Human Resources — Carol Fraser, SCP, ACC, CCP

Republished with permission from the FrUn SMB AI Guide. Canonical chapter: Superhuman Docs → · All chapters

Carol Fraser, SCP, ACC, CCP · Founder, C4 Talent | Fractional HR Executive · LinkedIn

1. Mindset — What an SMB Should Believe Before Spending a Dollar on AI

The most expensive belief an SMB can hold is that AI can replace HR judgment.

It can draft, summarize, and organize — but HR is risk, relationships, compliance, culture, and trust. Use AI to prepare a first draft of an onboarding checklist; do not use it to decide a termination, a classification, or an accommodation.

The rule worth writing on a sticky note: use AI to reduce the blank-page problem, never to avoid the human-responsibility problem.

Where AI actually helps in HR:

• First drafts of job descriptions, onboarding checklists, interview guides, and policy FAQs
• Organizing scattered HR knowledge into something findable
• Helping managers prepare for hard conversations — structure and talking points, not the decision itself
• Synthesizing what you already know into clearer documentation

Where it merely feels productive:

• Using AI to decide terminations, classifications, or accommodations
• Generating HR policy or employment documents and treating the output as finished
• Producing volume — ten versions of a handbook nobody will read — when the business hasn’t decided what it actually needs

What has to be true before AI pays off:

• Basic HR discipline: defined roles, a document system for policies, and clarity on which decisions require human judgment
• A process clear enough that AI can help you articulate it — if your process is unclear, AI will help you create polished confusion faster
• Willingness to review every output before it touches an employee, a candidate, or a compliance question

Get those three in place and a modest tool stack will outperform an enterprise HR AI suite in the hands of someone who skipped them.

2. The SMB Tool Stack — Start Cheap, Stay Simple

Keep it simple. HR AI for an SMB is not a separate product category — it’s a general assistant, the document system you already have, and a payroll platform that keeps employee data accurate.

Three to start:

• A secure general AI assistant ($0–30/user/mo). For drafts and synthesis: job descriptions, interview guides, onboarding checklists, policy FAQs, and manager prep for difficult conversations. Use a business or team tier for anything sensitive — not the free consumer version.
• Your existing document system (Drive, SharePoint, or Notion). This is your HR knowledge base. Policies, templates, onboarding plans, and interview packets live here — findable, versioned, and owned by someone on the team.
• A basic HR/payroll platform (Gusto, Justworks, BambooHR). Keeps employee data accurate and gives AI something coherent to work from. AI cannot fix a messy employee record any more than it can fix a messy chart of accounts.

Skip or postpone:

• Enterprise AI HR suites — priced and scoped for companies with dedicated HRIS admins
• AI recruiting tools before your hiring process is clean — automation on top of a broken process just ships bad hires faster
• Surveillance and productivity-scoring tools — they erode trust and create compliance risk an SMB cannot afford
• AI policy generators with no legal review — employment law is state-specific; a polished wrong policy is worse than no policy

Zero budget:

A free general AI tool, used only with non-confidential information, plus your existing document system. That combination costs nothing and still gets you past the blank page on job descriptions, interview guides, and onboarding checklists — as long as a human reviews everything before it goes live.

3. The First 30 Days — Where to Start

Start with AI-assisted hiring and onboarding — where SMBs lose the most time and money. Don’t try to AI-enable your entire HR function in 30 days. Pick one open role (or your next hire) and run the playbook on that one workflow.

Week 1: Clean up one job description

• Pick the role you’re hiring for (or will hire for next)
• Use AI to draft or refresh the job description — then review it like a human: right title, realistic requirements, exempt vs. nonexempt classification
• If AI can’t write a clean job description, the business may not have made the role decisions yet — that’s a signal, not a failure

Week 2: Build a structured interview guide and scoring rubric

• Ask AI to draft an interview guide based on the job description
• Add a simple scoring rubric so every interviewer evaluates candidates on the same criteria
• Human review: does this guide actually predict success in the role?

