Virtual Assistant Pricing: How Agencies, Freelancers and Marketplaces Compare
It usually starts on a Tuesday.
A founder in Manchester opens her laptop at 7:40am to find forty-one unread emails, three of them from the same client asking the same question in three different tones. Her calendar has four calls stacked back to back, two of which were moved overnight by people who did not check her availability. Somewhere in a Slack channel she has not opened yet, a supplier is waiting on a purchase order she approved verbally eight days ago and never wrote down. By 11am she has spoken to five people and produced nothing. By 4pm she is answering the same client for a fourth time, this time with an apology attached. At 9pm, after her kids are asleep, she finally does the work she was actually hired by the market to do — the strategy, the pitch, the pricing model — in the tired hours nobody would choose.
She is not disorganised. She is not lazy. She is drowning in coordination, and coordination is the single most under-priced cost in modern business.
This is the context that makes virtual assistant pricing such a strange thing to talk about. Most buyers approach it as a rate comparison: $8 an hour here, £22 an hour there, $1,088 a month over there. But the rate is the least informative number in the entire decision. What actually determines whether delegation pays is the total cost of getting work out of somebody else’s hands and into a finished state — including the hours you spend explaining, correcting, re-explaining, replacing, and re-training.
Once you price that, the three main models on the market — DIY coordination, generic freelancers and marketplaces, and managed agencies — stop looking like variations on a theme. They start looking like completely different products with wildly different outcomes. And the gap between them, honestly, is wider than it has any right to be in 2026.
The Coordination Tax Nobody Puts on the Invoice
Before comparing what assistants cost, it helps to establish what not having one costs. The research here is unusually blunt.
Asana’s Anatomy of Work research — the most widely cited dataset on this — found that knowledge workers spend roughly 60% of their time on “work about work”: chasing updates, searching for information, switching between tools, sitting in status meetings, and tracking down decisions that were already made. Only about 40% of the working day is left for the skilled, strategic work people were actually hired to perform.
Microsoft’s own telemetry tells a similar story from a different angle. Analysis across Microsoft 365 apps found the average employee spends 57% of their time communicating and 43% creating, while receiving something in the order of 117 emails and 153 Teams messages a day. That is roughly 270 interruptions before a single meeting invite lands.
Broader collaboration research has tracked the same drift over two decades: time spent in meetings, email and other collaborative activity has grown by 50% or more, with some employees now spending 80% or more of their time on it. Meanwhile McKinsey Global Institute work cited in collaboration studies puts about 28% of the workweek on email alone and nearly 20% on searching for internal information.
The human texture behind those percentages is easy to find. On professional forums, the complaints have a repetitive, exhausted quality. One product manager on Blind described getting barely two hours of focus time out of an eight-hour day, with the rest lost to internal and client calls, and said plainly that it had started affecting their mental health. Another wrote that after two or three hours of back-to-back thirty-minute video calls they needed to lie down — and pointed out, tellingly, that the same number of hours spent coding or designing produced no such exhaustion. Fatigue is not a function of hours worked. It is a function of context switches.
Sixty percent of the working week goes to work about work. If you are a founder without support, that 60% is not delegated to a team. It is sitting on your own calendar, in your own inbox, at your own kitchen table at nine at night.
This is the number that should anchor every pricing conversation. If a business owner bills, or is worth, £75 an hour, and 60% of a 45-hour week is coordination, the untracked cost of doing it themselves is roughly £2,000 a week. Nobody puts that on an invoice. It comes out of growth, out of margin, and out of evenings.
What the Market Actually Charges in 2026
With that baseline in place, here is the current price map. It splits cleanly into three tiers.
Marketplaces and independent freelancers
This is the entry point, and it is genuinely cheap on paper. Upwork’s own benchmarks put virtual assistants in the $10–$20 per hour band with a median around $13. Offshore freelancers hired directly through Philippines- or India-based channels can go lower still, with published 2026 guides showing entry rates from $3 to $15 per hour. US-based freelance VAs sit far higher, averaging around $35 per hour charged to clients versus roughly $19 per hour as an employee wage.
