"Should I hire a VA or use AI agents?" is the most common question I get from sellers under $500k revenue. The honest answer is "both, for different jobs." This page breaks down the actual cost per task, where each wins, where each loses, and the hybrid model most successful sellers settle into.
VAs win for judgment-heavy work (supplier negotiation, customer disputes, brand voice decisions). AI agents win for repetitive, multi-step, well-defined work (listing optimization, image generation, PPC analysis). The hybrid stack uses AI for execution work and humans for judgment work. Don't replace your VA with AI. Have your VA review AI output.
Let's start with concrete pricing. As of mid-2026. VA labor-rate references cross-checked against industry coverage from About Amazon on Seller Assistant cost savings and reporting at EcommerceBytes.
These numbers matter because the unit economics differ wildly.
Three concrete tasks, both approaches priced out.
| Approach | Monthly cost range | Strengths | Weaknesses | Best for revenue tier |
|---|---|---|---|---|
| AI tools only | $50 to $500 | Speed, consistency, scale. Cheap per run. No PTO, no onboarding. | Bad at supplier calls, escalations, judgment-heavy work. Needs human review. | Under $250k a year. Side-projects. |
| Filipino VA only | $700 to $1,500 (full-time at $4 to $10 an hour) | Judgment, relationship work, anything legally binding. Owns context across tasks. | Slower at bulk repetitive work. Sleep cycle and PTO. Bandwidth ceiling. | $100k to $500k a year. Multi-product launches. |
| Hybrid (VA + AI tools) | $1,000 to $2,500 | VA reviews and ships AI drafts. Best of both. Scales without hiring a second VA. | More tools to manage. Requires standard operating procedures so the VA and the AI don't step on each other. | $250k a year and up. Most serious sellers settle here. |
Talking to a Chinese supplier about a price reduction, a delivery delay, or a quality issue. AI doesn't have the cultural context, the relationship history, or the negotiation instincts. This stays human.
A frustrated buyer threatening a negative review. A return dispute that involves judgment about whether the seller is at fault. These need empathy and judgment AI doesn't reliably produce.
Deciding what your brand sounds like, what your packaging should say, what your A+ content should emphasize. AI can execute on creative briefs. It can't write the brief.
When something goes wrong with your account (a listing suspension, a policy violation warning, a Section 3 review), the work involves judgment about what to say to Amazon. VAs with Amazon experience are valuable here. AI is a draft helper, not the decision-maker.
Contracts, terms, anything where being wrong has legal consequences. Humans.
Auditing 100+ ASINs for keyword coverage. Drafting new bullets. Checking style compliance. AI does this faster, more consistently, and (for technical compliance checks) more accurately than a VA.
VAs don't generate images. They edit existing photos. AI generates new lifestyle scenes, infographic variants, and creative iterations. Different category of work entirely.
Adjusting hundreds of bids based on multi-dimensional performance data. The right answer involves more math than a human can do in real time. PPC AI tools win here at scale.
Multi-variable forecasting (sales velocity, seasonality, lead times) is exactly the shape of work AI handles better than humans at the speed required.
Pulling data from SP-API, Amazon Ads, Helium 10, and a competitor's listing, then producing a unified report. Tedious for a human, fast for an agent.
The pattern I see most successful $250k-$2M sellers settle into:
Cost: ~$400-1,000/month all-in for the AI stack + VA labor. For a seller doing $30k+ per month in revenue, that's a small percentage of revenue for what amounts to a full operational team.
Here's the part that gets missed in "AI vs VA" debates. The same VA reviewing AI-generated output is dramatically more productive than the same VA doing the work from scratch. A listing optimizer running 50 ASINs in an hour, with a VA doing 5-min reviews, is a 6x productivity lift on the VA's hourly cost.
So the real question isn't "VA or AI?" It's "how do you compose VA and AI so each does what it's best at?" The hybrid model isn't a compromise. It's the answer.
The Amazon BSA Agent Policy effective March 4, 2026 affects how AI agents can act on your account. It doesn't restrict VAs (who are humans using your account). The implications:
Both, for different jobs. VAs win on judgment-heavy work like supplier negotiation, customer disputes, and brand voice decisions. AI agents win on repetitive multi-step work like listing optimization, image generation, and PPC analysis. The hybrid stack most successful $250k to $2M sellers settle into uses AI for execution and humans for judgment, with the VA reviewing AI output before it ships.
Amazon-experienced VAs in the Philippines or India typically run $4 to $10 per hour. Latin America-based VAs run $8 to $20 per hour. US-based VAs and agencies run $25 to $75 per hour. For comparison, pay-per-run AI tools on marketplaces like SellerShorts are $0.50 to $10 per run, and subscription PPC AI tools are $50 to $300 per month.
Not in any reasonable timeframe. The judgment-heavy work VAs do (negotiation, customer empathy, brand decisions, account-health triage with Amazon) isn't where AI is improving fastest. AI is taking over the repetitive execution work, which frees VAs to focus on the higher-judgment work where humans are irreplaceable. The same VA reviewing AI-generated output is typically 6x more productive than the same VA doing the work from scratch.
Under $50k per year: 0 VAs, 1-2 AI agents (you do the judgment work). $50k to $250k: 0-1 part-time VA at 10 hours per week, 2-4 AI agents. $250k to $1M: 1-2 VAs at 20-40 hours per week total, 4-6 AI agents. $1M to $5M: multiple VAs plus a custom AI stack or agency. $5M+: an internal team plus AI infrastructure and possibly custom builds.
No. The March 4, 2026 BSA Agent Policy affects automated AI agents acting on your account. VAs are humans using your account access, same as before. The hybrid model actually gets easier under the policy because human-in-the-loop review (VA reviewing AI output) is exactly the pattern Amazon prefers.
SellerShorts is the marketplace where AI Tool builders publish tools that handle the repetitive execution work so your VA can focus on judgment. Pay per run, capability-scoped, BSA-compliant. Don't see a tool for the work you'd like to automate? Post a request.