24/7 Employee Support
Always-on availability inside the employee portal, on desktop and mobile web, from nine entry points that all open the same conversation. Sessions persist across pages and expire after 24 quiet hours.
Meet Athena — Your Always-On Benefits Assistant
Smart. Fast. Empathetic. Always Available. Athena is the AI-powered benefits chatbot that makes employee support effortless.
Athena answers from your own benefit records, your company's approved policy documents, and the enrollment field you're looking at — with a citation, a next-step button, and a thumbs-up or -down on every answer. It explains and points. It never enrolls anyone, and it never sees anyone else's data.
Intent-first, not keyword-firstCites its sourcesReads only your records20+ languagesNever enrolls
Smarter Conversations. Happier Employees. Lighter HR Workload.
"HR departments spend a large share of their time on the same questions — when coverage starts, whether a dependent is eligible, what a paycheck deduction means. A generic chatbot answers those with a paragraph and a hope. Athena answers with the employee's own numbers, the company's own policy, and the button that finishes the task — and hands off to a person the moment it isn't sure."
Instant Answers, No Wait Times24/7, on every page of the portalContextual UnderstandingIntent, confidence, and a "did you mean?"From Confusion to ClarityYour records, your policy, citedGuided Assistance for Every StepButtons and to-dos, not just textActionable IntelligenceEvery thumbs-down teaches the setup
3ways an answer is grounded — the employee's own records, approved policy documents, or the field on screen — never free association
9places in the employee portal that open the same Athena conversation, from the floating icon to the "i" on a plan tile
0elections made by Athena — plan changes always go through the enrollment engine, with Athena pointing the way
20+languages, with tone that adapts to the employee's sentiment, across desktop and mobile web
Working flow
Who Clicks What, in Order
The employee chatbot as it runs today — the order of clicks on both sides of the conversation, and the rule the platform enforces at each one.
| # | Who | What they do | What the platform enforces |
|---|---|---|---|
| 1 | Administrator | Creates an intent — name, description, example questions, a confidence threshold — and marks whether it may read the employee's data, search policy documents, show buttons, or several of those. | An intent with no examples or the wrong flags produces a weak answer no matter how good the model is; turning an intent inactive hides that path without deleting any chat history. |
| 2 | Administrator | Uploads policy documents (SPDs, guides) and writes the approved help text and common questions behind every enrollment control that has an "i" button. | Documents are chunked and indexed for retrieval; field help is the only source a field question may use. |
| 3 | Employee | Opens Athena from any of nine places in the portal. | The launcher is shown only to employees with the Athena permission; the panel restores an active session or greets the employee by first name. |
| 4 | Employee | Types a question — or clicks a button, a common-question chip, or a pending action. | A typed question is classified; a click already says what the person means, so it skips classification and runs that path directly. |
| 5 | Athena | Classifies the question against the intent catalog with the last few turns as context and returns an intent, a confidence and a sentiment. | Whether the intent may read data, search documents or show actions comes from the administrator's configuration, never from the model. |
| 6 | Athena | Applies the confidence gate: sure → route; unsure but recognisable → "Did you mean…?" with Yes / No; unsure and unknown → small talk. | Yes is handled as that intent; No falls to a friendly reply; a handler failure falls back to an apology, sample buttons or Contact HR — never a blank bubble. |
| 7 | Athena | Answers from one of four sources: the employee's own records (read-only), cited policy passages, approved field help with the live tile values, or configured buttons and to-do cards. | Personal answers never print identifiers or internal names; a policy answer without a matching passage says so instead of pretending; two modes on one intent run both and merge. |
| 8 | Employee | Acts on the answer — opens the Dependents overlay, resumes enrollment, uploads a proof — or rates it with a thumb. | Elections, waivers and changes happen on the enrollment engine with its own rules and audit trail; Athena only points there. |
| 9 | Employee · HR | Thumbs-down → reason → retry, ask differently, contact HR or create a case; ends the chat or lets it expire. | Every turn, source, thumb and button click is logged; the session closes on End Chat or after 24 quiet hours; Contact HR carries the transcript into BR Resolve. |
One Conversation, Five Steps
Follow an employee named Jane from opening the panel to rating the answer. Each step is a decision Athena makes in the open — and one an administrator can tune without a developer.
