AI That Helps With Benefits Should Explain Itself — and Never Decide for You.
Athena answers questions. IntelliSuggest suggests plans. IntelliMap proposes mappings. None of them changes a record without a person, and all of them show their reasoning.
Benefits decisions touch health, family and pay. So we set rules for our AI before we set it loose: what each helper may see, what it may do, how it is tested, and how a person stays in charge. This page states those rules in plain words, and tells you how to check them.
What We Promise About Every AI Helper in the Platform
Each commitment comes with the place in the product where you can see it working.
A person can always override the AI
No helper enrolls, changes, sends or pays on its own. It proposes; a person approves, changes or rejects; and the decision is kept with the name and the reason.
We explain every suggestion
A suggestion without a reason is a guess. Every helper shows what it looked at and why it suggests what it does, in words an employee can check against their own situation.
Your employees' data never trains public models
We use commercially available large language models under agreements that stop the provider keeping or learning from your data. One customer's data never improves another customer's results, and it is never sold.
We test for bias before and after launch
Before release and on a schedule afterwards, we check whether answers and suggestions differ for people who should be treated the same. Where they do, we fix or narrow the helper, and we keep the record.
We tell you what the AI cannot do
Athena says "I don't know" rather than inventing an answer. Helpers are limited to the topics and the records they were built for, and we say so — on this page and in every demo.
How we hold ourselves to it
A named owner for AI on our team; a written checklist mapped to the NIST AI Risk Management Framework that every helper passes before release; a register of every helper; a feedback path from every answer; and a review after any reported problem.
Every Helper, What It Sees, What It May Do — and What It May Not
This is the public version of the register we keep for customers. Anything not on this list is not an AI feature of the platform today.
| Helper | What it sees | What it may do | What it never does |
|---|---|---|---|
| Athena — for employees | The employee’s own record, your plan documents, deadlines on file | Answer questions, show where the answer came from, point to the next step, take feedback | Enroll, change a record, see another employee’s data |
| Athena — for administrators | The records the administrator’s role already allows | Answer questions about cases, deadlines and totals; draft a reply for review | Change a record, close a case, send a message on its own |
| BR IntelliSuggest | The employee’s situation as entered, plan rules and costs | Rank plans and show why each one fits | Choose for the employee, hide any plan the employee is eligible for |
| BR IntelliMap | A carrier’s file layout and the platform’s fields | Propose a field mapping with confidence shown | Activate a mapping without an administrator confirming it |
| BR IntelliEdit | The text an author is writing | Suggest clearer wording; draft from an outline | Publish anything; change plan facts |
| BR Spotlight | Enrollment, billing and case data the viewer may already see | Point at what needs attention and say why | Take the action; hide what the system could not prevent |
| BR Insights | A number already on a dashboard and its sources | Explain the number in plain words with sources | Calculate, edit or predict on its own |
Each helper can be turned off for your organization, and some can be limited to certain groups or topics. The platform works without them.
An Answer You Can Check, a Button Only You Can Press
This is how Athena handles the most common employee question. Notice three things: the source is named, the next step is offered rather than taken, and the employee can say the answer was not helpful — which goes to a person.
- Source shown — the plan document and the employee's own record, not a guess
- Action offered, never taken — the employee opens the life event; Athena does not
- Feedback goes to a person — unhelpful answers become a case your administrators see
Five Gates Before an AI Helper Reaches an Employee
We are a small team — the people on the About Us page — so our governance is a checklist with a named owner rather than a committee with a name. Every helper passes these gates, and the evidence is kept.
- 1 · PurposeWhat it is for, and what it is not for. Written down first: the questions it answers, the records it may see, the actions it may propose.
- 2 · DataLeast data, kept apart. The helper sees only what the signed-in person may see. Nothing leaves to train a public model.
- 3 · TestingAccuracy and fairness. Answers checked against the plan documents; suggestions checked for differences between groups who should be treated the same.
- 4 · ExplanationCan a person check it? Every answer names its source; every suggestion shows its reasoning; every limit is stated.
- 5 · WatchingAfter release. Feedback from every answer, quality reviewed on a schedule, and a review after any reported problem — with the option to switch the helper off.
See the AI Explain Itself on Your Own Plans
Ask Athena a question your employees actually ask. Watch IntelliSuggest show its reasoning. Then see the history entry a person’s decision leaves behind.
- Live demo — as your employee, then as your administrator, on your plan structure
- Security review pack — the security and privacy answers your IT and procurement teams will ask for
- RFP response — every line answered against the platform, not a brochure
Questions HR, Legal and IT Ask About AI in Benefits
Which features use AI, what they may do, what they learn from, how they are tested and how to switch them off.
Athena (the assistant that answers employee and administrator questions), BR IntelliSuggest (plan suggestions with the reasoning shown), BR IntelliMap (proposes how a carrier file maps to the platform), BR IntelliEdit (helps write and tidy employee communications), BR Spotlight (points at what needs attention) and BR Insights (explains a number in plain words). Each one is listed, with what it may do and see, in the AI register we share with customers.
No. Every AI helper proposes; a person decides. Athena never enrolls anyone. IntelliSuggest ranks plans and the employee chooses. IntelliMap proposes a mapping and an administrator confirms it. Every decision is written to the history with who made it and why.
Your data is not used to train public AI models. We use commercially available large language models under agreements that stop the provider keeping or learning from your data. We do not use one customer’s data to improve another customer’s results.
Before a helper is released and on a schedule afterwards, we test whether its answers and suggestions differ for people who should be treated the same — by age, sex, location, pay band, language and plan. Where they do, we fix the helper or narrow what it may do, and we record the test and the result.
Every Athena answer shows where it came from — the plan document, the employee’s own record, the deadline on file — so it can be checked. Employees can mark an answer unhelpful and say why; those reports go to your administrators and to our quality loop, and repeated problems become a case for a person to fix. Athena also says plainly when it does not know.
Yes. Each helper can be turned off for your organization, and some can be limited to certain groups of employees or certain topics. The platform works without them; they make it faster and easier. Ask on the contact page and we will show you the switch.
We review every helper against the NIST AI Risk Management Framework — a recognised, plain-language standard for managing AI risk — and we keep the evidence. Independent certification of our AI management is planned; we will say so on this page when it is done, not before.
AI Proposes. A Person Decides.
Explained, tested, listed and switchable — with your employees’ data kept to itself. SAP-native, ready for any HRIS.