Practice · AI for HR (powered by Claude)

AI for your HR team — built, adopted, and kept.

You still need the subject-matter expertise. Claude amplifies the execution. I help HR teams put it to work on the jobs that actually eat their week.

I built and shipped this entire site with Claude — static HTML, no framework, no developer.

HR-team-scoped
Human in the loop
Risk-literate
The thesis

An HR operator builds this better than a generic AI consultant.

I build Claude against the work that consumes your HR team’s hours — turning a pile of HR data into an executive-ready story, running the math and rationale on every offer, cleaning the HRIS export nobody trusts, getting policies and JDs out the door. It’s HR-team-scoped, not an org-wide IT rollout: I’m not deploying Claude to 300 people across seven functions — I’m giving your team back the hours this work costs today.

The edge is knowing those workflows from the inside. I led global compensation for $760M+ in total direct compensation across 7,800 employees at Lucid Motors, and ran HR operations and total rewards through a nine-country public-company merger at Orthofix. I know where AI saves real hours — and where it quietly creates risk. A consultant who has never run a comp cycle or reconciled an HRIS can’t tell the difference.

Claude is the tool. The judgment about what to build, and what not to, is the practice.

Already have an AI seat?

Many HR teams already do — Copilot in Microsoft 365, maybe ChatGPT — and a blank chat box to go with it. The seat was never the hard part. The work here is Claude built against your real data, templates, and leveling framework: the actual job done, with the guardrails — plus the working tools and scheduled reports a sidebar assistant doesn’t give you.

Skills + surfaces

Build a Skill once. Your whole team runs it the same way.

Claude isn’t one app — it’s a set of surfaces, from Excel and Word to chat and an agent that runs multi-step projects. A Skill sits above them: encode a repeatable HR process once, and your team runs it the same way across the surfaces they already use. Everything below runs on the Team plan; a couple are still beta or preview, flagged on each card. Keep genuine PHI out of these surfaces.

The multiplier · Skills
GA

A Skill is a packaged process Claude loads on demand to do one HR task consistently — JD generation tied to your leveling framework, offer-analysis math, merit-cycle comms, board-report formatting. You build it once through a guided creator (no coding), share it across your team on the Team plan, and everyone gets the same output every time — across the surfaces your team uses every day: chat, Cowork, Excel, Word, and more.

Build once A Skill One repeatable HR process, encoded once.
drives
Claude chat Cowork Excel Word + more
Your whole team runs it the same way, every time.
C

Claude chat + Projects

Where you start: ask across your HR docs, draft, and summarize — and a Project holds your handbook as standing context for every answer.

GA
Most powerful for HR

Cowork

Not a chat window — a workspace. It works in your actual folders, reads many documents at once, and pulls org context from your historical files to build what you need end to end: a merit-comms package, a board pack, a full open-enrollment kit.

GANo PHI

Connectors

Pulls context from Drive, Gmail, Calendar, GitHub, Microsoft 365, and Slack directly, so you don't copy-paste.

GA
X

Claude for Excel

Build and audit comp models, headcount plans, and pay-equity checks right in the workbook.

GA
W

Claude for Word

Draft offer letters, policies, and job descriptions; redline handbook language in place.

GA
P

Claude for PowerPoint

Build leadership people-metrics decks and open-enrollment presentations from the data you already have.

GA
>_

Claude Code

For HR-ops analysts: light HRIS scripting, data-file cleanup, and automation. Ships on every Team seat.

GAFree on every Team seat

Scheduled reports

Recurring runs on a schedule — a weekly people-metrics report, or a pre-payroll data check that catches dirty records before they cause pay errors. Pulled for you, not by you.

Preview

Artifacts — live tools

Describe a tool and Claude builds it — a comp-band calculator, a headcount-and-attrition dashboard — shareable at a link your team opens and edits. Not a static spreadsheet emailed around.

GA in chatLive share: beta

None of this is a stack of apps to evaluate one by one — it’s one Claude Team workspace met where HR already works, and a Skill makes your team run it the same way across all of it.

Get the shareable Claude-for-HR guide

What I build

Concrete builds against real HR work.

Each build targets a specific job an HR team already does — and bridges into the other two practices.

