Agentic Studio

Two employees. One question. Two right answers.

Grounded HR agents in Slack and Microsoft Teams. Every answer is assembled from the policies and plans that apply to the person asking, and nothing else.

Book a scoping callSee how grounding works

Both ask the same question

  • Slack
  • Teams
Both ask

How much parental leave do I get?

  • MRMaya R.Warehouse associatein Slack

    1Who is asking, from the Ontology

    • Hourly
    • Texas
    • 18 months
    • Plan B

    2Which resources apply

    • Parental Leave Policy 2026
    • Hourly workforce handbook
    • Salaried handbook
    • Texas state supplement
    • California state supplement
    • Tenure and eligibility rules

    4 of 6 resources apply

    3Answer

    Six weeks paid, then unpaid state leave. You passed the twelve month mark, so you qualify now.

    • Scoped to her
    • Sources on the record
  • DKDan K.Corporate, salariedin Teams

    1Who is asking, from the Ontology

    • Salaried
    • California
    • 4 years
    • Plan A

    2Which resources apply

    • Parental Leave Policy 2026
    • Hourly workforce handbook
    • Salaried handbook
    • Texas state supplement
    • California state supplement
    • Tenure and eligibility rules

    4 of 6 resources apply

    3Answer

    Twelve weeks paid, plus the California top-up. You are eligible immediately

    • Scoped to him
    • Sources on the record

Same agent, same question. The eligibility rules decide which resources it is allowed to read.

It answers where your workforce already is

No new tool for anyone to learn, and no portal for anyone to forget.

Slack

A direct message or a channel. Live today.

Microsoft Teams

Same agent, same rules, same audit trail. Live today.

Your own tooling, over MCP

Your assistant calls our agents while HR keeps the data and the agent in Aragorn.

Two ways teams use it

Answer the questions, or build the agent

One of these is for the team drowning in questions. The other is for the team running the operation.

  • For the team answering the questions

    Employee-facing agents

    Benefits, policy, general HR, or one you define. Employees ask in Slack or Teams and get an answer built from the plans and policies that apply to them.

    • Benefits

      Plan, eligibility, dependants, dates.

    • Policy

      Leave, time off, conduct, location rules.

    • General HR

      Where do I, who do I ask, what happens next.

    • Custom

      Your own scope, your own rules.

  • For the team running the operation

    Your own agents, your own rules

    Build an agent for an internal workflow. You pick the resource it reads, you write the rules it obeys, and you set the questions it should expect. The data never leaves HR.

    Live today, at a national healthcare employer

    Internal hire eligibility

    A recruiter needs to know whether an internal applicant can interview. Two rules: six months in the current role, and a recent review above a threshold. Answering means touching pay and performance data, which is why it could not be built anywhere else.

You configure the agent. Not us, and not a consultant.

Watch one get built: name it, point it at a governed resource, write the rules, then see a rule refuse something it should refuse.

  1. Name it
  2. Point it at governed data
  3. Say how people will ask
Building

Name

Internal hire

Answers employment detail questions about employees applying to a new internal position, including whether they are eligible for the position they applied to.

Governed resource

Integration Studio → Ontology

Internal Hire Applicants

Fed by Integration Studio. One resource per agent, so every answer traces back to a single governed source.

Rules

  • Individual employee requests
  • Unknown employee
  • Reasoning

Plain language, and it can reference fields on the resource. For example: if {{recruiter_email}} is not the asker's email, do not return the record.

How will people ask for this?

  • Is Steph eligible to be hired in this position?
  • Has Mark been in his current role for six months?
  • Give me Jane's current employment details

The questions you expect, in your own words. This is what makes a narrow agent accurate instead of a general one guessing.

Internal hirein Slack
  • Is Steph eligible to be hired in this position?

    Yes. Fourteen months in her current role, and her last review is above the threshold.

    Rules passed · sources on the record
  • What is everyone in the pipeline paid?

    I can only answer about one named applicant you own. That request is outside what you can see.

    Blocked by: individual employee requests
Why it is safe to deploy

Why an HR team can put this in front of employees

The first question every buyer asks is not whether an agent can answer. It is whether someone can reach data they should not see.

Scope is enforced in the model, not the prompt

An agent sees only what the person asking is allowed to see. Set per group, point in time, held in the Ontology.

It tells you why it answered

The reasoning and the resources behind any response are on the record, so an answer can be checked rather than trusted.

It holds its behaviour

Rules and grounding are configured, not remembered, so it behaves the same in month six as on day one.

Every new agent is a configuration

The second use case is not a second project. Same resources, new rules.


How it is grounded

It reads the same governed model your reporting reads

Integration Studio connects your HR systems and feeds the Ontology. Agentic Studio sits on top of it, with the same permissions and the same lineage.

  1. Your HR systems

    HRIS, payroll, benefits, ATS, and your own files.

  2. Integration Studio, into the Ontology

    One governed model, with permissions and history.

  3. Agentic Studio

    Agents answering in Slack, Teams and your own tooling.