The ROI of Hiring a Virtual Assistant: Numbers You Need to See
It usually starts on a Tuesday.
You open your laptop at 6:40 a.m. to “just clear the inbox” before the day begins. Ninety minutes later you’re still there, having answered eleven emails, rescheduled two calls, chased an invoice, and booked a flight you’ll probably have to change. Then the actual workday starts. By four o’clock you’ve been in five meetings, three of which nobody prepared for, and the strategic work you blocked out on Monday has quietly slid to Thursday. Where it will slide again.
Microsoft has now put a number on this. Analysing anonymised Microsoft 365 telemetry alongside a survey of 31,000 knowledge workers across 31 markets, its 2025 Work Trend Index found that the average worker is interrupted every two minutes during core hours — roughly 275 times a day — by a meeting, an email, or a chat. The same research found 40% of people are already reading email by 6 a.m., nearly a third are back in their inbox by 10 p.m., and meetings starting after 8 p.m. are up 16% year over year. Microsoft called it the infinite workday.
Here’s the part that should bother you more than the burnout headlines: almost none of that fragmentation is the work. It’s the coordination of the work. Scheduling, confirming, chasing, formatting, reconciling, following up. It’s the connective tissue of a business, and in most small and mid-sized companies it’s being done by the single most expensive person on the payroll — the owner.
What follows is an attempt to price that properly, using what the research actually shows in 2024–2026, and to explain why the gap between businesses that have solved this and businesses still grinding through it has widened into something genuinely difficult to look at.
1. The Coordination Tax Nobody Puts on the P&L
Every business has a line item it never writes down. Call it the coordination tax: the hours senior people spend on work that is necessary, low-judgement, and endlessly interruptive.
The scale of it is easy to underestimate because it never arrives as a block. Microsoft’s telemetry shows the average employee receiving 117 emails and 153 Teams messages a day, with roughly 57–60% of meetings happening ad hoc, outside any calendar invite. Reclaim’s calendar research puts managers at 23-plus hours a week in meetings against about six for individual contributors — a figure that tracks with the MIT Sloan estimate cited by Axios, which found the average US executive now spends around 23 hours weekly in meetings, up from roughly 10 in the 1960s.
Now layer on the switching cost. Paul Graham’s 2009 essay on the maker’s schedule versus the manager’s schedule made the structural point that a single mid-afternoon meeting doesn’t cost an hour — it costs both halves of the day, because the morning gets truncated in anticipation and the afternoon gets spent recovering. Sophie Leroy’s work on attention residue supplied the mechanism: part of your mind stays with the previous task, so the hour after the interruption isn’t a full hour of anything.
Ask people how this feels and the language changes register immediately. On Blind, the anonymous professional forum, one engineer described sitting through 16 of 40 hours in meetings a week and wrote that they were <q>feeling absolutely burned out</q> — only to be told by a commenter that 16 hours <q>seems like a pleasure cruise</q> compared with their colleagues’ calendars. A product manager on the same platform described getting barely two hours of desk time out of an eight-hour day and asked, plainly, whether they were overreacting or whether this was simply the job now.
Microsoft’s data shows 275 interruptions per working day. That is not a busy week. That is a structural condition — and every one of those interruptions is a decision someone else could have made for you.
The honest version of the problem is this: most owners aren’t short of hours. They’re short of contiguous hours. And contiguous hours are where revenue actually gets made.
Why “I’ll just be more disciplined about it” fails
The standard response is a productivity system. Time-blocking, inbox zero, no-meeting Wednesdays, a stricter calendar. These help at the margin, and then they collapse — because the underlying volume doesn’t go anywhere. It’s the same tension the Blind thread surfaced: the poster’s company had blocked creative time on Tuesdays and camera-free Wednesdays, and scrum masters simply booked over both.
You can’t discipline your way out of a workload problem. You can only route it somewhere else, automate it, or drop it. Most owners try automation first, which is where the next set of numbers gets interesting.
2. Running the Actual Arithmetic
Let’s price this rather than gesture at it.