HR time-suckThe Claude buildBridges to
Board & leadership reportingHR data → a written, board-ready people-narrative: walk into the meeting with the story, not just a chart dump — built in an hour, not a day.HR Systems
Offer analysis & justificationTooling that runs the offer math and drafts the candidate-facing rationale for TA.Total Rewards
Dirty HRIS dataReconcile and anomaly-check HR data on an ongoing basis, not just at migration.HR Systems
Writing & leveling job descriptionsA JD generator tied to your leveling framework — consistent, on-band JDs in minutes.Total Rewards
Annual merit commsManager talking points, FAQs, and letters generated for the whole cycle.Total Rewards
Tribal knowledge & SOPsTurn how-we-do-X into documented runbooks before the person who knows leaves.People Ops
Reports nobody has time to pullA weekly people-metrics report or a pre-payroll data check that runs on a schedule — ongoing monitoring, not a one-time cleanup.HR Systems

A Sprint bundles two or three of these against your real work. The result is hours returned to your team — and a team that can keep building after I leave.

The trust signal

What I won’t build — and why that protects you.

I won’t build an employee-facing HR assistant that answers LOA, FMLA, ADA, or benefits-eligibility questions unsupervised. A confident wrong answer harms an employee and creates liability for the company. Anything legally sensitive is configured to flag and escalate to HR or legal, with a human in the loop before a reply goes out.

The same line holds on the everyday work, not just the legally sensitive cases: Claude drafts the offer letter; the comp logic, the band, and the number are yours. It audits the spreadsheet; whether the pay-equity finding is actionable is your call. The expertise stays with the HR team.

An AI shop without HR judgment will happily ship the reckless version — an always-on bot that sounds authoritative and is sometimes wrong. I won’t. The licensing is HR-team-scoped anyway, so an org-wide employee bot isn’t the fit; the point is that I wouldn’t build it even if it were.

This is the protection you’re buying. Knowing where the legal landmines are — and building around them — is the difference between AI that helps HR and AI that creates a problem HR has to clean up.

The guardrail, step by step
01
A request comes inAn HR or employee question, or a task for Claude to draft.
02
Claude drafts the reply or runs the analysisA first pass — never the final word.
03
A human reviews — alwaysEvery build keeps the HR professional as the reviewer. Claude never owns what goes out the door.
If it’s legally sensitive — LOA, FMLA, ADA, benefits eligibility — it’s flagged and escalated to HR or Legal, never answered automatically.
04
The reply goes outReviewed, approved, and on the safe side of the law.
Adoption

Getting Claude is the easy part.

The platform is quick and safe to stand up — and that’s deliberately not where the work is. The build is.

For IT & Legal

It doesn’t train on your data

On the Team plan, your content isn’t used to train Claude’s models — by default and in the contract. That default only flips on the personal consumer plans. For the rest of IT and Legal’s questions, a Sprint produces a data-governance SOP — the same shape as the approval kit and data-classification framework in the library.

Same-day

It’s a self-serve setup

No procurement cycle, no IT project. Team is self-serve: sign up → add your team with SSO → set spend caps. You can be live this afternoon.

What it costs

A 13-person HR team — say 10 Standard and 3 Premium seats — runs about $500 a month, roughly $6,000 a year (about $38 a seat): the whole platform, bundled and predictable. Open the full calculator →

That gets you the platform. Turning it into the tools your team uses every week — built into your data, with the guardrails — is the Sprint.

How a Sprint works

In about two to four weeks I build two or three of these and train your team to keep them.

The Sprint is $12,000 fixed — $9,000 for orgs under 50 staff or nonprofits. Full pricing lives on the Engagements page.

01Pick the buildsWe choose two or three off the menu against your real work — the ones the assessment surfaced as worth doing first.
02Build into your contextI build each one into your actual data, templates, and leveling framework — and, where it fits, package it as a reusable Skill your team keeps. Not a throwaway demo.
03Add the governanceEscalation rules, human-in-the-loop checkpoints, data-classification scoping — plus a data-governance SOP your HR team can hand to IT and Legal to clear their security and compliance questions.
04Train your teamI train the people who’ll use it so the capability stays after I leave. The point is an AI-capable HR team, not a dependency on me.
After the Sprint

The Sprint stands on its own — it doesn’t have to be the last time we work together. Some teams keep me on a monthly retainer, so the builds stay current as your HR work evolves — across hiring, onboarding, comp, policy, and the HRIS underneath. Others bring me back on-call when a specific question lands. You own the work either way — I’m there when it helps.

Proof

I built all of this with Claude. Click into any piece.

I don’t describe what an HR operator can build with Claude. I built it and put it where you can use it.

Start here

Start with the free assessment.

Bring the work that’s eating your team’s week — the board deck, the merit cycle, the HRIS cleanup. We’ll find where Claude is a real fit, where it isn’t, and which two or three builds are worth doing first.

Book a 45-minute call The first one’s free.