The panel opens where the question is
Jane clicks the floating Athena icon — or "Chat Now" on the dashboard, "Ask Athena" on a plan tile, "Need help deciding?" on the waive banner, or "Start chat" on an HR contact. It is the same conversation everywhere. Athena restores a session that is still active or greets Jane by first name with a time-of-day welcome and three shortcuts: My Tickets, My Documents, Pending Actions.
What happens
- Active session restored with its history, so a follow-up from another page still makes sense
- Welcome intent comes from an administrator-configured action, not a hard-coded string
- Launcher is shown only to employees with the Athena permission
- If the greeting service is slow, a static welcome keeps the composer usable
Why it matters
Employees never hunt for help; help is already on the page.
Closing the panel keeps the session; "End chat" or 24 hours of silence closes it.
Athena decides what you meant before it answers
Jane types "how much do I pay for dental?" Athena loads the last few turns for context and asks a large language model to classify the message against the company's own intent catalog — each intent with example questions an administrator wrote. The model returns an intent, a confidence score, a sentiment and any entities. Whether that intent may read employee data, search policy documents or show buttons is decided by the configuration in the database, not by the model.
How it works
- Intent catalog: name, description, example questions, allowed data tables, allowed documents
- Confidence gate: below the intent's threshold Athena asks "Did you mean…?" with Yes / No
- Buttons and FAQ chips skip classification — a click already says what the person means
- Sentiment is tracked per message; repeated frustration can escalate to HR
Why it matters
A guess with a score beats a confident wrong answer.
Turning an intent inactive hides that path without deleting chat history.
Four kinds of answer, each grounded differently
A data question becomes a read-only, employee-scoped database query that Athena then explains in plain English — Jane sees a cost card, never a query or a personnel number. A policy question is embedded, matched against chunks of your uploaded SPDs and guides, and answered only from passages that clear a similarity bar, with the document cited. A field question from a plan tile is answered from the approved help text and the live numbers on that tile. A "take me there" question returns a button. When two modes apply, Athena runs both and merges.
The four paths
- Personal data — "what are my plans? how much is dental?" → your rows, SELECT-only
- Company policy — "is acupuncture covered? HSA deadline?" → cited document passages
- Field help — the "i" on an enrollment tile → approved definition + your live amounts
- Navigation & to-dos — "show my pending actions" → cards and routes
Why it matters
No document found means Athena says so — it never pretends.
Answers are instructed not to print identifiers, table names or internal codes.
Guided assistance for every step — without leaving the chat
Under every answer are the buttons an administrator configured for that intent: open the Dependents overlay, show my tickets, compare plans, contact HR. Pending-action cards surface what is waiting on Jane — start open enrollment, upload a proof document, name a beneficiary — with the same navigation path as a button. During enrollment, the "i" on any plan tile opens Athena with a banner ("Answering questions about HSA per-paycheck contribution") and up to four common questions as chips; a typed follow-up is answered only from the approved field text and Jane's live amounts.
What the employee can do
- Navigate to a portal page, overlay or external link — the chat minimizes and can reopen after
- Re-run a data or policy lookup from a button without re-guessing the intent
- Work through pending-action cards, paginated, with "you're all caught up" at the end
- Get step-by-step help on plan selection and document verification during enrollment
Why it matters
Fewer incomplete submissions — the live estimate is up to 40% fewer.
An open-enrollment deep link only works if Jane actually has an open offer; otherwise Athena says so.
Turning conversations into actionable intelligence
Jane can thumb an answer up or down. Thumbs-down opens a reason panel first — wrong plan, wrong benefit, incorrect — then a recovery choice: retry (Athena asks again and stores the new answer linked to the old one), ask differently, contact HR, or create a case. Every turn writes an interaction log: the intent, the confidence, the mode that answered, the sources used and the timings. Button clicks are tracked silently. Those records feed the Insights view for the employee and the analytics administrators use to tune intents, documents and field help.
What is captured
- Thumbs, reason codes, and the recovery the employee chose
- Intent, confidence and sentiment per message; low-confidence clusters flag missing examples
- Which document or FAQ each answer cited — and which questions found no document at all
- Button shown → clicked → completed, so navigation actions are measured, not assumed
Why it matters
A bad answer becomes a configuration fix, not a rumor.