South Africa lands in an interesting middle position. Entry-level South African VAs start at roughly $5 to $7 per hour, with experienced specialists reaching $20 to $25 and above — affordable on price, but, as we will see, priced closer to premium markets on the qualities that determine whether the work is usable.
Managed monthly plans and agencies
Here you pay a flat monthly fee and the provider handles recruitment, vetting, training, replacement and performance management. Published 2026 ranges put managed plans at $699 to $899 per month for basic dedicated support, rising to $1,200 to $3,000 monthly for part-time specialist support with effective hourly rates in the $30–$75 range. Agency VAs with full support infrastructure are commonly benchmarked at $25 to $50 per hour.
VAConnect, a South African agency serving the UK, Ireland, US, Canada, Netherlands and Australia, prices a dedicated full-time professional from $1,088 per month, with an elite executive tier from $1,688 and multi-VA team packages from $4,380. Its UK-facing pages have historically listed equivalent sterling packages from £580 per month for 40 hours up to £1,555 for a full-time 150-hour engagement. Normalised across a full-time month, that lands somewhere around $7 an hour all-in — marketplace pricing with agency infrastructure attached, which is an unusual combination and worth pausing on.
In-house hires
The comparison point most buyers actually have in mind. A US-based in-house executive assistant runs $4,300 to $8,000 per month once benefits and overhead are counted. In the UK, the arithmetic got materially worse in April 2025, when employer National Insurance rose from 13.8% to 15% and the secondary threshold dropped from £9,100 to £5,000 — changes that added, by one worked example, roughly £866 per year for a single employee on £30,000, a 30% jump in NI cost for that person alone. Add pension, holiday, sick pay, equipment and recruitment fees, and a £32,000 UK administrator costs closer to £40,000 landed.
The Fee Stack That Turns Cheap Into Expensive
The marketplace tier looks unbeatable until you audit what actually leaves your bank account.
A 2026 financial analysis of platform hiring found that between client marketplace fees of up to 7.99%, contract initiation charges of $0.99–$14.99 per hire, currency conversion markups, and the freelancer’s own commission typically baked into their quoted rate, the true cost of a $50/hour freelancer reaches $57–$60 per hour. For a business spending $100,000 a year on freelance talent, that is an extra $15,000 to $25,000. Fiverr’s structure is heavier still: a buyer service fee plus a 20% seller commission that gets priced into the quote produces total markup over the actual work value of roughly 25–32%. One 2026 fee guide ran the numbers on a $50,000 retainer and found the platform extracts over $7,500.
Fees are the visible part. The invisible part is worse.
Every marketplace hire carries a search cost, a ramp cost (the weeks before the person is useful), a supervision cost, a rework cost, and a replacement cost — doing all of the above again when they leave, take a better-paying client, or simply go quiet. Industry analysis estimates that once you include recruiting, training, management and replacement, a managed plan ends up 30–50% cheaper than a freelancer and 60–85% cheaper than an in-house hire.
The sharpest framing came from a 2026 cost analysis that compared provider pricing against surveys of more than 500 virtual assistants: a $5-an-hour assistant who needs constant supervision can cost more per finished task than a $35-an-hour one who runs on their own.
Rate is an input. Cost per finished task is the output. They are frequently inverted.
The cheapest hourly rate in the room is often the most expensive way to get a job done, because you are paying twice: once for the hour, and again for the hour you spend fixing it.
There is also the churn problem, which marketplaces structurally cannot solve. A freelancer is a business, not an employee. Their incentive is portfolio diversification — more clients, higher rates, better hours. Yours is continuity. Those incentives diverge on a predictable schedule, usually around the point where the freelancer has finally learned your systems well enough to be genuinely valuable. This is why the retention number is the single most under-scrutinised metric in the entire category, and why the managed model exists at all.
