are at base. Move a lever to see which one drives the effect.

CSM glide pathi

Trace any group back to its workbook rows

Contract group	Portfolio	Model	CSM	Loss component	Net effect
{{ g.disp }} Turns onerous {{ g.icg }}	{{ g.pf }}	{{ g.appr }}	{{ g.csm }}	{{ g.lc }}	{{ g.d }}

Readingi

Cash flows, discount curves and contract-group attributes are real rows from the SAS WW_INSURANCE sample workbook, shown as MYR with HLA-style portfolio names. CSM and loss component are estimates until SAS IFRS17 engine output is loaded. The discount shock is an exact re-discount; lapse and severity are approximations applied to base cash flows until Prophet sensitivity runs are available.

Evidence packages

Illustrative demo data. Figures and generated answers are not production-calculated.

Last review {{ evSel.reviewed }} by {{ evSel.reviewer }} · {{ evSel.items }} evidence items

Confidence, decomposedi

Metric	Owner	Validation
{{ r.name }} {{ r.code }} · {{ r.usedIn }}	{{ r.owner }}	{{ r.validation }}

No metrics match that search.

{{ regSel.code }} {{ regSel.validation }} v{{ regSel.version }}

Business definition

Source measures

Source table

Owner

Last reviewed

Used in

Illustrative demo data. Figures and generated answers are not production-calculated.

Input → measures → events → journal → GL → disclosurei

End-to-end traceability

Pick a contract group and walk every hop from the input measures the engine calculated to the disclosure note the balance lands in. Each hop shows the real keys and row counts for that group.

Loading the traceability index...

{{ trGroup.chainLabel }} Projects: {{ trGroup.projectList }} Source currency: {{ trGroup.currencyList }} As at 31 Dec 2017

Click a stage to collapse or expand it. Click a record to open its table in Data Hub. Scroll to zoom, drag to pan.

Workbook lineage · SAS WW_INSURANCE samplei

Contract group

{{ twHead.pf }} Source portfolio: {{ twHead.src }} Model: {{ twHead.appr }} As at {{ twHead.period }}

Loading workbook tables: INSURANCE_CASHFLOW, GL_ACCOUNT_BALANCE_SEGMENT, INSURANCE_CONTRACT_GROUP, INSURANCE_CONTRACT_GROUP_DETL, QUOTES...

{{ s.label }}: {{ s.value }}

{{ c }}
{{ cell }}

Engine extract lineage · SAS IFRS17 test entityi

Insurance contract group

{{ trGroup.chainLabel }} Projects: {{ trGroup.projectList }} Source currency: {{ trGroup.currencyList }}

Loading the traceability index across ALL_MEASURES, ACCOUNTING_EVENT, IFRS17_SUBLEDGER_JOURNAL and IFRS17_SUBLEDGER_BALANCE...

{{ s.label }}: {{ s.value }}

{{ c }}
{{ r.key }}	{{ r.value }}

Chain integrity across all contract groups

Which of the 16 posted contract groups can be walked all the way from input measures to a disclosure note.

Contract group	Measure rows	Events	Journal lines	Balance rows	Chain
	{{ c.measRows }}	{{ c.events }}	{{ c.lines }}	{{ c.balRows }}	{{ c.chainLabel }}

Ask for a lineage

Name a contract group and ask for its lineage. The answer is composed from the same index the hops above read, so it cannot drift from the data.

LINEAGE COPILOT

{{ c }}
{{ cell }}

Columns: {{ datahubTable.columns.length }}

Rows: {{ datahubTable.rows }}

Size: {{ datahubTable.size }}

Updated: {{ datahubTable.updated }}

Demo data

Loading extract…

#	Name	Data Type	Description
{{ c.n }}	{{ c.name }}	{{ c.type }}	{{ c.desc }}

Showing {{ datahubSampleRows.length }} of {{ datahubTable.rows }} rows

{{ col }}
{{ cell.val }}

Generative AI Insight — {{ datahubTable.name }}

Thinking…

Guided Workflow

Follow these steps to stand up retrieval-augmented generation over your enterprise documents. Each step unlocks once the previous one completes.

