HR AI is worth what it does to employee lifetime value.
Performance and tenure, across everyone's time with you: that's eLTV, the number every HR investment answers to in the end. The framework scores whether an AI idea raises it: which number it moves, what trust it has to earn, and whether your function is ready. Miss any of the three and it moves the other way.
The 10X HR-AI Framework. 10X is the size of the move: an order-of-magnitude gain in eLTV across your function.
Eight questions. Two minutes. An honest tier.
Impact
Which number does it move?
eLTV
performance × tenure
Trust to earn
How much does it have to earn first?
Readiness
Is your function ready?
Why score before you build: every HR AI idea ends up in one number.
Employee lifetime value is the outcome a person creates over their whole time with you: how well they perform, and how long they stay. Hire well and they start higher. Onboard well and they get there sooner. Manage well and they stay longer, nearer their peak. Every HR program you have run was, in the end, a bet on one of those.
AI is the same bet with a new tool. An agent that takes the volume off your team gives the attention back to people, and the number rises. An agent people don't trust, or a pilot that stalls, is an experience too, and the number moves quietly the other way: people disengage first, and the dashboard notices last.
The framework asks the three questions that decide which way it goes, before anything is built: which number it moves, what a wrong answer costs, and whose name goes on the result. It's the thinking behind our strategy month, cut down to one idea at a time so you can run it yourself.
-
Impact Raises it.
Faster ramp, fewer regretted exits, hours returned to the work only people can do.
-
Trust Protects it.
A system your people don't trust changes how they behave long before it changes any dashboard.
-
Readiness Keeps it.
A pilot that stalls spends more than its budget. It spends the credibility the next one needs.
How each lens feeds the number
If you've read our five questions, you'll recognize them here. Impact carries the first. Trust carries the second and fourth. Readiness carries the third and fifth. The five questions
Impact, trust, readiness. Each one can move eLTV either way.
-
01 Impact
Which number does it move, and how much of the work does it take on?
Start from a number you already report on, and that rolls up into eLTV: attrition in the roles you can't afford to lose, time-to-productive for new hires, the hours your team spends answering the same questions every week. Then the share of the work the agent would carry as it stands today. A real number and a real slice of the load is what gets funded. A nicer experience with no number underneath tends to fade.
- Scores high when
- A number the CFO already watches, and most of the volume that eats your team today.
- Scores low when
- It's not clear which number moves, or the gain is felt more than counted.
-
02 Trust to earn
What happens when it's wrong, who decides, and how sensitive is the data? The less trust it has to earn, the sooner it can be built.
This is the lens that is different in HR. Your data is the most personal in the company, your users are the most skeptical, and your mistakes don't stay quiet. So the framework asks three things: what a wrong answer costs, whether a person decides or the AI would, and how sensitive the data is. An idea that shapes a decision about a named person, or touches pay, performance or protected data, can still be built. It just has trust to earn first, and it usually isn't the first thing you build. This is also where eLTV gets lost quietly: people who don't trust a system route around it, and some of them leave.
One thing the framework will not do: green-light an idea that, as imagined, makes a decision about a person on its own. AI may inform, surface, rank, summarize or draft. A person decides. If your idea scores that way, the framework stops and says so, and the fix is usually to reshape it so a named person decides.
- Scores high when
- A wrong answer costs a correction. No decision about a person in it. Nothing personal in the data.
- Scores low when
- It would decide something about a named person, or needs health, pay or grievance data.
-
03 Readiness
Where does the data live, and whose name is on the number?
The unglamorous lens, and the one that stalls more good ideas than either of the others. If the data lives in spreadsheets and inboxes, that gets fixed first, and it's usually a matter of weeks. If nobody in your team would put their name on the number and defend it in the exec meeting, that's the thing to settle first, before anyone spends money.
- Scores high when
- Clean data in one trusted system, and a named owner who would defend the number upstairs.
- Scores low when
- Spreadsheets and shared drives, and nobody has been asked to own it yet.
The four tiers: how the three scores combine.
Each tier maps to a step in how we work, so the result comes with a next move.
-
Tier 1 Build now
Real impact, little trust to earn, ready.
Start with the proof step; go to production when the numbers hold.
-
Tier 2 Proof first
Worth doing and safe to try.
Run it on test data for a couple of weeks and decide on what you watched work.
-
Tier 3 Build with guardrails
Real impact, but trust to earn.
Design the guardrails, the notice and who decides before anything is connected. Usually not your first build.
-
Tier 4 Not yet
A lens is short, or the impact is not worth the trust it would have to earn.
You get the one thing to fix first. Fix that and score it again.
Not yet is a result, and it comes with the thing to fix.
Scored: an interview scheduling agent for a company hiring at pace.
A recruiting team spends a chunk of every week coordinating panels, candidates, rooms and reschedules. The idea: an agent that proposes slots, sends invites and prep packs, and handles reschedules, with the recruiter setting the panel and able to override any booking.
- Impact 4/5
- Moves time-to-schedule and recruiting-coordinator hours; a number the TA lead already reports. Takes on most of the coordination volume.
- Trust Little
- A wrong slot costs a reschedule. No decision about a person in it: it schedules, and the recruiter still picks the panel. Calendars and availability, nothing sensitive.
- Readiness 4/5
- Candidate stages live in the ATS, availability in calendars, both trusted. The recruiting-ops lead would take on time-to-schedule.
Tier 1 Build now
A first build: real number, little trust to earn, ready. It's on our map as a proven agent.
Run the 10X HR-AI Framework on one idea.
Eight questions, answered the way you'd say it aloud. About two minutes, and it will tell you if the answer is not yet.
Which number would this move?
Start from something you already report on. The strongest ones roll up into employee lifetime value: performance or tenure.
Nothing you enter is stored or sent anywhere. The score is computed in your browser.
One idea, scored. Your function has a list of them.
The framework scores one idea at a time. It can't tell you the order to build them in, which ones your data can support this year, how to sequence by trust so the first win earns the second, or what the whole list adds up to in eLTV. That's the strategy month.
How the strategy month worksThe map of HR agents, area by area
Thirty minutes. Bring your scorecard and the number it moves.