Take a UK business owner or director whose time is worth, conservatively, £75 an hour when applied to sales, delivery, or strategic decisions. Suppose 15 hours a week currently go to inbox triage, diary management, supplier chasing, CRM hygiene, invoice follow-up, travel booking, research, and social scheduling. That’s 15 hours × £75 = £1,125 a week of opportunity cost, or roughly £54,000 a year — before you count the compounding effect of never having a clear four-hour block.
Now the alternatives.
Hiring a UK-based PA or office manager. The salary is the smallest part of the story since April 2025, when employer National Insurance rose from 13.8% to 15% and the secondary threshold fell from £9,100 to £5,000 a year — meaning employers now pay NI on a far larger share of each salary. Sage’s worked example showed one employer’s secondary Class 1 bill rising roughly 30% year on year on identical headcount. Add auto-enrolment pension, holiday and sick cover, equipment, recruitment fees, and the desk itself, and a £32,000 salary lands somewhere north of £40,000 in true cost.
Hiring a managed South African VA. VAConnect publishes a full-time dedicated VA rate starting at $1,088 a month (roughly £860), against its stated comparison of £2,900-plus a month for a UK-based PA before employer NI, pension, and office overheads. The company’s headline claim on its own landing page is savings of $25,000-plus per year per placement, with no PAYE, no employer NI, and no auto-enrolment admin, because employment and compliance sit on the agency’s side.
The industry-level figure supports the direction of travel. BPESA, South Africa’s national GBS association, puts offshore delivery from South Africa at 55–65% cost savings versus source geographies including Manchester, Dallas, and Sydney, with government incentive programmes reducing operating costs by a further 7–10% for qualifying engagements.
So the crude ROI: if a £10,300-a-year placement (12 × £860) reclaims even 10 of those 15 hours, you’ve bought back £39,000 of owner time for a tenth of that in cost. The payback period is measured in weeks.
The arithmetic isn’t marginal. It’s a 3–4x return before you count a single extra deal closed with the reclaimed hours — which is why the gap between firms that have done this and firms that haven’t stops looking like a preference and starts looking like a structural disadvantage.
But — and this is where most ROI articles stop and most real implementations fail — the arithmetic only holds if the work actually gets done to standard. Cheap hours that produce rework are not cheap. Which brings us to the thing everyone assumed would solve this by now.
3. The Human in the Loop: Why Pure Automation Keeps Sending the Bill Downstream
The promise of the last three years was straightforward. Generative AI would absorb the admin, the drafting, the summarising, the scheduling — and the coordination tax would evaporate.
It hasn’t worked out that way, and there is now hard evidence about why.
In September 2025, researchers from BetterUp Labs and Stanford’s Social Media Lab published findings in Harvard Business Review introducing the term “workslop”: AI-generated output that looks polished but lacks the substance to move a task forward. Surveying 1,150 US desk workers, they found 40% had received workslop in the previous month, spending an average of 1 hour 56 minutes per instance cleaning it up. Extrapolating from respondents’ salaries, they estimated an invisible tax of about $186 per employee per month — over $9 million a year for a 10,000-person organisation.
The reputational finding was sharper still. Around half of respondents rated colleagues who sent workslop as less creative, less capable, and less reliable. More than half reported feeling annoyed by it; 22% said they felt offended.
By June 2026, HBR had escalated the diagnosis. Oxford’s Matthias Holweg and Babson’s Thomas Davenport described “knowledge decay” — the process by which unreviewed AI output enters reports, knowledge bases, and operating decisions, and the organisation’s collective understanding degrades from the inside. A February 2026 survey from Zety found 66% of workers spending up to six hours or more a week correcting AI-generated errors.
The customer-facing picture is, if anything, less flattering. Metrigy’s Customer Experience Optimization 2025-26 consumer study found 84.7% of participants would prefer a human over an AI agent — and that even when guaranteed their issue would be resolved, 80.1% still preferred the human. A HubSpot and SurveyMonkey study across seven markets found 82% of customers would choose human support even with identical wait times. Verizon’s CX Annual Insights Report, covering 5,000 consumers in seven countries, recorded an 28-point satisfaction gap: 88% satisfaction with human agents versus 60% with AI.
