Section A · Orient

Positioning From Scratch

How to interview honestly when one half of your background — AI agents or HR-tech — is strong and the other half is thin. The posture, the language, and the moves that work.

The trap

This role wants four things in one person: senior software, applied AI/agentic systems, HR/People-tech, and Workday at depth. Most candidates have two of the four cleanly. The trap is trying to simulate the missing two. That fails for two reasons:

The interviewers detect bluffing in 90 seconds. Whoever runs the AI portion has built agents; whoever runs the Workday portion has spent years inside a tenant. A vague answer is a tell.
The bluff-failure mode is much worse than the gap. An honest gap is a hire-with-ramp signal. A caught bluff is a credibility loss the rest of the loop can't recover from.

The single mindset shift

Your job in the loop is not to prove you have all four. It's to demonstrate the two you have cleanly, signal you understand the two you don't, and show the velocity you've already applied to close the gap.

The two archetypes who interview for this role

Roughly, candidates fall into one of two camps. Identify which one is you before the loop.

	Archetype A: AI-strong, HR-light	Archetype B: HR-strong, AI-light
Background	ML engineer, applied AI engineer, agent builder. Maybe shipped a RAG product, maybe an MCP server, maybe production LangGraph.	People Engineering, HRIS administrator, Workday consultant, People Analytics lead. Has touched Workday Studio or EIB in anger.
Strength shown easily	Tool-use patterns, eval design, prompt caching, MCP architecture, model selection.	Workday business processes, EIB schemas, payroll-calendar awareness, HR vocabulary, jurisdictional nuance.
Visible gap	Workday and HR-domain depth.	Recent agentic AI production work.
What you must show	You've taken the JD's Workday list seriously and can talk about it cleanly even if you haven't shipped it.	You've shipped at least one agent with HITL gates, structured outputs, and basic evals — even if it's a side project on top of Workday sandbox data.

Archetype A — AI-strong, HR-light

Your interview is mostly about whether you respect the HR domain enough to be trusted with payroll-adjacent automation. The failure mode is sounding like an AI tourist.

Language that works

"My production depth is on the AI side — agents, tool use, evals, MCP servers, prompt-cached long-context workflows. I'd be ramping on the depth of Workday's business-process model and the specifics of payroll calendar integration in my first 90 days. I've spent the last [N weeks] going deep on Workday's worker model, calculated fields, EIB flows, and the integration patterns (Studio, Core Connectors, web services) so I can speak the language with HRKX from day one, even before I've shipped my first build."

Pivot moves when asked about Workday specifics

"I haven't configured a business process condition rule in production. My understanding is [X]. Want me to reason about how I'd approach [the specific case they asked], or is there a related thing you'd rather dig into?"
"My closest experience is [adjacent — Greenhouse API integration, BambooHR, manual SQL on a People warehouse]. From that I'd extrapolate [Y]. Tell me where that extrapolation breaks down for Workday."

Don't do this

Anti-patterns

Don't claim "I've worked with Workday" if you've only read the docs.
Don't describe an EIB flow you've never run. Say "I understand the shape conceptually but haven't run one in anger."
Don't invent calculated-field syntax. Say "I'd write that as a calculated field but I'd reach for the reference for the exact function names."

Archetype B — HR-strong, AI-light

Your interview is mostly about whether your AI work is real or PowerPoint. The failure mode is "we love AI" without having shipped anything.

Before the loop — build one thing

If you have any time, ship a small thing. Concretely:

A Python script that calls Claude with tool use to read a Workday-shaped CSV and propose a JSON action plan for a hire scenario.
An MCP server that exposes a fake worker-record store. Wire it to Claude Desktop. Watch the agent navigate it.
An n8n workflow that reads from a survey CSV, calls Claude to cluster comments, posts the summary to Slack with an approval step.

This isn't to fake AI experience. It's to have a concrete answer when they ask "what have you built lately?"

