Workforce Deployment

AI Worker Job Description Template: How to Define the Role Before You Deploy

Every failed AI deployment we have investigated had one thing in common: the role was never defined before configuration started. Without a job description, configuration becomes guesswork. The template below is what CC uses as the starting point for every AI worker deployment.

Why the Job Description Matters More Than the Technology

When you hire a human SDR, you write a job description first. You define what they are responsible for, what decisions they can make autonomously, when they escalate, and how you will measure their performance. You do not give someone access to your CRM and calendar before clarifying those things.

AI workers require the same clarity — but businesses skip this step because AI feels more like software than a hire. The result: an AI worker configured to do everything poorly rather than a narrow set of tasks well.

What happens without a JD

→Configuration covers every possible scenario instead of the highest-value ones

→Escalation rules are undefined — the AI either escalates everything or nothing

→KPIs are never established — there is no way to know if it is working

→Scope creep: the AI gets asked to do more and more until it fails visibly

→No clear owner for the AI worker's performance post-deployment

The AI Worker Job Description Template

Complete all five sections before configuration begins. This is not a technical document — it is a business document. A non-technical operator should be able to fill it out.

Section 1: Scope

□Job title (e.g., 'Inbound Lead Qualifier', 'Customer Success Agent', 'Content Writer')
□Primary channel (voice / web chat / email / SMS — pick one for v1)
□Hours of operation (24/7 or business hours only)
□Languages supported
□What this AI worker is explicitly NOT responsible for

Section 2: Core Workflows

□List the 3–5 specific tasks this AI worker will handle (ranked by volume)
□For each task: input trigger, output/action, and completion criteria
□Which tasks are fully autonomous vs. which require human approval
□What systems the AI worker reads from and writes to

Section 3: Escalation Rules

□Trigger 1: escalate when [condition] (e.g., 'customer mentions legal', 'deal > $50k')
□Trigger 2: escalate when [condition]
□Who receives the escalation (name, role, Slack channel, or email)
□What information gets included in the escalation handoff
□Max time AI worker works a case before escalating automatically

Section 4: Success Metrics

□Primary KPI (e.g., 'response time < 60 sec', '% of leads qualified per week')
□Secondary KPIs (2–3 supporting metrics)
□Baseline: what does this metric look like today without the AI worker?
□Target: what does good look like at 90 days?
□Review cadence (weekly / monthly)

Section 5: Guardrails

□Topics the AI worker must never discuss (competitors, pricing not in approved list, legal commitments)
□Tone/persona: formal, casual, brand-specific language
□What the AI worker says if it does not know an answer
□Data it can access vs. data it must never touch

Examples by AI Worker Type

Maya (Inbound Voice + Chat)

Scope

Qualify inbound leads via phone and web chat during business hours. Handle initial objections. Route qualified leads to calendar. Route unqualified to nurture email sequence.

Primary KPI

% of inbound leads qualified within 60 sec

Escalate when

Any mention of legal / contract / litigation; deal size signals > $100k; customer states 'I want to speak to a person'

Sage (Analytics + Reporting)

Scope

Weekly automated performance reports across Google Ads, HubSpot pipeline, and GA4. Deliver to Slack and email. Flag anomalies. No autonomous media buying.

Primary KPI

Report delivered by 8am Monday, anomaly detection rate vs. baseline

Escalate when

Any metric moving > 30% week-over-week; any attribution conflict

Atlas (Outbound Prospecting)

Scope

Research prospect companies in target ICP, draft personalized first-touch emails, enrich CRM records. No autonomous sending — drafts only.

Primary KPI

Drafts produced per week, reply rate on sent drafts

Escalate when

Prospect is a current client / active deal / known account

Handing Off to Configuration

The completed job description becomes the input to the configuration process. Every configuration decision should trace back to a JD field. If a configuration question is not answerable from the JD, the JD is incomplete — revise before configuring.

JD → Configuration handoff checklist

✓JD is approved by the business owner (not just the technical implementer)

✓All 5 sections complete — no blank fields

✓Success metrics have a documented baseline

✓Escalation recipient has confirmed they are the right owner

✓Guardrails reviewed by legal or compliance if applicable

FAQ

How long should completing the JD take?

For a single AI worker with a focused scope (e.g., inbound lead qualifier), 1–2 hours for a business owner who knows their process. If it takes longer, the scope is too broad — narrow it before proceeding.

Can we start with a simpler version and expand later?

Yes — and this is the recommended approach. Deploy one workflow well (e.g., Maya only handles inbound web chat leads, 9am–5pm, qualification only). Expand scope after 30 days of data. Expansion is cheaper and lower-risk than building too broad on day one.

Do we need a separate JD for each AI worker?

Yes. Each AI worker deployed should have its own JD even if deployed on the same plan. The JD governs the operational role — one AI worker can have multiple workflows within its JD, but the scope, KPIs, and escalation rules should be specific to that worker.

Start my AI worker deployment →

Ready to stop doing this manually?

We map your workflows, deploy the right AI Worker, and guarantee the math pencils out before you sign.

Start my AI worker deployment →Back to Workforce Deployment