What this means practically
None of this is an argument against AI. It’s an argument about where the judgement sits.
A capable assistant using AI tools is a fundamentally different unit than an AI tool operating unattended. The assistant knows that your biggest client hates being called “mate,” that the Thursday board pack needs the revised margin figures not the draft ones, that the supplier who missed two deadlines gets a firmer email than the one who’s been faultless for four years. That’s context, and context is exactly what workslop lacks.
Run the same admin through an unsupervised model and you get an output that reads well and is subtly wrong — a meeting summary that misses the decision, a client email with the right tone and the wrong commitment, a CRM entry that’s tidy and inaccurate. You don’t notice for a fortnight. Then you notice all at once.
This is the case for the human in the loop, and it’s an empirical case, not a sentimental one. The rework tax on unsupervised AI output — two hours per instance, six hours a week, $186 a month per head — is precisely the cost you were trying to eliminate. Delegation to a trained person who uses AI removes it. Delegation to AI alone relocates it, usually onto you.
VAConnect’s model leans on this directly: assistants are recruited through its own pipeline, upskilled through VAVarsity, a proprietary training platform the company describes as Udemy-style, and supported through wellness programmes it calls Atomic Energy and Two-Way Happiness. The bet is that a trained, supported, accountable human who is fluent with modern tools produces less rework than either an untrained human or an untended machine. On the current evidence, that bet looks well placed.
4. The South African Advantage: Why the Cape Beat the Obvious Options
For UK and European businesses, the offshore decision used to be a choice between cost and convenience. South Africa has quietly collapsed that trade-off, and the numbers explaining why are more specific than most buyers realise.
The timezone is the whole ballgame
South Africa sits at GMT+2 — one to two hours ahead of the UK depending on the season. That means a South African assistant’s working day overlaps almost completely with a British one. No graveyard shift. No 6 a.m. handover. No waiting overnight for an answer to a question you asked at 4 p.m.
Compare that with the Philippines (UTC+8, typically eight to nine hours ahead of the UK) or India (UTC+5:30, five and a half hours ahead in winter). Those markets have real strengths, but real-time collaboration during UK office hours requires someone to work through the night — and night-shift work carries a well-documented attrition and quality penalty.
The timing matters more than it did in 2019, because Microsoft’s own data shows cross-timezone coordination is now a primary driver of the infinite workday: 30% of meetings span multiple time zones, a figure up eight percentage points since 2021, and late-evening meetings are climbing with it. Choosing a delivery location that removes timezone friction rather than adding it is no longer a nice-to-have.
VAConnect’s UK-facing materials describe the overlap slightly more conservatively — one to two hours of direct morning overlap with strong asynchronous coverage across the rest of the day — which is a fair description of how it actually works in practice for clients who don’t need someone glued to a shared screen for eight hours.
Language and cultural affinity
English is South Africa’s business language, full stop. The country ranks 13th globally and first in Africa on the EF English Proficiency Index 2025. For UK client-facing work, that’s not a technicality; it’s the difference between an assistant who can draft a nuanced apology to an unhappy customer and one who can only follow a script.
There’s a softer layer too, and it’s worth naming honestly because it’s the reason so many UK founders report the relationship “just working.” South African professional culture — Commonwealth legal heritage, shared spelling conventions, similar business etiquette, overlapping media and sporting reference points — reduces the friction that shows up in a hundred micro-interactions a week. Nobody needs to explain what “end of play Friday” means, or why a two-line email isn’t rude.
Compliance that doesn’t create work
South Africa’s Protection of Personal Information Act (POPIA) is closely aligned with GDPR. For a UK business handling customer data, that alignment removes an entire category of compliance engineering that other offshore destinations create.
