AI tooling has gotten cheap and capable enough that a lot of sellers are asking whether virtual assistants are still worth the payroll line. This interview between two long-time wholesale/arbitrage operators (9+ years selling, 500+ VAs placed) works through where each actually earns its keep — and the cash-flow math on when to hire at all.
The core framing: neither replaces the other
- Treat VAs as remote staff, not robots. A human piling through hours of manual comparison work suffers attention fatigue and output quality drops; AI does not.
- Conversely, AI confidently produces wrong answers. Someone with domain knowledge — who already understands FBA, reselling, and your supplier quirks — needs to be the one saying “that’s not right, we don’t do it that way.”
- Rule of thumb from the discussion: repetitive, high-volume data tasks → automate. Creative work, opportunity spotting, supplier relationships, judgment calls → human.
- Garbage in, garbage out applies equally to both. Vague instructions produce a useless VA and a useless AI workflow.
Where AI genuinely replaced manual hours
Concrete examples the operators shared from their own businesses:
- Buy sheet vs. warehouse sheet reconciliation. Checking that purchase data copied correctly and that supplier confirmation emails match ordered quantities used to take 1–2 hours a day. AI does a first pass in minutes; the VA spot-checks.
- Invoice retrieval. A VA mapped out that there were effectively five distinct ways invoices arrived (supplier portal download, email request, auto-attached, email link, one custom case). Once documented, the process became automatable more or less permanently. The mapping took a VA ~20 hours for a 5-hour job — worth it once.
- Deduplication during sourcing. A small internal tool that flags whether a product has already been reviewed (and how recently) stops the VA re-checking the same SKUs daily, roughly doubling products reviewed per hour.
- Capital-aware sourcing. Mapping which suppliers accept which payment method, so the sourcing VA only hunts brands the business can actually pay for this week rather than surfacing good deals that can’t be funded.
- Automated Amazon-to-Amazon flip bots generating hundreds of leads daily — far past human volume, though most sellers are capital-constrained long before they’re lead-constrained.
The cash-flow test before hiring
- Sourcing VAs fill the bath (generate revenue). Admin VAs plug leaks (reimbursements, unpaid invoices, repricing misses). If you can only pick one, pick the tap over the plug.
- You can afford a VA’s wage and still not afford a VA. If they surface deals you have no capital to buy, you’ve bought frustration.
- The suggested threshold: don’t put ~$500/mo into a VA unless you’re clearing meaningfully more than that in free cash flow (~$2,000/mo was the number floated). Below that, spend on AI tooling and speed up your own process first.
- Even at $300–500k/month revenue, capital stays tight — bigger orders eat bigger amounts of cash. It doesn’t get easier, it gets bigger.
Implementation notes worth stealing
- Don’t buy tools without an objective. It’s easy to spend a night writing a Python script that reads your emails when you could have just read your emails. Repetitive task → build it. One-and-done → just do it.
- Make two lists: tasks you’re doing that add no value when you do them, and tasks that would generate revenue if someone else owned them. The second list is the job description.
- Shared AI access beats per-seat seats. One team wired their AI into their internal chat (Slack/Google Chat) as a bot the whole team can query, rather than buying $200/mo subscriptions for everyone.
- Miscommunication is the real failure mode. One example: a new WMS rollout produced duplicate SKUs because “one SKU per product” was never stated explicitly. Poor briefing breaks humans and machines identically.
- Human ownership of metrics still matters. A daily end-of-day report from the admin VA on units shipped into FBA could have been automated — but the point was making a person accountable for pushing that number.
- The “dad’s dead test”: if the owner disappeared for a month, does the business keep running? A buyer isn’t purchasing your deal list; they’re buying documented processes.
How Squatio helps
Most of the pain above is really about where a human’s limited hours should go. The answer is judgment work — but judgment needs good inputs, and that’s where tooling should carry the load:
- Instead of a VA (or an AI script) grinding through supplier catalogues product by product, pre-filter the universe with Prospect (Product Database) so the shortlist your team reviews is already screened on your criteria.
- Before committing capital to a brand or category, Cortex (AI niche & category analysis) handles the category-level read that would otherwise be a multi-hour manual research task.
- For the per-ASIN verification step in any sourcing SOP, Squatio Decode (ASIN lookup, listing visibility & Sentiment IQ review analysis) gives your VA a consistent place to check listing health and review sentiment rather than assembling it from five browser tabs.
- On the demand side, Neuron (AI keyword search & interpretation) and Reveal (Reverse ASIN) turn keyword validation into a repeatable step a delegated team member can run without needing your judgment on every call.
- If your VA also owns advertising, Squatio Optimize (PPC intelligence) surfaces the campaign signals worth acting on so their time goes to decisions, not data assembly.
If you’ve split work between a VA and AI, which task surprised you most by not being worth automating?