Completed workflow runs

Collection	Documents	Embedding Model	Destination	Status	Updated
{{ rc.name }}	{{ rc.docsCount }}	{{ rc.embedModel }}	{{ rc.destination }}	✓ {{ rc.status }}	{{ rc.updated }}

Previously Created Workflows

Agent / Workflow	Collection	LLM	Destination	Deployed	Status
{{ ag.name }}	{{ ag.collection }}	{{ ag.llm }}	{{ ag.destination }}	{{ ag.deployedAt }}	✓ {{ ag.status }}

Step 1 · Upload Enterprise Documents, Videos & Tutorials

Audience for this upload

Drag & drop files, or click to browse

Documents: .pdf .docx .txt .csv .xlsx .html .json .md .pptx · Video & tutorials: .mp4 .mov .webm

Step 2 · Organize & Process Documents

Embedding model

Destination (vector store)

Chunking, embedding and writing vectors to the destination store…

Collection

Vectors

{{ ramCollection.rowCount }} chunks

Embeddings size

Status

✓ {{ ramCollection.status }}

Step 3 · Run Evaluation to Choose the Best Retrieval Settings

Large language model

Retrieval top-K: {{ ramTopK }}

Running retrieval evaluation across the collection…

Recall@{{ ramTopK }}

{{ ramEvalRecall }}%

Overall score

{{ ramEvalScore }}%

🏆 Champion configuration

Step 4 · Interact with Documents via Chat

Ask a question grounded in your uploaded document(s) — e.g. "How are onerous contracts treated?"

Source: {{ c.source }}

Retrieving relevant passages…

Step 5 · Develop Agents to Automate Complex Workflows

Agent name

Tools

search_docs — retrieve grounded passages summarize — condense long documents compare_clauses — diff policy vs. regulator text generate_chart — visualize retrieved metrics

LLM: {{ ramAgent.llm }} · Collection: {{ ramAgent.collection }}

✓ Workflow complete — your enterprise knowledge base is live. Explore it via the Chat, Collections and Agents tabs above.

Dashboard

Read-only · authorized users see queries made by all users

Type

Show

Period

▮ {{ ch.barLabel }}— {{ ch.lineLabel }}

User Feedback %

▮ Positive▮ Negative▮ No response

Positive feedback climbed thirteen points as the champion configuration moved to overlapping chunks. The August dip tracks the orchestrator's failed delegations.

Filters apply to Entity Name, Entity Type and Feedback

Entity Name	Type	Query	Response	Start Time	Duration	Cost	Feedback
{{ q.entity }}	{{ q.type }}	{{ q.query }}	{{ q.response }}	{{ q.start }}	{{ q.duration }}	{{ q.cost }}	{{ q.feedback }}

Licensed data volume

{{ ramStorUsed }} of {{ ramStorLicensed }} · {{ ramStorPct }}

Embedded Models Allocation

{{ ramStorEmbedUsed }} of {{ ramStorEmbedAlloc }} · {{ ramStorEmbedPct }}

Source Files Allocation

{{ ramStorFilesUsed }} of {{ ramStorFilesAlloc }} · {{ ramStorFilesPct }}

Embedding Model	Used By	Disk Space
{{ m.name }}	{{ m.configs }}	{{ m.size }}

Source	Files	Disk Space
{{ m.name }}	{{ m.files }}	{{ m.size }}

Big 4 Technical Guidance is the largest source at 403 GiB across four PDFs — the OCR and table extraction on those handbooks accounts for most of it. It is also the least authoritative content in the corpus.

Start {{ ramQDetailStart }}

Duration {{ ramQDetailDuration }}

Cost {{ ramQDetailCost }}

Query submitted

System prompt generated

Response

Knowledge Base Data

Configuration	Destination	Embedding Model	Row Count	Embeddings Size	Status
{{ rc.name }}	{{ rc.destination }}	{{ rc.embedModel }}	{{ rc.rowCount }}	{{ rc.embeddingsSize }}	{{ rc.status }}

No collection yet — complete Steps 1–2 of the Workflow to see it here.

Average Retrieval Time by Model

Sources

Sources provide the raw materials for vectorization and search. A source cannot be removed while a collection is using it.

Name	Type	File Update Schedule	Files	Associated Collections	Last Updated
{{ src.name }}	{{ src.type }}	{{ src.schedule }}	{{ src.fileCount }}	{{ src.collections }}	{{ src.updated }}

In use — {{ ramSelSourceCollections }}

STORE_FULL_TEXT is off. Document metadata, issue and effective dates, amendment inventory and machine-written summaries are vectorized; verbatim IFRS 17 paragraph text is not stored, because it is IFRS Foundation copyright. Retrieval therefore cites and characterises the standard rather than reproducing it.