Employees can download their own transcript and view their own Insights from the ⋮ menu.
Ground rules
Rules of the Assistant
The behaviours an employee can rely on and an auditor can check — written from the conversation engine as it runs, not from a policy document.
| Rule | How it behaves | Why it matters |
|---|---|---|
| Explains, never enrolls | Athena reads and points. Premiums, eligibility dates and elections stay in the enrollment engine; "Enroll" is always a button that opens that screen. | No benefit record is ever written from a chat. |
| Your rows only | Every personal question carries the signed-in employee's own identifier; the read is limited to the tables the intent allows and is read-only. | An employee cannot see anyone else's data by asking, and answers never print identifiers or table names. |
| Documents or nothing | A policy answer is written only from passages that clear the similarity bar, with the document linked; below the bar, Athena gives a general answer and says no document was found. | No invented policy, no invented page number. |
| Field help is approved text | The "i" on a plan tile opens Athena with a banner and common-question chips; a chip returns curated text with no model call, a typed follow-up uses only the approved definition and the live tile numbers. | The employee reads the same words the benefits team approved, with their own amount filled in. |
| Configuration decides the mode | Data, policy and action flags — and the confidence threshold — are read from the intent, not guessed by the model. | A classifier can be wrong about the meaning; it cannot grant itself access. |
| Ask before guessing | Below the threshold Athena asks "Did you mean…?"; a handler error apologises and offers sample buttons or Contact HR. | A wrong-but-confident answer is the one failure mode the design refuses. |
| Buttons come from setup | Every action under an answer — message, route, link, overlay, data or policy re-run, pending to-dos — is a row an administrator configured. | Nothing appears in the chat that the organisation did not allow. |
| The session belongs to the employee | Closing the panel keeps the conversation; End Chat or Refresh closes it; 24 quiet hours expire it; the employee can download the transcript and see their own Insights. | Continuity across pages without a conversation that never ends. |
| No deploy for a fix | A new intent, example, document, button or field-help text goes live from the configuration console. | The thumbs-down you got this morning is a configuration change this afternoon. |
Core Capabilities That Power Athena
Availability, understanding, grounding, guidance, analytics and security — the six things an employee benefits chatbot must do, each one configurable by a benefits administrator.
Contextual Understanding
An intent catalog written in your words — with example questions, confidence thresholds and a "did you mean?" clarification — interprets each question using the last few turns, the employee's eligibility and their location.
Grounded Answers with Citations
Personal answers come from read-only, employee-scoped queries; policy answers from your uploaded SPDs and guides with the document linked; field help from approved text plus live tile values. Nothing is invented to fill a gap.
Enrollment & Form Assistance
Guides benefits selection and document verification step by step: plan-tile "Ask Athena" help with FAQ chips, pending-action cards for uploads and beneficiaries, and a hand-off to IntelliSuggest for a ranked recommendation.
Feedback & Analytics Engine
Tracks usage, sentiment and issues: thumbs with reason codes, retry outcomes, low-confidence clusters, unanswered policy topics and button completion rates — the signals that tell you which intent, document or field text to fix next.
Secure by Design
Encryption in transit and at rest, employee-scoped data access, answers instructed never to print identifiers, audit logs for every turn, and a permission that decides who sees the launcher — HIPAA-aligned.
A Generic Chatbot vs. Athena
Most AI can answer a question about benefits. Very little AI can answer a question about your benefits, from your employer's documents, and then open the right screen.