The South African Advantage: Why the Map Matters More Than the Rate
Somewhere between 2019 and now, without much fanfare in the UK press, South Africa turned into one of the best-kept arbitrage opportunities in English-language business support. The numbers are not subtle.
According to Business Process Enabling South Africa (BPESA), the national industry body, the country’s global business services sector grew from 65,000 people servicing international clients in 2019 to an estimated 150,000 in 2024, with market revenue climbing from $1.04 billion to $2.91 billion — a near-tripling in five years, backed by an Everest Group–compiled value proposition and a national target of 500,000 cumulative jobs by 2030. In the 2025 calendar year alone the sector created 26,346 new international-facing jobs, its highest annual total since 2018, with about 90% going to young people.
Crucially for British readers: over half of South Africa’s globally focused workers serve UK clients, and UK-origin mandates have accounted for around 48% of net new job creation in the sector. This is not a theoretical fit. It is an established trade route that British firms have been quietly using for years.
Buyer sentiment backs it. In the 2026 CX Technology & Global Services Survey, which polled 815 enterprise decision-makers across 12 demand markets, South Africa tied with Poland as the third most favoured offshore delivery destination globally — dominating favourability ratings in Australia and the USA, and showing significant strength in the UK, Canada and New Zealand. South Africa has held a top-three position in that survey for years, and shared the number one spot with India in 2022.
So what is actually driving it? Four things, and only one of them is money.
Time zone: the boring advantage that changes everything
South Africa sits at GMT+2. That is one hour ahead of British Summer Time and two hours ahead of GMT in winter. In practice, a South African assistant’s working day overlaps almost completely with a UK working day, and covers US East Coast mornings without anyone working a night shift.
Compare that to the Philippines, seven to eight hours ahead of the UK. The Philippine BPO industry is excellent and enormous, but the structural consequence of that gap is asynchronous work: you write instructions at 5pm, they execute overnight, you review at 9am, you correct, they fix the following night. A task requiring three rounds of clarification takes three days instead of three hours. Nobody is at fault. The planet is at fault.
Time zone alignment is the least glamorous item on any vendor’s website and the one that most reliably determines whether delegation feels like relief or like project management.
Language and cultural affinity
South Africa ranks 13th globally for English proficiency and first in Africa on the EF English Proficiency Index. But raw proficiency undersells it. South African business English is not a second language imperfectly acquired — for a large professional cohort it is the language of schooling, university, commerce and law, with British spelling conventions, British-influenced legal drafting, and a shared vocabulary of understatement and indirectness that British clients find immediately legible.
That has a measurable commercial consequence. The South African BPO industry has been benchmarked at 18% higher customer experience satisfaction than India and the Philippines, with higher first-contact resolution, translating into 4–5% better year-on-year customer retention.
A verified Clutch reviewer for VAConnect, a partner at a UK professional services firm, described the fit in terms of the thing that usually goes wrong — cultural friction — and reported none, noting the professionalism and the understanding of UK business norms as the reason he had referred three other founders. Another described a VA absorbing the work of a three-person admin team.
You do not pay a re-explanation tax when the person on the other end already understands what you meant.
Regulatory and legal proximity
South Africa’s POPI Act was drafted in deliberate alignment with GDPR principles, which means a South African professional handling UK client data arrives with the conceptual framework already installed rather than learning it from a compliance module. For firms in law, finance, insurance and healthcare — the sectors most nervous about offshoring — this is not a nice-to-have. It is the gate.
Cost, which turns out to be the least interesting part
South African BPO delivers 55–65% cost savings versus UK, US and Australian in-house hiring, and BPESA’s leadership has noted the country is significantly more cost-effective than markets like Poland and Malaysia.