Type

Cron expression

Last updated

Associated collections

Custom source template

Repository URL

Document Name	Size	Tags
{{ f.name }}	{{ f.size }}	{{ tg.label }} —

Supported file types: {{ ramSupportedExt }}

Selected file

Ingested summary

Retrieved from

Timestamp	Trigger	Status	Job Log
{{ j.ts }}	{{ j.trigger }}	{{ j.status }}	{{ j.note }}

LLM Alias	Retrieval Settings
{{ a.alias }}	{{ a.retrievalSetting }}

Configuration	Vectorized Status	Last Job	Last Auto Eval	Last User Eval	Configuration Status
{{ cf.name }} {{ cf.embedModel }} · {{ cf.chunks }} chunks · {{ cf.size }}	{{ cf.vectorized }}	{{ cf.lastJob }}	{{ cf.autoEval }}	{{ cf.userEval }}	{{ cf.statusIcon }} {{ cf.status }}

#	Model	Retrieval Settings	Visibility in Chat
{{ l.order }}	{{ l.name }} {{ l.provider }}	{{ l.retrievalSetting }}	{{ l.visibility }}

Question	Context	Ground Truth	Generator LLM	Cost
{{ d.q }}	{{ d.ctx }}	{{ d.gt }}	{{ d.gen }}	{{ d.cost }}

Timestamp	Configuration	Eval LLM	Critic LLM	Critic Embedding	Status	Ragas Score	Cost	Details
{{ r.ts }}	{{ r.config }}	{{ r.evalLLM }}	{{ r.critic }}	{{ r.criticEmbed }}	{{ r.status }}	{{ r.ragas }}	{{ r.cost }}

Timestamp	Test Case	Eval LLM	Configuration	Status	Score	Cost
{{ r.ts }}	{{ r.test }}	{{ r.evalLLM }}	{{ r.config }}	{{ r.status }}	{{ r.score }}	{{ r.cost }}

Tags

A tagged query retrieves only from sources carrying that tag.

Previous chats

Response Details

Total duration {{ ramTraceTotalDuration }}

Total cost {{ ramTraceTotalCost }}

Every action taken to produce this answer, in order, indented by depth. Colour is the action type: query, LLM call, retrieval, MCP tool, delegated agent.

LLM

Provider

Model

Agent called

Experiment used

Tool server

Tool

Input supplied by the agent

{{ ramNodeInput }}

Output returned by the tool

{{ ramNodeOutput }}

Agentic retrieval decision

Queried against

Configuration size

Retrieval prompt

Documents retrieved from

System prompt

Prompt

Response

{{ ramNodeResponse }}

Response Duration

Response Cost

Retrieval is {{ nf.NT018 }} of the elapsed time but almost none of the cost. The cost sits in the completion tokens across four LLM calls — which is what agentic retrieval reduces, by not searching collections that cannot hold the answer.

Documents

No documents uploaded yet — use the Workflow tab.

Chatting with: {{ ramChatSelectedDocName }}

Ask a question grounded in your uploaded document(s).

Source: {{ c.source }}

Retrieving relevant passages…

Agents

Agents extend retrieval with custom workflows and tools. Orchestrator agents decompose a query and delegate to the agents whose A2A cards claim each task. An agent needs a champion experiment before it can be started.

Name	Type	Experiments	Status	Last Modified
{{ a.name }} {{ a.desc }}	{{ a.kind }}	{{ a.experiments }}	● {{ a.status }}	{{ a.lastModified }}

{{ ramSelAgentKind }} ● {{ ramSelAgentStatus }} Chat initializer: {{ ramSelAgentInitializer }}

Experiment	Type	LLM	Retrieval Setting	Agentic Retrieval	Footprint	Last User Evaluation	Status
{{ e.name }}	{{ e.type }}	{{ e.llm }}	{{ e.retrievalSetting }}	{{ e.agentic }}	{{ e.footprint }}	{{ e.userEval }}	{{ e.statusIcon }} {{ e.status }}

System prompt

A system prompt set on an experiment overrides the agent-level prompt and the prompt in Retrieval Settings.

LLM

Retrieval setting

Virtual CPUs

Container memory

Tools

No tools assigned — retrieval only.

Delegates to

Only agents with a configured A2A agent card can be delegated to.

Collections

Agentic retrieval

Enabled

Disabled

Agent description

The orchestrator reads this card to decide what this agent is for. A vague description is the most common cause of a task being routed to the wrong agent.

Skills

Query examples

“{{ x.text }}”

Orchestrator agents do not carry an A2A card — they read the cards of the agents they delegate to. See the Delegates to list on the Summary tab.

