The Practical AI Checklist for a Two-Person HR Team

If your HR function is you and one other person, you have probably already heard some version of "you need an AI strategy." Most of that advice is written for organizations with a Chief People Officer, a data science team, and a legal department on speed dial. None of that applies when there are two of you covering hiring, benefits, performance, compliance, and whatever fire is burning that week.

The good news is that a two-person team does not need an AI strategy. It needs a short list of decisions, made once, that keep you out of trouble and free up real hours. Here is what actually matters.

1. Find out what is already AI before you buy anything new

Your ATS, payroll provider, and scheduling tool have probably shipped AI features in the last year, resume ranking, chatbot screening, automated scoring, without much fanfare. Before evaluating a new tool, spend an afternoon checking what your existing stack already does. You may be using AI in hiring decisions right now without a policy for it.

2. Separate decision support from decision making

AI is genuinely useful for summarizing feedback, drafting a first pass at a review, or surfacing patterns across survey responses. It should not be making the call on who gets hired, promoted, or let go. Draw that line explicitly and hold it, especially under deadline pressure, when it is tempting to let a tool's recommendation stand in for a human review.

3. Know the compliance floor, even at your size

Laws like NYC's Local Law 144 and guidance from the EEOC on Title VII, the ADA, and the ADEA already treat AI hiring tools as subject to existing anti-discrimination rules, and a growing list of states are adding their own disclosure and audit requirements. Company size does not exempt you if you are recruiting candidates in those jurisdictions. Ask every AI vendor for their bias audit results and a plain-language explanation of how the tool scores people. If they cannot produce one, that is your answer.

4. Build the paper trail as you go, not after an audit request

For any AI-assisted decision that touches a real person, hiring, review ratings, terminations, keep a short record of what the tool suggested and what a human reviewed or changed. This does not need to be elaborate. It needs to exist before a regulator, a candidate, or your own legal counsel asks for it.

5. Pick one high-friction task, not a platform overhaul

With two people, you do not have the bandwidth to roll out an AI initiative. You have bandwidth to fix one recurring headache: synthesizing open-ended survey comments, drafting first-round review language from manager notes, or turning a stack of 1:1 notes into something you can actually act on. Start there, prove it saves time, then decide if it is worth expanding.

6. Get managers on board before employees see anything

In a lean HR setup, managers are your delivery mechanism. If a new AI-assisted process adds a step to their week without clearly saving them time, it will quietly die, or worse, get used inconsistently across teams. Pilot with one or two managers first and ask directly whether it made their job easier.

7. Revisit the setup every quarter

Models change, vendors ship new features by default, and what counted as low-risk six months ago may not still be. A quarterly ten-minute check, what is the tool doing now, has anything changed, are we still comfortable with it, is enough to keep pace without turning this into a project.

None of this requires a data team or a big budget. It requires deciding, in advance, where AI earns its keep and where a human stays firmly in charge. That is also the design principle behind tools built specifically for small HR teams: platforms like Peoplelogic lean on guided setup and templates rather than sprawling configuration, so a two-person team gets the insight without taking on a second full-time job just to run the software.

The teams that get this right are not the ones with the most AI. They are the ones who decided, early and on purpose, exactly where it fits.