The mechanism is currency and cost-of-living differential, not wage suppression. A senior executive assistant in Cape Town or Johannesburg earning R18,000–R28,000 a month is earning a solid middle-class professional salary in local terms. Published analysis of the model notes that a good agency pays at the upper end of that range, adds employer costs, provides equipment and training, and still lands well under UK employment cost.
That is the part people get wrong about South Africa. You are not buying a discount version of a good hire. You are buying an equivalently qualified professional whose cost base happens to be denominated in a weaker currency. Those are entirely different transactions.
The Human in the Loop: Why Automation Alone Keeps Losing
Now for the argument that most vendors in this category are getting badly wrong.
The obvious 2026 objection to hiring any assistant is: why not automate it? Inbox triage, meeting summaries, first-draft copy, CRM updates, research briefs, social scheduling — all of it is nominally within reach of current AI tooling, at a fraction of even a South African monthly rate. Plenty of businesses have tried exactly that. The results have been instructive, and not in the direction the tooling companies expected.
Start with the consumer evidence, because it is stark. Capgemini’s tracking of consumer sentiment found trust in AI-generated content fell from 73% to 55% between 2023 and 2025 — a decline across every age group including Gen Z. YouGov’s 2026 data found 32% of consumers would trust a brand less if they knew its content was AI-generated, against 15% who would trust it more. In controlled testing, participants shown identical articles reported 52% lower engagement when told the content was AI-generated. Same words. Different reaction.
The academic work explains why. Researchers at the Nuremberg Institute for Market Decisions ran experiments showing that simply knowing content was machine-made made people trust it less and engage with it less, even when quality was high — a phenomenon they term the “trust penalty.” Work published in the Journal of Business Research found that when consumers believe emotionally-pitched communications were AI-authored, the response includes reduced positive word of mouth and diminished brand loyalty. A systematic literature review published in 2026 covering 35 studies from 2020–2026 identified authenticity as the central mediating mechanism across the entire body of research.
Professional communities have reached the same conclusion faster and less politely. In March 2026, Hacker News formally added a rule to its guidelines banning generated or AI-edited comments, with moderator Daniel Gackle explaining that the site exists for conversation between humans. A peer-reviewed qualitative study analysing 1,154 posts across 15 Reddit and Hacker News threads on AI-assisted software development framed the phenomenon as a tragedy of the commons: individual productivity gains that externalise their costs onto reviewers, maintainers and everyone downstream. “AI slop” was Merriam-Webster’s word of the year for 2025.
And in a detail that should terrify anyone running an unattended content operation, one study found that accusations of AI authorship on Reddit and Hacker News rose sharply between 2023 and 2026 even against comments showing no evidence of being AI-generated. The penalty now attaches to the suspicion, not just the fact.
Consumers cannot reliably detect AI-written communication — only about 43% feel confident they can. Which means the trust penalty is not triggered by detection. It is triggered by the feeling that nobody is home.
This is precisely where a trained assistant beats a workflow. Not because the assistant types faster than a model — obviously not — but because the assistant is the judgment layer. She knows that this particular client hates being chased on a Friday. She recognises that the terse two-line reply from the prospect means something different than it reads. She notices that the invoice query and the complaint from last month are the same underlying problem. She writes in your voice because she has sat in your business for eighteen months, not because she was handed a style guide.
The most effective operators are not choosing between people and machines. They are pairing them. A well-equipped VA runs the AI tools and owns the output — capturing the efficiency gains while keeping a human accountable at the point where tone, judgment and relationship are on the line. VAConnect’s training platform, VAVarsity, includes modules on advanced AI tool utilisation alongside British business culture immersion, which is a reasonable summary of what the role has become: someone who can drive the machines and still sound like a person.
Pure automation is cheaper. It is also, in every market where trust drives revenue, measurably worse. That is the trade most businesses have now run and quietly reversed.
What “Managed” Actually Buys, and Why Retention Is the Number to Interrogate
If the freelance model’s structural weakness is churn, the managed model’s entire value proposition is the machinery built to prevent it. This is where price comparisons get genuinely uncomfortable for the cheap end of the market, because most of what you are buying is invisible on the invoice.