| The question | A generic AI chatbot | Athena in Benefits Reimagined |
|---|---|---|
| "How much do I pay for dental?" | Explains how dental premiums work in general. | Your number. A read-only query of your own elections, explained as a cost card with your next deduction date. |
| "Is acupuncture covered?" | Answers from the open internet or its training data. | Your SPD, cited. Only passages from your uploaded documents that clear a similarity bar — or an honest "I couldn't find that in your documents". |
| "What does this field mean?" | Cannot see the screen. | Approved field help + your live values. The "i" on a plan tile opens Athena with context and FAQ chips. |
| "Add my spouse" | Describes the steps in prose. | A button. Opens the Dependents overlay; the election itself goes through the enrollment engine and its audit trail. |
| "I don't understand" | Tries again with a longer paragraph. | "Did you mean…?" then reason codes, retry, or Contact HR / create a case — with the conversation attached. |
| Whose data can it see? | Whatever was pasted into the prompt. | Yours only. Every personal query carries your employee identifier; answers never print identifiers or table names. |
| Who tunes it? | A developer, with a prompt. | A benefits administrator — intents, examples, documents, buttons and field help in the configuration console, no deploy. |
How administrators teach Athena
Configured in Your Words, Not in Code
Athena can have every document uploaded and the best model connected and still give a weak answer if an intent is missing or mis-flagged. So the setup is a benefits task, done in the configuration console in a recommended order — connect, catalog the intents, add buttons and a welcome, load the knowledge, then prove it as a test employee.
- Intents — a name, a description, example questions, a confidence threshold, and whether the intent may read data, search policy, show actions, or several at once
- Table hints — which employee tables a data intent may read (health, insurance, savings elections), never anything broader
- Buttons and welcome — messages, in-app routes, external links, data or policy re-runs, and the one allow-listed function that loads pending to-dos
- Documents and field help — SPDs and guides chunked and embedded for retrieval; approved definitions and FAQ chips for every enrollment control with an "i"
- Prove it — ask the example questions as a test employee, thumb-down the wrong routes, adjust; a new intent, document or button needs no deploy
Is acupuncture covered on my medical plan?
Athena · from your documentsYes, on your PPO Standard plan acupuncture is covered at 80% after deductible for up to 20 visits per plan year when performed by a licensed provider. Chiropractic visits share the same 20-visit limit.
Find an in-network providerWhat counts toward my deductible?
👍 👎 Copy
What if I switch to the HDHP next year?
Athena · from your documentsThe HDHP covers acupuncture at 100% after you meet the deductible, with the same 20-visit limit. Because the deductible is higher, most of those visits would be paid from your HSA first. Want a side-by-side for your usage?
Ask me anything about your benefits…
Why Organizations Choose Athena
HR Efficiency
Reduces ticket volume by up to 60% with instant self-service support — the questions that used to be tickets end in the chat, with the next step attached.
Employee Experience
Improves satisfaction with responsive, personal interactions: the employee's own numbers, their employer's own words, and a button instead of a paragraph of directions.
Scalability
Serves global workforces without expanding HR teams — 20+ languages, sentiment-aware tone, and one configuration that a shared-services center or PEO can run for many populations.
Compliance
Built with HIPAA alignment: employee-scoped reads, logged turns, cited sources, and no elections ever made outside the audited enrollment engine.
Insight Generation
Transforms support data into strategy intelligence — top intents, low-confidence topics, unanswered policy questions and thumbs-down reasons by region or department.
Seamless Integration
Works within the same portal as enrollment — inside SAP SuccessFactors and any other HRIS Benefits Reimagined syncs. No second login.
Where Support Meets Empathy, Intelligence, and Scale
Athena is one member of the BR Intelligence family and the front door to several others. It opens IntelliSuggest for a ranked recommendation, files a case in BR Resolve when a person is needed, explains a Communication HUB notice in the employee's context, and reads the documents the Document Management System already holds. Administrators get their own Athena — Athena for Admins inside the modules they configure, and AI-Powered Configuration for building the benefits blueprint — so the employee chatbot stays focused on the employee.
- Employee: a conversation that continues — across pages, after an overlay closes, and the next morning if the session is still active
- Employee: Common Questions on enrollment — a searchable FAQ library in the right rail with one-shot answers and recommended plan cards, a companion to the chat
- Employee: My Insights — from the ⋮ menu, seven read-only views of their own Athena history: overview counts, messages by day, topics asked, usage time, question success rate, repeated questions and a journey timeline — each with a drill-down to their own rows, never anyone else's
- Admin: the Athena console in the configuration console — intents, examples, table hints, documents, actions, welcome, field help and FAQs
- Admin: conversation and usage analytics — intents by volume, confidence distribution, thumbs by reason, sources cited, and button completion
- HR: escalations with context — "Contact HR" and "create a case" carry the transcript into BR Resolve so nobody starts over
Four Stories, End to End
Four moments from one employee's month, each traced through the flow above.
