FORWARDED THIS? Subscribe here to get the next one.
ISSUE 05 · AI + COMPENSATION
AI will build on your worst data as happily as your best
TL;DR
Why a recruiter bot fired the wrong numbers into live offers, the one boring step that stops it, and a quick game to spot which job architecture an AI built. Plus what's moving in comp this week.
A NOTE FROM GIAC
A few years back I watched a recruiter send an offer built on a salary band we'd retired months earlier. Nobody caught it until the candidate had signed. Cleaning that up took weeks and one very awkward conversation.
Arif Ender brought the whole thing back this week. He runs compensation across 65 countries at Palo Alto Networks and he's been doing this for nearly two decades. When his team built a chatbot so recruiters could answer their own policy questions, it worked beautifully. Then it answered a couple of them from an outdated policy.
Same mistake I made by hand, except his ran at machine speed. That's the thread through this whole issue. AI doesn't rescue a messy setup. It scales whatever's already there.
Grab a coffee. Arif's one of the sharpest guests I've had on, and this one's worth the 10 minutes.
A quick postcard before you go. It's World Cup final day here in Madrid, Spain against Argentina tonight, and the city has been dressed in red all week. Last night we watched the sun drop behind the Almudena with a cold Mahou, the beer this city is fiercely proud of. The third photo is Moncloa the last time Spain lifted this trophy. Half my new neighbourhood is Argentinian, so whatever happens tonight, somebody near me wakes up celebrating.
 |  |  |
| The Almudena at sunset · Mahou doing its job · Moncloa in 2010. Fans photo Daniel Dionne, CC BY-SA 2.0, Wikimedia Commons. | ||
Giac
THE BREAKDOWN
Before you build a comp bot, get one clean version of every policy
Arif's team had a real problem. Recruiters kept raising tickets for the same basic questions. What's the car allowance in this country, is this location eligible for RSUs, what's the relocation policy here. Every one of them waited on an SLA for an answer that already lived in a document somewhere.
So they loaded the policies and pay tables into NotebookLM and turned it into a chatbot. Recruiters could ask in plain language and get an answer in seconds instead of opening a ticket. Genuinely useful.
The trouble came from the source documents. A couple were outdated versions, and the bot answered from them with total confidence. Those answers reached live offers before anyone spotted them, and unpicking a signed number costs far more than any ticket ever would.
Standing up the bot isn't the hard part anymore. Any of us can do that in an afternoon. The hard part is the unglamorous work underneath it, one current version of every policy, dated, with the retired ones pulled out of reach. Do that first and the bot is a gift to your recruiters. Skip it and you've handed your worst mistakes a megaphone.
We wrote the full guide, how to build a recruiter facing policy assistant on your own data without shipping a single outdated number, with the version check to run before you switch it on.
AI doesn't fix a messy foundation. It builds on it faster.
RANGE PODCAST
 | Arif Ender Director of Compensation EMEA and LATAM, Palo Alto Networks |
Arif Ender runs comp across 65 countries. He thinks reward pros are turning into architects.
Arif Ender is Director of Compensation for EMEA and LATAM at Palo Alto Networks, covering more than 65 countries. He trained as a mechanical engineer, spent time at Meta, Mars and Mercedes, and has been building with AI since before we called it that, back when he ran sentiment analysis on employee pay surveys at Facebook.
His main argument is about identity. For years comp people have called themselves program managers or analysts. Arif thinks that frame has aged out. The work is closer to product management now, and the person doing it is an architect. His warning is blunt. If your value is that you build a tidy model in a spreadsheet, the next decade will be hard. If your value is judgement, business partnering and knowing what good looks like, it'll be the best stretch of your career.
He also killed an $8,000 vendor model by rebuilding it in Claude in 15 minutes, and he's got a sharp take on the one question to ask any vendor before you buy, the one nobody markets to you. Worth a listen whether you build or not.
SPOT THE AI
Arif thinks reward pros are becoming architects, so here's a test. One of those career ladders is real. The other an AI built in seconds. Reply with A or B and I'll tell you which, plus the three tells that give it away every time.
ON THE RADAR
AI skills now carry a 62% pay premium. A 2026 global jobs barometer, built on more than a billion job ads across 27 countries, found workers with AI skills earn 62% more than peers in the same roles, up from 57% last year. In the US, postings asking for AI skills grew about 144% in a year. If your survey data lags that shift, you're mispricing AI-skilled roles right now. Read → the AI jobs barometer
AI is everywhere in the conversation and almost nowhere in the actual work. A benchmarking study of over 525 total rewards teams scored the average team at just 4.3 out of 16 AI capabilities. More than half have adopted fewer than 5. The gap is the whole opportunity, and whoever gets their data house in order first gets a head start nobody else has yet. Read → the AI maturity benchmark
The EU's deadline for AI in hiring just moved to the end of 2027. Under the new Digital Omnibus, now in force, the high risk rules covering AI used in recruitment and employment decisions slip from 2 August 2026 to 2 December 2027. The transparency duties, like labelling AI generated content, still land this August. More runway before the hard rules bite, but the direction is set, so build your human oversight trail now while it's cheap. Read → the client alert
FROM THE COMMUNITY
Welcome to the room, Urvashi Agarwal and Evert Kraav. Both went public about joining this week, which still counts for a lot when the whole thing runs on trust.
That's what the room is for. Somewhere to pressure test a build before it touches a real payslip, with people who've already hit the wall you're walking towards.
RECENT EDITIONS
Reading this because a friend forwarded it? Get Range Weekly in your own inbox every Sunday. Subscribe free |
 That's me for this week. If Arif's story poked at something you're about to build, hit reply and tell me where you're stuck with AI. I read every one. Until next week, Giac |
 |
When you're ready, here are three ways Range can help you.
1. Join the community
Find your people. Inside Range, reward pros learn to build with AI together, the prompts, the setups, the band workflows that actually hold up. Free is the signal, the room is the substance. Apply to join →
2. Grab the free resources
Everything free we've made now lives in one place, the Comp AI Vault. The guides, the prompt library and the frameworks, all in the open. Browse the Vault →
3. Reward Rewired 2026
Our in person gathering for the people actually building this in comp. Register your interest →