No A2A card configured. Until one exists this agent cannot be delegated to by any orchestrator.

Experiment	LLM	Last User Evaluation	Cost per run	Status
{{ e.name }}	{{ e.llm }}	{{ e.userEval }}	{{ e.cost }}	{{ e.statusIcon }} {{ e.status }}

Compare the scores here to decide the champion. Where two experiments score equally, cost per run decides — which is why the mini model holds champion on the Regulator Watch Agent at a fifth of the cost.

User Eval Tests

Reusable test scenarios, not tied to any collection or agent. Each prompt carries assertions and a passing threshold — the percentage of assertions that must pass for the response to count as correct.

Name	Prompts	Passing Threshold	Updated At
{{ t.name }}	{{ t.promptCount }}	{{ t.threshold }}	{{ t.updated }}

Passing threshold {{ ramSelTestThreshold }}%

Test prompts

Prompt

Passing threshold (%)

Assertions

Metric	Value
{{ a.metric }}	{{ a.value }}

A good practice is to include safeguard prompts — a question the documents cannot answer, asserting that the model replies that it does not know rather than inventing a figure.

Automation

Connect sources, configurations and agents into scheduled workflows that run without intervention. A connection from a source to a configuration re-vectorizes on source update; a connection from a configuration to an agent re-initializes the agent when vectorization finishes.

Configured pipelines

Available pipelines

Configured

No connections — drag between connection points to build the pipeline

Sources

Configurations

Agents

No agent is configured against this collection, so the pipeline ends at vectorization.

If you connect only one source to a configuration, every source in the collection is re-vectorized when that source updates. Use the update strategy to avoid duplicate embeddings.

Automation Settings

Resources for automated vectorization jobs only. Manual vectorization keeps its own settings.

CPU

Memory

Execution provider

OpenVINO device

Code Templates

Predefined blueprints for the components developers extend. A template must be published before it appears when adding an agent, source or tool server. Versions are immutable — saving changes creates a new one.

Name	Type	Latest Version	State	Used By
{{ t.name }}	{{ t.kind }}	{{ t.version }}	Published Unpublished	{{ t.usedBy }}

Viewing a superseded version — read only

Version history

Template type

Version shown

LLM aliases

Enabled

Consumed by

{{ ramSelTplCode }}

Custom source and agent templates must define exec(). Credentials, endpoints and file paths belong in environment variables, never in the code — a hard-coded key here becomes a key in every source built from the template.

Name	Description	Default Value	Secret
{{ e.name }}	{{ e.desc }}	{{ e.value }}	{{ e.secretLabel }}

LLM Alias	Description
{{ a.alias }}	{{ a.desc }}

Aliases let the code name models by role rather than by ID, so a model swap is a mapping change on the experiment rather than a code change here.

This template calls no language model, so it declares no aliases. Tool server templates expose functions to an agent; the agent's own experiment decides which model calls them.

Application Settings

Administrator role required

Name	Type	URI	Connection	In Use
{{ d.name }}	{{ d.type }}	{{ d.uri }}	● Connected ● Failed	{{ d.configs }}

Connection failed — the host is unreachable from the cluster

URI

Port

Database

Schema

Create if not exist

Deployment type

gRPC port

Name	Provider	Dimension	Max Tokens	State	In Use
{{ m.name }} {{ m.modelId }}	{{ m.provider }}	{{ m.dim }}	{{ m.maxTokens }}	{{ m.status }}	{{ m.configs }}

Model ID

Dimension

Max tokens

Deprecated. Existing configurations continue to consume it, but no new configuration can be created with it. Deprecation cannot be undone.

Dimension and max tokens are inferred when the model is verified. Verify them — an incorrect inferred value means the model silently fails to embed.

Name	Provider	Deployment	Connection	Token Cost	Groups with Access
{{ l.name }}	{{ l.provider }}	{{ l.deployment }}	● Healthy ● Failed	{{ l.cost }}	{{ l.groups }}

Connection failed — check the API key and that egress to this endpoint is permitted

Provider

Deployment name

URI

API version

API key

••••••••••••

Metered cost

Groups with access

A user can chat with a collection only if they belong to the User group, the collection's own group, and a group with access to the model. All three, or the collection simply does not appear for them.

Admin group

sas_ram_admin_role on client sas-ram-app. Full read, create, update and delete across every view.

User group

sas_ram_user_role on client sas-ram-app. Chat only, and only with collections they have been granted.