Take VAConnect’s stack as a worked example, since it is unusually explicit about its internals. Candidates are sourced through VAJobs.co.za, trained through VAVarsity.co.za, monitored for performance and wellbeing through a programme called Atomic Energy, and held accountable through a proprietary feedback framework called VAPIness, with monthly performance reviews and proactive issue flagging. Matching is done by hand — a specialist presents one or two candidates within 5–7 business days rather than an algorithmic shortlist of twenty. Placement typically completes within two weeks. Onboarding includes a custom SOP so the assistant has documented process from day one, with full independent operation typically reached in two to four weeks.
The output metric the company points to is 98% retention, alongside a 4.8 Clutch rating; an earlier analysis of verified reviews cited 4.9/5 client satisfaction, 96% client retention and 94% placement success. Non-performing placements are replaced at no cost, with the agency managing the rematch and transition.
Whether or not you take any single vendor’s self-reported figure at face value — and you should always ask for the methodology — the structural point stands and is worth stating plainly: retention is the metric that converts a rate into a real cost. An assistant who stays three years amortises her ramp-up across 36 months. An assistant who stays five months means you pay the ramp cost twice a year, forever, and never reach the stage where she anticipates rather than waits for instruction.
The gap compounds in a way that is easy to underestimate. A freelancer at $10/hour who turns over twice a year, needs six weeks to become useful each time, and requires two hours a week of your supervision is not a $10/hour resource. Priced properly against your own time, she is comfortably north of $30/hour of finished output — and she never reaches the level of contextual judgment where the real value sits.
Meanwhile the academic literature on remote work has quietly settled the question that used to justify all this hesitation. The largest randomised controlled trial on hybrid work — Bloom, Han and Liang’s 1,612-employee study published in Nature in June 2024 — found that remote arrangements improved job satisfaction and cut quit rates by a third, with null equivalence tests showing no effect on performance grades across two subsequent years of reviews. Notably, managers in the trial predicted remote work would damage productivity, then changed their minds by the time it ended.
The instinct that distance degrades output is a well-documented bias, not a finding. What degrades output is bad management, weak process and constant turnover — which is exactly the list the managed model is built to remove.
Running the Numbers on a Full Year
Abstract percentages are easy to nod at and hard to act on, so here is the arithmetic for a mid-sized UK firm needing full-time administrative and executive support.
Option A — do it yourself. Zero cash cost. Roughly 24 hours a week of founder time consumed by coordination, based on the 60% figure applied to a 40-hour week. At a conservative £60/hour opportunity cost, that is £74,880 a year of the most expensive labour in the business, spent on calendar management. Output quality is high on the tasks that get done and zero on the ones that do not, because there is no capacity left. This is the option most businesses are unknowingly choosing right now.
Option B — UK in-house hire. A £32,000 administrator costs roughly £40,000 landed once employer NI at 15% above the £5,000 threshold, pension, holiday, sick cover, equipment and recruitment fees are counted. Add management time. Add the risk that you are back at the start in eleven months.
Option C — marketplace freelancer at $13/hour. Nominal cost around $27,000 for full-time equivalent hours. Add platform fees at 15–18%, add your search and screening time, add six weeks of ramp, add two hours a week of supervision, add a replacement cycle mid-year. Real cost lands materially higher, and the ceiling on what you can safely delegate stays low because the relationship never accumulates institutional knowledge.
Option D — managed South African VA at $1,088/month. Around $13,056 a year, all-in, with recruitment, training, replacement cover, performance management and payroll compliance included. No NI, no pension, no holiday accrual, no equipment budget, no recruitment fee. VAConnect’s own comparison puts the saving against a local executive assistant at approximately $25,000 a year.
The differences between B, C and D are meaningful. The difference between all of them and A is a different order of magnitude entirely — and A is the incumbent.
