Every AI tool demos well. The demo is rehearsed; your desk is a live year. So evaluate the way you'd evaluate a candidate: ignore the interview polish and check the things that don't flex. Eight questions do it.
1. Does candidate data come included?
Start here because it decides everything downstream. A tool without a candidate database can summarize, draft, and organize, but it cannot put a new name on your call sheet, so sourcing still lives in a separate stack you pay for and switch into all day. What good looks like: search built in, described in plain English, with contact enrichment you control. RecruiterClaw searches 400M+ professional profiles from a sentence typed in Slack, hands back a ranked shortlist in about 90 seconds, and enriches verified emails and direct phones only for the candidates you pick. The full method is in our guide to sourcing passive candidates with AI.
2. Is the pricing flat or metered?
Meters tax habits, and habits are the whole point. If every task, message, or scheduled run spends from a credit balance, you will hesitate exactly when you should be exploring, and your busiest months will be your most expensive. But be suspicious of "unlimited everything" too, because verified contact data costs real money and someone is paying for it. The honest answer states one limit plainly. RecruiterClaw is flat rate: conversations, tasks, and scheduled automations are unlimited, and only candidate sourcing and contact enrichment draw from an allowance, which is sized generously. One placement more than covers the year.
3. Can it damage your ATS?
Your ATS is the accumulated value of your firm: every candidate you've ever met, every note from every call. Before any tool gets write access, ask what it is structurally unable to do. "Our agents are careful" is a hope. "Delete is impossible by design" is an architecture. The standard to demand: additive-only writes, meaning create, note, add to pipeline, and update, with delete, merge, and archive off the table entirely, plus approval modes you choose, from read-only to confirm-each to trusted. That is how RecruiterClaw ships, with Loxo included out of the box and any API-capable ATS wired in during install.
4. Where does your client data live?
A recruiting desk runs on confidential information: fees, offer numbers, who is quietly about to resign. Ask whether that sits on your machine or on the vendor's cloud, and if it's the cloud, who else can reach the connected accounts. RecruiterClaw's answer is local: memory, client profiles, conversation history, and candidate files all live on your machine, credentials sit in your encrypted keychain, and RecruiterClaw operates no servers holding client data. You own your data. We don't.
5. Will outreach sound like you?
Candidates and clients answer you, not a template, and AI mail that reads like AI mail burns the very trust you're paid for. Ask two things. First, what do the drafts learn from? The good answer is your own real sent mail. Second, what stands between a draft and a send? RecruiterClaw will not send anything until you type an exact confirmation phrase, and deleting email is impossible by design because the permission is never even requested. Nothing goes out on its own.
6. Does it remember your desk?
A tool with no memory makes you the memory: re-explaining clients, fees, and history every session. Ask what it retains and whether it acts on that memory on a schedule. The reference cadence looks like this: a morning brief at 7 AM with email triage, calendar, pipeline alerts, and deals needing a touch at 7 and 14 days; daily wins at 5 PM; a weekly roll-up on Friday. Ask whether it coaches too, the way RecruiterClaw's Field Manual does when a counteroffer lands. That standing rhythm is the difference between an assistant that answers questions and a chief of staff that runs the desk; we draw that line properly in what is an AI chief of staff.
7. Who actually sets it up?
Most recruiting firms have no IT department, and solo recruiters are the IT department. Ask who does the install, how long until it's earning, and what happens when something breaks. RecruiterClaw's setup is white-glove: a 30 minute discovery call, an onboarding call that goes deep on your desk, clients, and niche, then a remote install session where we drive and you watch. You are live within 48 hours and never touch a terminal. There's more on why this matters for small firms in AI for solo recruiters.
8. What happens when you leave?
The quietest question and the most revealing one. If your notes, memory, and candidate files live on a vendor's servers, leaving means exporting whatever they let you export, on their timeline. If everything lives on your machine, leaving is just stopping. RecruiterClaw is built the second way: there is nothing to request and no one to ask, because the data never left your machine in the first place. Any vendor's answer to this question tells you how they think about who owns your desk.
What the checklist won't tell you
No tool closes your candidate for you. The judgment layer, reading a client, handling the hard conversation, the trust that wins retained work, stays yours; the tools compete to take the admin layer off your plate. That is the honest version of the replacement question, and we take it head-on in will AI replace recruiters. So run all eight questions, then judge the vendor the way you judge a candidate: not by the interview, but by the references and the work.
What is the first question to ask an AI recruiting tool vendor?
Whether candidate data comes included. An AI tool without a candidate database can organize your work but cannot source, so the core of the job still happens somewhere else. RecruiterClaw includes plain-English search across 400M+ professional profiles with a ranked shortlist in about 90 seconds.
Why does the pricing model matter so much?
Because meters change behavior. When every task or scheduled run spends from a credit balance, you start rationing the daily habits that make the tool worth having. Look for flat pricing with limits stated honestly: RecruiterClaw meters only candidate sourcing and contact enrichment, because verified contact data costs real money, and the allowance is sized generously.
What does ATS safety actually mean?
That the tool is structurally unable to hurt your database. The standard: writes are additive only, meaning create, note, add to pipeline, and update, while delete, merge, and archive are impossible by design, with approval modes you choose from read-only to confirm-each to trusted.
How do I know AI outreach will sound like me?
Ask what the drafts are learned from. The good answer is your own real sent mail, not a generic template library. Then ask what stops a send: nothing should go out until you approve it, and RecruiterClaw requires you to type an exact confirmation phrase before any email sends.
What does white-glove setup mean?
A human team does the installing. With RecruiterClaw that is a 30 minute discovery call, an onboarding call that goes deep on your desk, clients, and niche, then a remote install session where we drive and you watch. You are live within 48 hours and never touch a terminal.