Per-collection groups

Created automatically with each collection — IFRS17KnowledgeBase, RegulatoryWatch, HomeIntelligenceFeed.

Task	Administrator	User
{{ r.task }}	{{ r.admin }}	{{ r.user }}

Users are managed in the Keycloak realm created at deployment. Administrators hold these permissions by default; users must be granted theirs explicitly by group membership.

Connect only to tool servers from secure, trusted sources. An untrusted server can expose the environment to data exfiltration, adversarial content injection or unauthorized data access. Every connection here must be vetted under internal security policy before it reaches a production agent.

MCP Tool Servers

Tool servers host the MCP tools an agent can call. Starting a server makes its tools selectable on the Add Agent and Edit Agent windows; stopping it makes them unavailable.

Name	Code Template	Source	Server Status	Ping Status	Tools
{{ sv.name }}	{{ sv.template }}	{{ sv.source }}	● {{ sv.status }}	{{ sv.ping }} {{ sv.ping }} {{ sv.ping }}	{{ sv.toolCount }}

● {{ ramSelSrvStatus }} {{ ramSelSrvTemplate }}

This is a third-party server outside the cluster. SAS does not vet the security, reliability or behaviour of third-party MCP servers — evaluating and approving it is the customer's responsibility, including the risk of data exfiltration and adversarial content injection.

Transport type

Server URL

Authentication

OpenAPI specification

Database vendor

Started

Tools used by

Environment variables

Name	Value	Secret
{{ e.name }}	{{ e.value }}	{{ e.secretLabel }}

Tools

Parameter	Field Type	Required	Default
{{ p.name }}	{{ p.type }}	{{ p.requiredLabel }}	{{ p.def }}

The tool list is populated only while the server is running. Start the server to discover its tools and make them selectable when creating or editing an agent.

Tool Server Logs

Retrieval Agent Manager Admin — Generative AI for Enterprise Knowledge Base

Dashboard

ING data

Configuration	Destination Type	Embedding Model	Row Count	Rows Size	Embeddings Size	Embedding Dim.
{{ r.config }}	{{ r.dest }}	{{ r.embed }}	{{ r.rows }}	{{ r.rowsSize }}	{{ r.embSize }}	{{ r.dim }}

Average Retrieval Time

Sources

Name	Type	Schedule	Associated Collections	Last Updated
{{ r.name }}	{{ r.type }}	{{ r.schedule }}	{{ r.collections }}	{{ r.updated }}

Name	Size	Last Updated
{{ f.name }}	{{ f.size }}	{{ f.updated }}

Job	Status	Started
{{ j.name }}	{{ j.status }}	{{ j.started }}

All Collections > Financial Docs

Configuration	Vectorization Status	Last Job	Last Auto Eval	Last User Eval	Configuration Status
{{ r.name }}	✓ {{ r.status }}	{{ r.lastJob }}	{{ r.evalPct }}%	✓ {{ r.userEval }}	🏆 Champion {{ r.configStatus }}

⚙ Basics

Embedding model

all-MiniLM-L6-v2

Destination

weaviate storage local

Update strategy

Append, sync, and delete

🔧 Settings

Document ingestion

Exclude extensions from processing

.pdf

☑ Clean text after ingestion

🔍 Text Extraction

☐ OCR enabled

☐ Export tables

Collections

New Collection

1 Source ⌄

Chat

Large language model

gpt-4.1 (Concise)

what products does Apple sell?

Apple sells the following products:

Smartphones:

iPhone 16 Pro
iPhone 16
iPhone 15
iPhone 14
iPhone SE

Personal Computers (Mac):

MacBook Air
MacBook Pro
iMac
Mac mini
Mac Studio
Mac Pro

MCP Tool Servers

Name	Template	Server Status	Ping Status
{{ r.name }}	{{ r.template }}	● {{ r.status }}	{{ r.ping }}

Tools — {{ adminSelectedMcp }}

Name	Description
› {{ f.name }}	{{ f.desc }}

Agents

Name	Version	Status
{{ a.name }}	{{ a.version }}	● {{ a.status }}

About

Name

Description

LLMs

default

Collections

default

Tools

default

This section mirrors the Sources view's data — switch here once you've configured it from the Workflow tab.

IFRS 17AI Studio

Welcome to SAS Viya

Productivity and Speed

Experience SAS Viya Copilot

Data & AI Life Cycle Quick Start

Solutions

Sign in

You have signed out.

IFRS 17
AI Studio