The sector is scaling, fast
This isn’t a cottage industry. According to BPESA’s refreshed national value proposition, South Africa’s Global Business Services sector grew from 65,000 workers in 2019 to roughly 150,000 in 2024, with revenue expanding from $1.04 billion to an estimated $2.91 billion over the same period. In calendar 2025 the sector created 26,346 new international-facing jobs — its highest annual total since 2018 — with about 90% going to young people, according to BPESA’s sector report. The national target is 500,000 cumulative jobs by 2030.
Ryan Strategic Advisory’s buyer research has South Africa now outranking the Philippines as the preferred offshore destination among US buyers, and UK businesses account for a substantial share of new GBS employment in the country.
South Africa’s GBS revenue nearly tripled in five years while its cost advantage against UK hiring held at 55–65%. That combination — rising capability, sustained price gap, and a timezone that requires no compromise — is rare enough that its absence from most UK boardroom conversations is the genuinely surprising fact here.
The infrastructure objection, answered
Load shedding is the reasonable question, and it deserves a straight answer rather than a dismissal. South Africa’s scheduled power interruptions have been a real operational factor. Serious agencies handle it structurally: backup power, redundant connectivity, distributed teams, and contractual uptime commitments. VAConnect’s published service-level position is a 99.9% uptime guarantee backed by redundant connectivity and distributed team structures. Broadband coverage nationally reaches around 85%, with fibre in major cities running well above 100 Mbps.
Ask any prospective provider exactly what their continuity arrangement is. A provider who can’t answer that in one sentence is telling you something.
5. Freelancer Roulette vs. Managed Placement
Here’s where most VA experiments die, and it has nothing to do with geography.
The default route is a marketplace — post a brief, sift 80 proposals, hire the one with a decent portfolio and a rate you like, and hope. Sometimes it works beautifully. Often it produces a pattern every founder recognises: strong first month, gradual drift, unexplained slowdown around month four, then either a rate renegotiation or a disappearance. You’ve now spent your reclaimed hours on recruitment, onboarding, and a second recruitment.
The structural problem is incentive alignment. A marketplace freelancer serving eight clients optimises for breadth, not depth. They have no institutional memory of your business beyond what’s in their own notes, no training pipeline, no one managing their workload or wellbeing, and no consequence for churn beyond a rating.
The managed model inverts this. VAConnect — which has operated since 2008, originally as Lime Tree Consulting Solutions, rebranding in 2014 under founder Karen van Zyl — describes its approach as explicitly not a marketplace: dedicated placements rather than shared gig workers, recruitment and vetting handled centrally, continuous upskilling through VAVarsity, and performance accountability held by the agency rather than the client.
The company’s published figures are worth stating plainly, with the caveat that they are self-reported: 17-plus years operating, 98% client retention, 250,000-plus hours delivered, and a 4.8 Clutch rating. Its UK site cites a 4.9/5 client satisfaction score and 96% retention in a separate write-up — the small variance between properties is the kind of thing worth asking about directly on a discovery call. On the placement side, VAConnect states it has placed over 2,400 South African assistants with UK-based clients since 2019.
The commercially significant term is the replacement guarantee: if a placement isn’t performing to the agreed standard, VAConnect rematches at no additional cost and manages the transition. That single clause reallocates the largest hidden risk in the whole category — the cost of a bad hire — from you to the provider. Marketplaces do not offer this. They can’t; it isn’t their business model.
What the evidence says about why this matters
Attrition is the quiet killer of offshore economics. An operation running at 15% annual attrition rather than 35% avoids roughly 20 retraining cycles per 100 full-time equivalents every year. Each cycle costs context, quality, and management attention. When you’re placing one assistant rather than a hundred, the maths is even less forgiving: one departure is 100% turnover, and every process you documented walks out with them.
This is why the managed model’s investment in retention — training, wellbeing programmes, career structure — is not soft HR decoration. It’s the mechanism that protects your ROI.
6. What the Research Actually Shows About Remote Output
There’s a fair objection lurking under all of this: does remote work actually deliver equivalent output, or are we all just agreeing not to look too closely?
