Forwarded this? Range Series is free. Subscribe and it lands in your inbox.
Every other function has its public AI builders. Talent acquisition, marketing, finance, engineering. Compensation has almost none. So I went and asked. Over the last few months I had over 100 conversations with comp and reward practitioners, inside our community and across the wider field, and mapped how they're actually using Claude.
Here's what I found, 11 workflows ranked by how often they came up. Most of us are still chatting. A quiet middle is pointing Claude at messy data and collapsing weeks into days. A rare few have shipped tools their managers self-serve with. The ladder is steep, and where you sit on it has very little to do with how clever you are.
THE 11 WORKFLOWS, RANKED
Here's the whole map at a glance, then the story behind each tier underneath.
CHAT STILL DOMINATES
The most common uses are the everyday ones, and there's no shame in that. It's where almost all of us start.
1. Sense checking a market read. The solo comp person with no peer to ask now has a thought partner at 11pm. Pressure test the decision, ask what you're missing, get a second opinion that doesn't cost a consulting day.
2. Levelling and grading job descriptions. Classification is the most hated, most backed up queue in comp. One team we spoke to spends about a third of its time on it. Claude clears the first pass, a human signs off.
3. Drafting the words. Comp comms, board papers, manager FAQs, policy explainers. Most of us prefer numbers to prose, and this quietly removes the writing tax.
4. Manager and HRBP assistants. Saved helpers, deliberately scoped to general explanations and never individual pay data, that field the endless why only 3% questions so you don't have to.
THE REAL STORY IS CLAUDE CODE AND COWORK
This is where the time really collapses. And the people doing it almost entirely can't code.
5. Survey crosswalks and grading pipelines. Hand Claude the messy multi source exports and the manual spreadsheet grind disappears. One comp team turned a benchmarking and grading cycle that used to run six weeks into five working days.
6. Pay equity where the data never moves. The deepest fear in the room is putting employee records in the cloud. The answer a growing number have found is to get Claude to write the statistical scripts and run them on their own machine. The model writes the maths, never sees the people, and what you're left with is a script you can walk an auditor through line by line.
7. Merit cycle automation. Award changes, budget drawdowns, currency conversion, leadership roll ups, letter generation. One comp lead built the whole thing on dummy data inside a locked down enterprise stack.
8. Build vs buy and vendor replacement. Several have rebuilt what a paid platform used to do, org charts, band building, in house HCM pieces. One team replaced a six figure org charting subscription with a build of their own.
9. Quant levelling frameworks. One reward leader co-developed a scoring model from awkward inputs, a long range revenue forecast among them. The model caught a structural quirk a one lens human read would have missed, and it landed well with the business.
THE EDGE
A small advanced core has crossed into production. It's the frontier, and it's lonely up there. Lonelier than spending the night all by yourself at the top of Mount Kilimanjaro.
10. Classification and evaluation agents trained at scale. One team trained a job classification agent on thousands of position descriptions. It didn't just speed up classification, it surfaced biases in the program they were already running.
11. Agents managers actually use. Tools wired into live systems that turn comp from a gatekeeper into something people serve themselves. A few have built equity and burn models ready for the board, market data layers that stay current on their own, even models running on their own hardware at home for total control.
WHERE THE WORK LIVES
Four surfaces and four very different users. Most comp people never leave the first one, and that's fine. The interesting movement is everything to its right.
THE THREE FEARS NOBODY POSTS ABOUT
The map above is the easy part. The reason most comp people stall isn't on it.
I can't put our data in the cloud. Half the room, near enough. Privacy is only half of it. The harder question is auditability. Could you walk an auditor from the input to the output and defend every step? That fear is exactly why the local scripts approach matters so much. And also why CompTech as a SaaS provider won’t magically disappear over the upcoming years.
I built it and I'm too scared to ship it. The skills gap is really a confidence and quality gap. People are producing working tools they can't fully read or maintain, so the tool lives on one laptop and never reaches the team. The line that comes up again and again is a version of the same thing. I don't know what I don't know.
It has to be right every single time. The moment a tool faces managers, one wrong number can burn trust you spent years building. An AI mistake feels less forgivable than a human one. That's why so much good work never leaves prototype.
WHAT ACTUALLY SEPARATES THE TIERS
It's rarely natural talent or budget. It's mostly permission, a worked example from someone like you, and somewhere to ask the daft question without judgement. Notice the pattern in every workflow above. The quality comes first. The time savings follow.
You still can't fire and forget a pay decision, and nobody should.
 | That's the map as it stands in 2026. The workflows are the part I can show you. The how, the prompts, the setups, the worked examples that survive real data and real policy, is the part we work through together inside Range. If you're the only comp or reward person trying to figure this out, come and do it in good company. Giac |
|  |
|
When you're ready, here are three ways Range can help you.
1. Join the community
Find your people. Range is a private, vetted and capped community of comp and total rewards pros learning how to build with AI together. Apply to join →
2. Grab the free resources
The Comp AI Starter Kit gets you running with Claude for compensation in five steps. Get the kit →
3. Reward Rewired, 17th November 2026, London (UK)
In person gathering for the people engaged with figuring out where AI should live in reward today and where it shouldn’t (yet). Register your interest →