Exemplary AIEnterprise
For creatorsDemoBook a demo

“I started in March — how much carry-over do I actually have?”

Manager ALeave §2.4
Manager BLeave §2.4
Manager CLeave §2.4
§2.4Joiners in the first half of the year carry over up to five days, pro rata.
  1. Question asked
  2. Three managers
  3. Three readings
  4. Retrieve clause
  5. One answer
  6. Cited
Human Resources

One policy. One reading of it. Every time.

Two managers, two interpretations of the same clause — a fairness problem long before it is an efficiency one. Exemplary AI answers from the handbook itself with the section attached, in the language the employee asked in, on infrastructure where personnel data is allowed to sit.

Talk to us
120+ languagesAnswers cite the policyPersonnel data stays internal
The unread handbook

The handbook is not the problem.

It is usually well written, current, and unread. What is missing is a way to ask it something.

The same questions reach the same few people

Employees ask a colleague because a colleague is faster than a document. Managers ask HR because nobody wants to be the one who got it wrong.

The people who maintain the policy spend the week repeating it.

Consistency depends on who you ask

One clause, a firm answer from one manager, a careful one from the next. Neither wrong on purpose. Employees compare notes. The gap arrives in HR attached to a name.

By then it is a case, not a question.

Not everyone reads the working language

A policy that exists in one language is a policy half the workforce receives second-hand, through whoever was willing to translate it in the moment.

Second-hand policy is how one company ends up with two rules.

Where it gets used

The policy answers, whoever is asking.

One indexed set of documents stands behind an employee's question, a new joiner's first week, and the letters and packs the HR desk has to produce. None of it touches an employment decision.

The handbook, answerable.

An employee asks in plain language and gets the answer with the clause behind it. A manager asks the same thing and gets the same answer. The answer changes when the policy changes, not when someone remembers to update a second copy.

  • Answers cite the policy and section they came from
  • Asked and answered in 120+ languages, from one maintained document
  • Available in web chat, in Slack, or through the browser extension
  • Updates the moment the source document is re-indexed
Policy questionemployee · web chatIllustrative
Asked in Portuguese

“I started in March — how much carry-over do I actually have?”

Joiners in the first half of the year carry over up to five days, calculated pro rata from the start date. Anything above five lapses at year end unless a manager approves an exception.

Leave policy · §2.4 Carry-over

Exception requests go to a person — the agent does not approve them.

One maintained source

The employee and the manager read the same clause.

Nothing is translated in advance. Answers are produced from the indexed document in the language of the question, so two people in two languages land on the same section of the same file. 120+ languages, one document to keep current.

One clause, two readingsIllustrative
Maintained onceLeave policy · §4.2 Sickness absence
Manager · asked in English

“How many days can someone be off sick before they need a doctor's note?”

Up to three consecutive days can be self-certified. From the fourth day a medical certificate is required, and it goes to HR rather than to the line manager.

Leave policy · §4.2 Sickness absence
Employee · asked in Portuguese

“Quantos dias posso faltar por doença antes de precisar de atestado médico?”

Até três dias seguidos podem ser autodeclarados. A partir do quarto dia é necessário atestado médico, entregue aos Recursos Humanos e não à chefia.

Leave policy · §4.2 Sickness absence
EnglishPortuguêsFrançaisதமிழ்Deutsch日本語+ 114 more

Read in the language the employee thinks in — from the same indexed document.

One question

A question arrives on a Sunday night.

The panel on the left is what the employee sees. Four things stand behind it — and the fourth is the one HR teams care most about.

Policy questionemployee · web chatIllustrative
Asked in English

“I am transferring to the Lisbon entity in June. Whose notice period applies to me?”

Notice is set by the contract of the entity that employs you on the day you give it. From the transfer date the Portugal handbook applies, and the sixty days in your current contract do not carry across.

Portugal handbook · §5.1 Notice

Contract changes go to a person. The agent does not amend terms.

01

They ask in their own words, in their own language.

Not a search query and not a form. A sentence, of the kind people actually type at half past ten on a Sunday when the office is shut.

02

The agent reads the policy, not the internet.

Grounded retrieval across the indexed handbook and the supporting documents your organisation maintains. A general model's idea of typical leave policy never enters the answer.

  • Scoped by group, so the entity that employs someone decides which handbook answers
03

The answer arrives with the clause attached.

Section and document, one tap away. The employee can check it, and so can the manager they forward it to, which is what stops a helpful answer becoming a disputed one.

04

Anything sensitive goes to a person.

Grievances, accommodations, anything touching an individual's circumstances — the agent hands over rather than interprets. It answers questions about policy; it does not make decisions about people.

Once the policy answers

The handbook stops being something you publish and becomes something people use.

Once the policy can be asked, the questions start telling you something. Every one the agent hands to a person marks a place the documents do not yet answer, and it lands with the team that owns them. HR stops finding out which clauses are unclear from the dispute they caused.

The people who maintain the policy spend their time improving it rather than repeating it.

An employee reads the answer in the language they think in, and it is the same answer their manager read.

The next entity and the next handbook are a folder and a group, not a project.

Personnel data

Personnel files never leave your systems.

HR holds the data with the least tolerance for a third-party processor — and usually the strictest works-council and data-protection commitments attached to it.

Nothing is sent anywhere for inference

The model runs inside your own infrastructure. Handbooks, correspondence and employee questions are never transmitted to a model provider, and are never used to train anything.

Scoped, revocable, logged

Groups and IAM decide what each agent can reach. Access is least-privilege by default, deny overrides allow, and revocation takes effect immediately.

Assistive, never evaluative

The platform retrieves and drafts. Hiring, performance and any decision about an individual stays with the people and processes accountable for it.

GDPR-ready, ISO 27001 and SOC 2 Type II — deployed inside your own perimeter.

How little it takes

A filestore, a group, and an agent.

It is a deliberately small amount of the platform — which is why HR is usually the second team onto a deployment rather than a project of its own.

Documents
Filestore folders, watched sync, upload or API
Reading
OCR and indexing across handbooks, PDFs and exports
Retrieval
Grounded retrieval, cited to the policy section
Language
120+ languages from one maintained source
Access
Groups and IAM — least privilege, deny overrides allow
Delivery
Web chat, Slack, or the browser extension
Systems
170+ integrations via MCP, plus custom MCP servers
Oversight
Hand-off to a person for anything about an individual
Deployment
On-premise, private cloud or air-gapped
Go deeper:File storeGroupsChatbotsAgents
Common questions

What HR teams ask first.

See Exemplary AI in action

Book a demo and we'll show you purpose-built agents grounded in your own knowledge — deployed on your infrastructure.

Book a demo
Exemplary AIEnterprise

Agentic AI for the enterprise — installed inside your network.

  • Sovereign deployment
  • Purpose-agnostic agents
Book a demo

Build

  • Agents
  • Model Hub
  • Knowledge Graph
  • File store

Connect

  • Integrations
  • Chatbots
  • Browser Extension

Govern

  • Management
  • Groups
  • Analytics

Company

  • About Us
  • Exemplary for creators
  • Book a demo

Use cases

By industry
  • Healthcare
  • Government
  • Financial services
  • Legal
  • Media
By department
  • Customer Support
  • IT & Engineering
  • Human Resources
  • Finance & Procurement
  • Compliance & Risk

© 2026 Exemplary AI. All rights reserved.

  • SOC 2 Type II
  • HIPAA
  • GDPR-ready
  • ISO 27001