The best available evidence is a randomised controlled trial, which is unusual and valuable in this field. Nicholas Bloom of Stanford and collaborators ran a six-month RCT with 1,612 graduate employees at Trip.com, published in Nature in 2024. Half were allowed to work from home two days a week; half stayed in-office five days. The results: job satisfaction improved, quit rates fell by roughly one-third, and null-equivalence testing showed no effect on performance grades across the following two years of reviews. Among software engineers, lines of code showed no significant difference.
The detail worth sitting with is the managers. Before the trial, they expected hybrid to reduce output by about 2.6%. Afterwards, the same managers estimated it had improved performance by around 1%. Direct experience moved them; assumption hadn’t.
A 2025 systematic literature review in SN Business & Economics, covering 12 peer-reviewed studies from 2020 to 2024 across multiple countries, reached a compatible conclusion for small and medium enterprises: flexible arrangements generally improve productivity through higher satisfaction, eliminated commuting, and better work–life balance, with hybrid emerging as the strongest model. The review was candid about the failure modes — inadequate digital infrastructure, communication breakdown, and cybersecurity exposure.
Fully remote work has a more mixed literature. Some studies estimate individual output reductions in the 8–19% range for fully remote arrangements, against roughly flat effects for hybrid. That finding is worth taking seriously rather than waving away.
But note what actually drives it. The remote productivity penalty in that literature is largely a coordination and management penalty — unclear expectations, weak feedback loops, isolation, and no structure around the work. Those are precisely the variables a managed agency controls: defined scope, monitored delivery, structured check-ins, a named account manager, and someone other than you responsible for performance.
Put differently: the research doesn’t say remote support underperforms. It says unmanaged remote work underperforms. Which is an argument for the managed model rather than against remote delegation.
7. The First Ninety Days: Where the Gap Actually Opens
Two businesses hire an assistant on the same Monday. Six months later one has bought back a day and a half a week and the other has quietly stopped using theirs. The difference is almost never the assistant.
Weeks 1–2: Document before you delegate. The most common failure is handing over a task you’ve never articulated. Spend a week recording yourself doing the five things you want to hand off — screen recording works better than written SOPs, because you’ll narrate the exceptions you’d never think to write down. Every good agency will ask for this; the good clients volunteer it.
Weeks 3–6: Start with the reversible. Inbox triage, calendar management, travel, data entry, CRM hygiene, research, scheduling, invoice chasing. High-volume, low-judgement, easy to check. Resist the temptation to start with client-facing communication — not because a capable assistant can’t do it, but because you need the shared context in place first.
Weeks 7–12: Move up the judgement curve. This is where the return compounds. First-draft client emails in your voice. Supplier negotiations within defined limits. Meeting prep with a decision memo attached. Report assembly. Pipeline follow-up. The tasks where the assistant’s growing context on your business makes them faster than you’d be, not just cheaper.
Throughout: measure the right thing. Not hours logged. Not tasks completed. Measure owner hours reclaimed and what you did with them. If you’re reclaiming 12 hours a week and spending them on different admin, you haven’t solved anything — you’ve just changed the admin.
The businesses that widen the gap treat the assistant as an operating relationship that deepens. The ones that don’t treat it as a transaction that should have worked immediately. Nine months in, the first group has someone who knows their business almost as well as they do. The second group is back on a marketplace, posting a brief.
The Competitive Gap, Stated Plainly
Set the evidence side by side and the picture is uncomfortable.
The coordination load on knowledge workers has intensified to the point where Microsoft’s own telemetry records an interruption every two minutes. AI has not absorbed that load; on the best current evidence it has added a rework tax of roughly two hours per bad output, six hours a week for two-thirds of workers, and a measurable erosion of organisational knowledge quality. Customers, meanwhile, prefer humans by margins in the 80–90% range and will switch brands over a single frustrating automated interaction.