The new hire and the dental question
Jane opens Athena from the floating icon on her first Monday. No session exists, so Athena greets her by name with My Tickets, My Documents and Pending Actions. She types "how much do I pay for dental?"; the data intent clears its threshold, a read-only query returns her own dental election, and the bubble shows a cost card with her next deduction date and a Compare plans button.
The plan tile and the HSA limit
During enrollment Jane taps the "i" on the HSA contribution tile and then Ask Athena. The panel opens with a banner naming the field and four common questions. She taps "What's the limit?" — the answer is the approved help text with her live per-paycheck amount and paycheck count filled in. No model wrote it; the benefits team did.
The thumbs-down that fixed the catalog
Jane asks whether acupuncture is covered. Athena cites the wrong dental summary, so she thumbs it down, picks "wrong plan" and taps Retry; the retried answer cites the right document and is stored linked to the first. In the administrator's analytics, "wrong plan" on dental is the top reason that week — two documents share a title, and renaming one is a configuration change, not a ticket.
Pending actions and the beneficiary
Jane clicks Pending Actions. Athena runs the one allow-listed function that loads her to-dos: name a beneficiary, upload a marriage certificate, resume the open offer. She opens the beneficiary overlay from the card; the chat minimises, she finishes, and the same session is still active when she reopens Athena that afternoon.
Ask Athena Something Only Your Data Can Answer
Bring a real SPD and three questions your help desk hears every week. We'll load the document, configure the intents in front of you, and let you ask as a test employee — then thumb it down and watch the fix take effect without a deploy.
Employee-scoped queriesCited sourcesHuman hand-offHIPAA-alignedNever enrolls
Athena FAQs
It's a virtual assistant that provides real-time employee support for benefits, HR policies, and enrollment. Athena goes one step further than a FAQ bot: it classifies what the employee means, then answers from that employee's own benefit records, from your approved policy documents with a citation, or with a button that opens the right screen — and it hands off to HR when it isn't sure.
By automating FAQs, providing guided enrollment, and reducing repetitive support tasks. Questions like "when does my coverage start?", "how much is dental per paycheck?" or "what do I upload for my new baby?" are answered inside the conversation, with the next step one click away, so they never become a ticket. The live site's estimate is a reduction in ticket volume of up to 60%.
Yes. Athena supports 20+ languages, answers in the language the employee writes, and runs inside the same portal your workforce already uses for enrollment — including SAP SuccessFactors and any other HRIS Benefits Reimagined connects to.
Athena is grounded, not generative-only. Personal answers come from read-only, employee-scoped queries of that person's records. Policy answers are written only from document passages that pass a similarity threshold, and each answer cites the document it used. When nothing qualifies, Athena says so and offers a general answer or a route to HR rather than pretending it found a document.
No. Athena explains, retrieves, and points. Elections still go through the enrollment engine with its eligibility rules and audit trail; Athena sends you there with a button. Every personal query carries your own employee identifier, so Athena cannot read anyone else's rows, and answers are instructed never to print internal identifiers or table names.
Employees see their own Insights (overview, activity by day, topics, usage time, success rate, repeated questions, journey) from the ⋮ menu — always scoped to the signed-in person. HR teams see the population view in Chatbot Analytics: which intents are asked most, where confidence is low, which answers get a thumbs-down and why, and which buttons employees click after an answer. Those signals tell you which policy document to update, which example questions to add, and which enrollment field needs clearer help text.
Administrators configure Athena in the configuration console: intents (what people mean), example questions, which employee data or policy documents each intent may use, the buttons that appear under an answer, and the field-help text behind every "Ask Athena" on an enrollment tile. A new intent, document, or button goes live without a code deploy.
It uses encrypted communication, privacy-by-design architecture, and compliance logging aligned to HIPAA. Conversations, sources, feedback and button clicks stay on the platform's own database, sessions expire automatically, and employees can download or end their own conversation at any time.
Explore Benefits Reimagined Today
An assistant that knows the employee, quotes the employer, points to the screen, and knows when to hand off — Athena is benefits support the way employees already expect it from their bank and their airline.