Week 3: Create the offer and preboarding workflow

• Map who owns each step from verbal offer to day one
• Use AI to draft the preboarding checklist (equipment, access, first-day agenda)
• Do not let AI draft the offer letter — employment terms need human and legal review

Week 4: Build the 30/60/90-day onboarding plan

• Draft the plan with AI: what the new hire learns, who they meet, what success looks like at each milestone
• Assign named owners to each section — a plan without owners is a document, not a process

What “good” looks like at day 30:

• One clean job description, human-reviewed
• An interview packet with a scoring rubric
• A preboarding checklist with clear ownership
• A 30/60/90-day onboarding plan with named owners
• Nothing published or sent to a candidate without a human sign-off

4. A Real SMB Use Case

The pattern most SMBs need first: hiring and onboarding where the blank-page problem eats hours and inconsistency costs good candidates.

What was broken: Job descriptions lived in email threads or were rebuilt from scratch every hire. Interviewers asked different questions and had no shared way to compare candidates. New hires arrived to a first day with no plan — whoever was available showed them around. Offer letters and employment documents were ad hoc.

What AI did: Drafted a clean job description from the owner’s notes about the role. Built a structured interview guide and scoring rubric. Produced a preboarding checklist and a 30/60/90-day onboarding plan with named milestones.

What the humans did: Reviewed the job description for the right title, realistic requirements, and exempt vs. nonexempt classification. Owned the offer letter and every employment term. Assigned named owners to each onboarding step. Kept terminations, classification decisions, and accommodations completely off the automation path.

Tools used: A secure general AI assistant (~$20–30/user/mo), the existing document system, and the HR/payroll platform already in use.

Result: One hiring workflow that runs end to end with AI on the drafts and humans on the decisions. Interview consistency improved because everyone used the same guide. New hires had a plan from day one. The owner stopped rebuilding the same documents from scratch every hire — and nothing went to a candidate or employee without review.

The part to take from this isn’t a specific tool. It’s the pattern: AI handles the blank page; humans handle the responsibility.

5. Guardrails — What an SMB Should NOT Do

Every guardrail here is the same idea in different clothes: HR is where trust, compliance, and people’s livelihoods intersect. AI drafts; humans decide.

The mistakes I see most:

• Pasting confidential employee data into free consumer AI tools. If you would not put it in a public document, do not put it into a free AI prompt. Names, compensation, performance notes, medical information, and investigation details stay out of consumer tiers — or get stripped before they go in.
• Using AI-only HR decisions. Terminations, classifications, accommodations, and compensation changes always get a human checkpoint. AI can help you prepare; it cannot own the call.
• Letting AI draft offer letters, handbooks, or termination notices without review. Employment law is state-specific. A polished wrong clause is one of the most expensive errors a small business can make.
• Automating before the process is clear. If your process is unclear, AI will help you create polished confusion faster. Fix the decisions first; then automate the drafting.

Keep humans on these, no matter how good the tools get:

• Offer letters and employment agreements
• Terminations and performance-related separations
• Exempt vs. nonexempt classification decisions
• Accommodation and leave decisions
• Anything involving personal employee data, health information, or investigation details
• Policy changes that employees will rely on

Privacy and data issues SMBs underestimate:

• Free consumer tiers may train on your inputs — use business or team tiers with clear data terms for anything sensitive
• Shadow AI: your team is already pasting company information into tools you’ve never approved. Give them a sanctioned option and a clear rule before a breach happens
• AI output is not legal advice and not a substitute for HR counsel on employment matters

6. Lessons Learned + One Thing You’d Tell Every SMB Owner

Hard-won lessons from using AI in HR work:

1. AI is excellent at helping you start, not at knowing when you’re done. The draft is the beginning; your review is the deliverable.
2. The better your input, the better your output. Garbage in, confident garbage out — especially dangerous when the output looks finished.
3. AI exposes operational mess. When it can’t write a clean job description, the business may not have made the role decisions yet. That’s a gift, not a failure.
4. It helps managers prepare for hard conversations — structure and talking points — but the conversation itself is always human.
5. It saves time only if you resist overproducing. Ten versions of a policy nobody asked for is not a win.

And the one thing, if you remember nothing else:

Use AI to make HR work clearer, faster, and better organized — and use human judgment for the decisions that affect people’s jobs, pay, dignity, safety, and trust.

Part of the FrUn SMB AI Guide. Read on canonical guide →