Agents Makers
Capability of Data AnalystDefault at launch

NL-to-SQL

Translates business questions into validated SQL against the semantic model.

  • Activation complexity

    Medium

  • Time to activate

    10-14 days

  • Volume share

    35-45% of role volume

  • Impact range

    Sub-minute on routine; under an hour on multi-source

Inherited pricing

€0.80 – €3.50 per query or report handled

This capability inherits the Data Analyst's pricing model. The role's launch fee + monthly retainer + role-level usage cover every capability under the role. Adding this capability to an active deployment does not change the price.

What this capability handles

How it works in detail.

NL-to-SQL turns a plain business question into a trustworthy answer without an analyst writing the query by hand. Mid-market data teams lose the week to ad-hoc requests: a stakeholder asks for a number, the question queues, and the answer arrives stale. This capability serves the operators who keep asking and the analysts who keep getting interrupted. The result is a fast, sourced answer that people can act on and trace back to its origin. It works through a clear sequence. First it reads the question. Then it resolves the metric through your semantic model, so a request for revenue or churn maps to the one agreed definition rather than a guess. It drafts SQL against the governed warehouse, then runs a validation pass to sanity-check the query before anything is trusted. Finally it returns the answer with lineage attached, so the requester sees not just the number but how it was produced. It runs inside your existing warehouse, semantic layer, BI tool, and messaging. It draws on the metric dictionary, the semantic model, prior queries, and the schema catalog, and what it produces per request is one validated, sourced answer. The decision logic is built on semantic-model matching and metric-dictionary rules. When a request maps cleanly to approved definitions, it drafts validated SQL and returns the result. When it does not, it flags the request rather than improvising. Novel metrics that have no agreed definition, cross-domain joins where the meaning is ambiguous, and PII-adjacent cuts all route to the analyst for review with the reasoning attached. Every query and its lineage are logged and reviewable, so the team can audit what was asked, what ran, and what came back. This fits teams whose semantic model is documented, whose metric dictionary is approved, and whose warehouse access is wired. Ad-hoc question throughput with query-lineage fidelity accounts for 35-45% of the role's impact, and this capability handles 35-45% of role volume. Turnaround is the measure that matters: sub-minute on routine questions, and under an hour on multi-source ones. That is the difference between answers that drive a decision and answers that arrive too late to use.

Workflow summary

Reads question, resolves metric, drafts SQL, validates, returns with lineage.

Stages

  1. 01read
  2. 02resolve
  3. 03draft
  4. 04validate
  5. 05return

Decision logic

Uses semantic-model matching and metric-dictionary rules to draft validated SQL and flag novel or ambiguous requests.

Systems and data

{warehouse,"semantic layer","BI tool",messaging}

{"metric dictionary","semantic model","prior queries","schema catalog"}

Exceptions & human handoff

Novel metrics, cross-domain joins, or PII-adjacent cuts route to the analyst for review.

Novel metric, cross-domain ambiguity, or PII-adjacent cut.

Readiness

Semantic model documented, metric dictionary approved, warehouse access wired.

Owner on client side · Head of Data

Impact contribution

35-45% of role impact is ad-hoc question throughput with query-lineage fidelity.

Primary KPI · Query turnaround · Sub-minute on routine; under an hour on multi-source

When this capability shows up

Real-shape scenarios.

Patterns where nl-to-sql is part of the launch set, with volume and pricing anchored to each company profile.

  • Mid-market SaaS with a governed warehouse and BI adoption

    SaaS · 300-800

    500 / mo

    A 500-person B2B SaaS company fields 500 questions and recurring reports a month. The analytics queue runs days long. Dashboards refresh late. Stakeholders ping analysts on the same metrics weekly.

    Data Analyst activates NL-to-SQL and report distribution. Validated answers return in minutes; recurring dashboards ship on cadence; analysts shift to modelling and insight.

    Expected outcomes at this volume: query turnaround sub-minute on routine, report distribution coverage above 98%, analyst hours recovered weekly.

    Monthly cost

    €400€1.8k

    vs human anchor

    €3.5k€12k

    Savings

    03%

  • Enterprise services firm with pipeline complexity and executive reporting

    Services · 800-2000

    1,200 / mo

    A 1500-person services firm handles 1200 queries and reports a month across engagement-mix, utilization, and margin metrics. Pipeline breaks surface days late. Anomalies land in leadership decks before the data team sees them.

    Data Analyst activates all four capabilities. Questions answer in minutes; recurring reports ship on cadence; pipeline breaks surface in hours; anomalies flag with contributing-factor context.

    Expected outcomes: cycle-time reduction 50-70% on ad-hoc queue, pipeline-health detection lead time in hours, metric anomaly time-to-flag in minutes.

    Monthly cost

    €960€4.2k

    vs human anchor

    €8.5k€28k

    Savings

    03%

  • Small subscriptions business running a lean data function

    Subscriptions · 40-100

    200 / mo

    A 70-person subscriptions business fields 200 questions and recurring reports a month with a two-analyst team. Stakeholders ping on the same cohort questions weekly. Board decks get assembled the night before.

    Data Analyst activates NL-to-SQL and report distribution. Validated answers return in minutes against the governed warehouse; board and operating reports ship on cadence; analysts shift to modelling.

    Expected outcomes at this volume: query turnaround sub-minute on routine, report distribution coverage above 98%, analyst hours recovered weekly.

    Monthly cost

    €160€700

    vs human anchor

    €1.5k€5.0k

    Savings

    03%

  • eCommerce brand with pipeline breaks and commercial anomaly risk

    eCommerce · 250-800

    2,200 / mo

    A 450-person eCommerce brand handles 2200 queries and reports a month across merchandising, marketing-mix, and margin metrics. Pipeline breaks surface in dashboards before the data team sees them. Anomaly patterns reach leadership without context.

    Data Analyst activates NL-to-SQL, data-quality monitoring and anomaly detection. Questions answer in minutes; pipeline breaks surface within hours of occurrence; anomalies flag with contributing-factor context before leadership sees them.

    Expected outcomes: query turnaround sub-minute on routine, pipeline-health detection lead time in hours, metric anomaly time-to-flag in minutes.

    Monthly cost

    €1.8k€7.7k

    vs human anchor

    €16k€52k

    Savings

    03%

All scenarios and cost ranges come from the Data Analyst role page.

Capability-specific integrations

Additional systems for NL-to-SQL.

Beyond the Data Analyst's base stack, this capability plugs into:

Last reviewed

Activate NL-to-SQL as part of a Data Analyst deployment.

Your free Agent Opportunity Audit opens with Data Analyst and NL-to-SQL pre-selected. We map the fit and the cost against the equivalent hire, with no obligation.