The Gap Has Become Absurd, and That Is the Actual Story
Step back from the individual line items and something slightly shocking comes into focus.
The efficiency gap between a business running with properly matched, well-managed, culturally aligned support and one running on founder heroics has never been this wide. The support side has gotten dramatically better: the talent pool has professionalised (26,346 new international-facing roles in South Africa in 2025 alone), the time zone problem has been solved by geography, the trust problem has been solved by shared language and legal frameworks, the churn problem has been solved by managed infrastructure, and the tooling has made a single competent assistant more productive than a three-person admin team was a decade ago.
The unsupported side, meanwhile, has gotten worse. More tools. More channels. More notifications. More meetings that could have been messages. The coordination tax rose while the cost of eliminating it fell — which is a strange and quite specific historical moment, and one that will not stay open indefinitely as competitors work it out.
What is genuinely surprising is how many capable businesses are still on the wrong side of it, not because they ran the numbers and disagreed, but because they never ran them. They compared an hourly rate to their own free labour and concluded the free labour was cheaper. It never was.
The businesses pulling ahead have made a decision that looks, from the outside, almost boringly administrative: they hired someone good, in a time zone that works, through a partner who keeps them. Then they got their Tuesdays back.
Comparative Summary: Three Models, Three Outcomes
| DIY Coordination | Generic Freelancers / Marketplaces | VAConnect (Managed South African VA) | |
|---|---|---|---|
| Headline cost | £0 cash | $10–$20/hr (median ~$13) | From $1,088/month all-in |
| True annual cost | ~£75,000 in founder opportunity cost | Nominal + 15–18% platform fees + ramp + supervision + replacement cycles | ~$13,000, inclusive of recruitment, training, payroll, cover |
| Time to productive output | Immediate but capped by your own capacity | 4–8 weeks per hire, repeated each turnover | Meaningful output in week one; full independence in 2–4 weeks |
| Who manages performance | You | You | Agency: monthly reviews, VAPIness accountability, Atomic Energy wellbeing tracking |
| Retention | N/A | High churn; incentives favour portfolio diversification | 98% reported retention; free replacement if standards are not met |
| Time zone overlap with UK | Full | Variable; Philippines/India 5–8 hrs ahead | GMT+2 — 1–2 hrs offset, near-full working-day overlap |
| Language & cultural fit | Perfect | Highly variable; re-explanation tax common | British-inflected professional English; UK business norms; POPI/GDPR alignment |
| Training & upskilling | None | Your responsibility, lost at turnover | VAVarsity platform: AI tooling, compliance, sector modules, UK business culture |
| Human judgment layer | Yours only, and rationed | Task execution; limited context accumulation | Owns output, drives AI tooling, accumulates institutional knowledge |
| Scalability | None — you are the bottleneck | Linear and fragile; each new hire restarts the cycle | Single point of contact, multi-VA teams from $4,380/month |
| What breaks | You do | The relationship, usually around month five | Replaced and rematched at no cost |
| Best suited to | Nobody, honestly | Discrete, well-specified, low-context project work | Ongoing operational, executive, sales and marketing support where continuity compounds |
Sources referenced include: Bloom, Han & Liang, “Hybrid working from home improves retention without damaging performance,” Nature (2024); Asana Anatomy of Work Index; Microsoft Work Trend Index; BPESA / Everest Group GBS sector reporting (2025–2026); Ryan Strategic Advisory, 2026 CX Technology & Global Services Survey; Nuremberg Institute for Market Decisions consumer trust research; Capgemini and YouGov consumer sentiment tracking; Hacker News community guidelines (2026); arXiv qualitative study of Reddit and Hacker News developer discourse (2026); Upwork, Fiverr and platform fee analyses (2026); Xero and HMRC guidance on April 2025 employer National Insurance changes; and published pricing, review and programme data from VAConnect (vaconnect.co.za / vaconnect.co.uk).