At the same time, a delivery market has matured that solves the problem directly. South Africa offers 55–65% cost savings against UK hiring, near-total working-hour overlap, top-15 global English proficiency, GDPR-aligned data law, and a sector that has nearly tripled in revenue in five years. A managed agency layer on top handles vetting, training, continuity, and replacement risk.
So the two businesses in question are not competing on equal terms. One has a director spending fifteen hours a week on work that costs them £75 an hour and could be done to standard for a fraction of that. The other has bought those hours back, redeployed them into sales and delivery, and — because the work is done by a trained human rather than an unsupervised model — isn’t paying the rework tax on the other end.
That gap doesn’t stay fixed. It compounds every quarter, because the reclaimed hours go into the parts of the business that grow.
The genuinely strange thing, reviewing the 2024–2026 evidence, is not that the gap exists. It’s how well documented it now is, and how many businesses are still on the wrong side of it.
Comparative Table: Three Ways to Handle the Coordination Load
| Dimension | DIY Coordination (you do it) | Generic Freelancer / Marketplace | VAConnect (Managed Placement) |
|---|---|---|---|
| True cost | ~£54,000/yr in owner opportunity cost (15 hrs/wk @ £75) | £8k–£20k/yr, plus 20–40 hrs of your time on sourcing, vetting, re-hiring | From ~$1,088/mo (~£860); company-stated savings of $25,000+/yr vs. local hire |
| Vetting & screening | None — you are the filter | You screen 50–100 proposals yourself; portfolio-based, unverified | Centralised recruitment and skills verification before you see a shortlist |
| Training | You train yourself, badly, on the fly | Whatever they brought; no ongoing development | VAVarsity continuous upskilling; pre-trained on Xero, HubSpot, Slack, Asana, M365, Google Workspace |
| Timezone overlap (UK) | N/A | Variable — often 5–9 hrs offset | GMT+2: 1–2 hrs direct morning overlap, full-day async coverage |
| Continuity risk | Total — it stops when you stop | High; churn is the norm, knowledge leaves with them | Free replacement and managed transition if performance slips |
| Accountability | You | The client (you) manages performance | Agency-held; account management and delivery monitoring |
| Compliance / employment admin | Yours | Yours to verify (contracts, IP, data) | Handled agency-side; no PAYE, employer NI, or pension admin; POPIA aligns with GDPR |
| Quality control | Whatever you have energy for | Ad hoc; rework absorbed by you | Structured QA, performance frameworks, documented standards |
| Scale-up speed | You can’t | New search each time | Placements typically filled within ~2 weeks (company-stated) |
| Realistic owner hours reclaimed | 0 | 4–8/week, net of management time | 10–15/week once ramped |
| Reported retention / satisfaction | — | Platform ratings, no verified continuity data | 98% client retention, 250,000+ hours delivered, 4.8 Clutch (company-reported) |
Sources referenced
- Microsoft, Breaking Down the Infinite Workday, Work Trend Index Special Report (2025) — microsoft.com/worklab
- Bloom, N. et al., “Hybrid working from home improves retention without damaging performance,” Nature (2024) — nature.com
- Niederhoffer, K., Rosen Kellerman, G., Lee, A., Liebscher, A., Rapuano, K. & Hancock, J.T., “AI-Generated ‘Workslop’ Is Destroying Productivity,” Harvard Business Review (September 2025) — hbr.org
- “The impact of remote and hybrid work models on SME productivity: a systematic literature review,” SN Business & Economics (2025) — link.springer.com
- BPESA, Refreshed National Value Proposition for South Africa’s GBS Sector (March 2025) — bpesa.org.za
- Metrigy, Customer Experience Optimization 2025-26 — Consumer Views — metrigy.com
- Practitioner sentiment: anonymous professional threads on Blind (burnout by meetings; zoom fatigued)
- VAConnect company-reported data — vaconnect.co.za and vaconnect.co.uk
- Sage / Xero UK guidance on April 2025 employer National Insurance changes
Figures attributed to VAConnect are as published by the company on its own properties and have not been independently audited.