KEZEL

solutions · K·03 · analytics module

Dashboards, queries, forecasts and audit, in one place.

Every surface below runs over governed views inside your boundary. Ask, forecast, rank, mask, trace: each panel is the product doing its job on sample data, in the same shape it takes in a deployment.

fig. 01the workbench

The working surfaces.

Live views over the system of record

The agent compiles the plan; SQL pushes down to engines that speak it, and the rest runs in the agent beside the source. What you chart is the table as it stands.

? churn by region, this quarter
→ plan · 3 tables · masked · approved

Plain language to a governed plan

LLMCore reads schema, not rows: it is built to plan without raw-row access. The plan names its tables, its masks and the policy that cleared it.

Ranked work, explained

Scores with reasons attached, so reviewers see why an item is at the top of the queue.

readeremailcard_no
analysta•••@corp.eu•••• 4412
fraud-rev[email protected]•••• 4412

Column rules, applied in the query

Row filters and column masks are part of the plan itself, scoped to the reader's role. Cleartext reads are themselves logged.

orders_eufx_ratesweekly_revtile 07

Lineage for every figure

Each published figure traces back through its transforms to its sources, as a graph.

10:42nl-querymasked viewj.rao
10:44forecastaggregateplanning
10:47exportdenied · policysvc-crm

Every read is logged

Question, rule, outcome, reader. One row per read, kept for the auditor.

fig. 02KEZEL Insights Studio

How much, which ones, when.

Introduced at Bengaluru Tech Summit 2025. Each question below has its own engine, shown here on demonstration data.

How much?

Price a decision before you make it

Fit the relationship, then move the lever and watch the outcome respond before you commit to it. Simulations recompute against the boundary's own tables.

simulate · compare · commit

campaign spend vs. weekly volume · fitted · sample data

VizEngine · dashboards · LLMCore · plain-language plans · time series · Prophet, ARIMA · classification · 8 model candidates, evaluated with and without outlier correction

fig. 03collaboration

Third parties work inside your boundary.

Suppliers, partners and service providers read governed views where the data lives. No records pool on a second platform, and every read they make lands on the same ledger yours do.

fig. 03 · readers on governed views

readerscopeoutcome
partner-logisticsfill_rate_weekly · governed viewreads logged
auditorledger + lineageread-only
svc-crmexport requestdenied · policy

governed views · in-boundary reads through the policy gate · sample data

Bring a real workload. We’ll answer it in place.

Start with a demo on our sample boundary. Move to a pilot in yours once your security review clears it.