.svg)
Using context “artifacts” to unlock AI for compensation
September 30, 2025 • 6 Min Read
But real change comes when you treat AI like a teammate, armed with context, engaged in conversation, and looped into your workflow. That’s where artifacts come in.
Let’s break it down.
Access to AI is everywhere.
Fluency is rare.
In a recent Compa Flash Poll, 90% of professionals said they have access to foundational models like ChatGPT, Claude, or Gemini. Yet nearly three out of four—73%—still rely only on ChatGPT.
AI mandates don’t equal mastery
Compa’s recent Flash Poll found that 70% of compensation teams have AI adoption mandates.
But their use remains narrow and focused on drafting docs, summarizing, and light ideation.
Fluency comes from using AI beyond work tasks
Compensation leaders use AI 2–4 hours a week for work and 1–2 hours for personal projects.
To increase fluency, practice on personal projects, not just work tasks.
How context artifacts
unlock AI
The core idea:
Feed your AI tools real-world context and build something together.
What’s an artifact?
A doc, image, spreadsheet, Slack thread, video—anything with substance.
Artifacts add depth to the conversation and turn ChatGPT into your analyst, Claude into your dashboard designer, and Gemini into your research assistant.
Building with artifacts, step by step
1.
Ingest Artifact
Drop your doc, image, or dataset into a Claude project (or ChatGPT thread).
2.
Analyze it
Start messy, ask questions, and don’t over-polish.
3.
Build a prototype
Turn analysis into dashboards, frameworks, or planning docs.
4.
Feed it forward
Every output becomes the next input, that’s the artifact chain.
In a recent Compa Flash Poll, 90% of professionals say they have access to foundational models like ChatGPT, Claude, or Gemini. But most default to just ChatGPT—73%, in fact.
3 ways context artifacts help comp teams unlock AI
Meta AI poaching blitz
Artifact:
Reddit data dump of 44 hires at Meta’s new AI team
Goal:
Build a retention risk framework
Output:
- Skills and role dashboard
- HRBP playbook
- Manager-facing retention scoring app
- CEO-ready risk report
What worked:
- Claude’s Canvas to create visual dashboards
- Project-based memory to stack insights
- Artifact chaining: each step led to the next idea
EU pay transparency directive
No starting artifact? No problem, you can build them from scratch.
Steps:
- Use Gemini for deep research
- Summarize in Claude
- Generate a strategic roadmap
- Draft a product requirements doc (PRD)
- Feed that into a no-code tool (e.g. VibeCode)
- Walk into the next meeting with a working prototype
Why it matters:
You go from “What is the directive?” to “Here’s the portal we’ll need in 12 months” in hours, not weeks.
Bonus program forecasting
Problem:
Bonus accruals use flat assumptions (100% payout).
Solution:
Monte Carlo simulations with LLMs.
Artifact:
Simple bonus spreadsheet
Tool:
Claude Canvas
Output:
Dynamic, simulation-based projections for finance.
AI tools aren’t just writers, they’re simulators, analysts, and app builders.
What used to take hours now runs in seconds.
Using agents in comp isn’t a productivity boost; it’s a second team.
One that doesn’t sleep and always enforces policy.
4 challenges comp teams face with AI and how to beat them
Barrier
Challenge
Solution
Systemic
IT, legal, security bottlenecks
Use personal licenses and start with public use cases.
Financial
Paid tools = better tools
Invest $20/month. ROI comes fast.
Functional
No time to learn new tools
Automate your busywork. Reinvest saved time.
Social
Org isn't ready
Become the pioneer and teach by showing.
Tools to kickstart your AI workflow
Deep research
Gemini
Ideation, drafting
ChatGPT
Dashboard + context
Claude
Prototype tools
VibeCode, Typedream
Notes → Artifacts
Otter, Notion
Visual storytelling
Excalidraw
What now?
You don’t need more prompts, you need better context, stronger ideas, and faster iteration.
That’s what context artifacts unlock.
Discover the future of compensation
When the market moves fast, you can't rely on surveys. Remove uncertainty and gain confidence with Compa.
.webp)
.webp)
.webp)