Language that works

"My depth is on the HR ops side — I've owned Workday business processes, EIB cycles, and the calendar coordination with Payroll. The AI architecture work is where I've been ramping. In the last [N weeks] I've built [the small thing above] to internalize the agent loop, MCP, structured outputs, and HITL patterns. I'd be the architect who is already fluent with the operational constraints; I'd be growing depth on the AI side, and I have a concrete plan for how I'd close it."

Pivot moves when asked about AI specifics

"I haven't shipped a multi-agent system in production. My side-project agent uses [tool use / MCP / structured outputs]. Walk me through what 'multi-agent' would look like in this org's context — I want to make sure I'm answering the right question."
"My eval experience is informal — I check accuracy on a labeled set I made. I know production-grade evals look different. What does your eval stack look like today?"

Your one-paragraph self-summary

Memorize one paragraph. Deliver it in 45 seconds, conversationally. Structure:

One sentence: who you are right now.
One sentence: the two strongest threads.
One sentence: what you've been pulling toward (this role).
One sentence: why this role specifically.

"I'm a [generalist engineer / People Engineering lead / applied AI engineer] with [N] years in software and [M] focused recently on [agentic systems / HRIS]. The two threads I keep coming back to are AI tool-use and the operational craft of running HR systems that touch payroll. I've been pulling toward roles where AI is doing real work against regulated employee data, with the human gates that implies. This role is exactly that — full-stack from opportunity to governance, Workday at the core, and the bar is 'no autonomous payroll-adjacent action,' which is the bar I'd want to be held to anyway."

Pre-loaded stories (3 to have ready)

You will be asked some version of these. Pre-write a 90-second answer for each in your own words.

A time you said no to a build (or "yes, but smaller"). Tests the halt-or-redesign muscle. Show the named risk, the proposed phased path, and the stakeholder loop-in.
A time you made a non-technical team self-sufficient. Tests "multiply the team's capability." Show the runbook, the training, the handoff metric.
A time something went wrong in production with people data. Tests judgment under stress. Show the detect → contain → fix → root-cause → prevention loop. Don't pretend you've never had an incident — the absence is the tell.

Honesty without apology

The pattern for any "have you done X" question where you haven't:

The structure

Name the gap once, plainly. "I haven't shipped X in production."
Anchor to the closest real reference point. "My closest experience is Y."
Offer a path forward. "Want me to reason about how I'd approach X from first principles, or pivot to the part I've done?"

Three sentences, no apologizing, no excuses, no "but I'm a fast learner." The fact that you didn't flinch is the proof you're a fast learner.

Never bluff Workday specifics

Workday vocabulary is dense and the interviewers will spot fakery instantly. Things you should know cleanly enough to talk about even if you've never touched the system (covered in 05-deep-dive-secondary and 14-domain-context):

The difference between Position, Job, and Job Profile
Supervisory organizations
The Worker object and effective-dated changes
Business process basics (initiation, approvals, sub-process, condition rules)
EIB (Enterprise Interface Builder) — what it is, when you use it, the file shape
Calculated fields (concept, not exact syntax)
Integration patterns — Studio, Core Connectors, RaaS, web services (SOAP/REST)
Why payroll changes have a calendar lock window

If you can speak these 8 cleanly, you've cleared the Workday-fluency bar even without hands-on. If you can't, do not fake it.

Confidence as posture, not content

Confidence in a senior interview is not "I know everything." It's:

Calm pace. You don't rush.
Owning gaps cleanly. You name them and move on.
Curious questions. You ask interviewer questions that signal you've thought about the work.
Comfort with silence. You don't fill thinking time with words.
Returning to structure. When asked a vague question, you frame ("Two things matter here…") before diving.

A closing line worth memorizing

For the final question of any round — "do you have anything you'd like to add?" — have a single closing line ready. Not a sales pitch; a confirmation of fit.

"Just to close the loop — the part of this role I find most interesting is that the constraints (payroll-adjacent HITL, employee PII, cross-jurisdictional privacy) force good architecture. I'd rather build AI in that environment than somewhere autonomous demos are the bar. If you decide to bring me back, that's the work I'd want to do."

Move on to 03-core-fundamentals when your posture and language for archetype A or B is solid in your own voice — not memorized, internalized.