AI Calls vs Human Agents: The Real Cost Comparison

A brokerage that puts ten agents on a dormant list pays for every hour they spend listening to ringing phones. The AI calls vs human agents cost question is really a question about what an attempt costs, not what an hour costs. Topcalls runs calls at $0.35 per minute all-inclusive, and the US Bureau of Labor Statistics puts the median customer service representative at $21.53 an hour, so the comparison starts with two real numbers and works forward from there.

This guide builds the AI calls vs human agents cost model a broker can defend in a budget meeting: the loaded hourly cost of a human desk, the per-attempt cost of an AI call, a worked 10,000-trader scenario, the costs that hide outside the wage line, and the calls where a human still earns the money.

Key Takeaways

• The median US customer service representative earned $21.53 an hour in May 2025 (BLS), which becomes about $30.80 an hour once benefits are added.
• Topcalls AI calls cost $0.35 per minute all-inclusive, so a 1.5-minute attempt costs about $0.53 and a three-minute conversation costs $1.05.
• At 12 dials an hour, a human attempt costs about $2.57 loaded, roughly five times the AI attempt before any manager or QA cost is counted.
• On a 10,000-trader dormant list called twice, the AI campaign costs about $10,500 and the human desk about $51,300 at 12 dials an hour.
• US federal rules allow residential telemarketing calls only between 8:00 a.m. and 9:00 p.m. local time, so a single-shift desk covers a fraction of a global broker's calling windows.

1. What does the AI calls vs human agents cost comparison include?

A fair AI calls vs human agents cost comparison prices one call attempt from both sides: wages plus benefits divided by attempts per hour for the human desk, and $0.35 per minute times call length for Topcalls. Anything else a broker adds (telephony, recording, transcription, QA, overtime, recruiting) is a second layer that only widens the gap, because the AI rate already includes the first four.

The comparison in this guide uses three public numbers and one assumption. The public numbers are the BLS median wage for customer service representatives, the BLS employer compensation split between wages and benefits, and the Topcalls per-minute rate. The assumption is dials per hour, and the sensitivity table in section 4 shows what happens when you change it.

Brokers who want the full profit side of the equation as well as the cost side, should read the AI calling ROI guide for brokers next. This article stays on cost so the numbers are easy to check.

2. What does a human calling desk cost a brokerage per hour?

A human agent on a brokerage reactivation desk costs about $30.80 an hour once benefits are loaded onto the median wage. The BLS Occupational Outlook Handbook reports a median wage of $21.53 an hour for customer service representatives in May 2025, and the BLS Employer Costs for Employee Compensation release for March 2026 shows wages and salaries make up 69.9% of private-industry compensation, with benefits the other 30.1%.

Divide $21.53 by 0.699 and the loaded hourly cost lands at $30.80. That's a US median, and a retention desk in Limassol, Dubai or Kuala Lumpur will sit below it while a London desk sits above it. Swap in your own payroll figure; the shape of the answer doesn't change.

The hourly figure is only half the input. What matters for a dormant list is how many attempts an agent completes in that hour. Manual dialing from MT4 or MT5 account exports, with CRM lookups and disposition notes between calls, rarely gets past 12 to 15 attempts an hour on a list where most numbers go to voicemail. A predictive dialer lifts that, at the price of the abandoned-call limits covered in section 5.

So at 12 dials an hour, one human attempt costs $30.80 divided by 12, which is about $2.57. At 20 dials an hour it's $1.54. Those two figures bracket most in-house desks.

3. What does an AI call cost a broker per minute?

A Topcalls AI call costs $0.35 per minute, and that rate covers the voice model, telephony, call recording, transcription and analytics. There are no per-seat fees, no setup fees, no minute bundles and no separate telephony or transcription add-ons. A 1.5-minute average attempt, which blends short no-answer calls with longer conversations, costs about $0.53. A full three-minute reactivation conversation costs $1.05. The AI voice agents product page lists what's included.

The per-attempt cost also stays flat at 2 a.m. The human desk cost doesn't. An AI campaign calling a broker's Australian clients at 6 p.m. Sydney time from a European office costs the same per minute as one calling London at lunch, and there's no overtime line, no night shift and no manager waiting up.

Language is the other place the per-attempt cost diverges. Topcalls speaks 32 languages, so a broker with Arabic, Thai and Portuguese segments on one list runs one campaign at one rate. A human desk either hires per language or leaves those segments uncalled. The multilingual voice agent guide for brokerages goes through how that works at campaign level.

4. How do the numbers compare on a 10,000-trader dormant list?

Calling 10,000 dormant traders twice each costs about $10,500 with Topcalls and about $51,300 with a human desk at 12 dials an hour. That's 20,000 attempts at $0.53 each versus 1,667 agent hours at $30.80. The human figure needs ten agents for a month; the AI figure needs one campaign brief and a suppression list.

The 1,667 hours is the number a head of retention should read twice. Ten full-time agents at 167 hours a month each can clear 20,000 attempts in a month. Two agents take five months, and the traders at the bottom of the list have gone another five months without contact. Topcalls processes 63,000+ AI calls a day across the platform, so a 20,000-attempt campaign is measured in days, not quarters.

Dials per hour is the assumption that moves the human figure most, so here's the sensitivity. The AI column doesn't change because the AI cost is priced per minute, not per agent hour.

Even at 30 dials an hour, a pace that only a dialer running back-to-back short calls reaches, the human desk costs roughly double. And the human cost in this table is wages and benefits only. Run your own list through the dormant trader revenue calculator to put deposit values on the other side of the ledger, or read how to measure cost per reactivated trader for the KPI that comes out the other end.

5. Which hidden costs change the comparison?

Five costs sit outside the wage line and all of them land on the human side: recruiting and training, management and QA, telephony and recording, calling-window coverage, and dialer compliance. The Topcalls $0.35 per minute rate already includes telephony, recording, transcription and analytics, so the AI column carries none of those as separate items.

Training time. The BLS lists 2 to 4 weeks of on-the-job training as typical for customer service representatives. For a brokerage that means two to four weeks of loaded wages per hire before the agent can hold a conversation about a margin call or a KYC document, and the BLS projects the occupation to shrink 5% between 2025 and 2035, so the hiring pool isn't getting deeper.

Management and QA. Ten agents need a team lead and someone listening to calls. Topcalls records and transcribes every call at no extra charge, so the compliance desk reviews transcripts instead of paying a QA analyst to sample recordings.

Calling windows. Under the US Telemarketing Sales Rule at 16 CFR 310.4(c), outbound calls to a residence are only allowed between 8:00 a.m. and 9:00 p.m. local time at the called person's location. A broker with clients across six time zones needs shifts to cover those windows with humans. An AI campaign schedules each trader inside their own window automatically.

Dialer compliance. The same rule sets a safe harbor of no more than 3% abandoned calls, measured over each 30-day period. A predictive dialer that pushes agents to 30 dials an hour does it by dialing ahead and dropping the calls nobody is free to take, which is exactly what the 3% limit polices. An AI agent answers every call it places, so the abandonment metric is close to zero by design.

Spreadsheets and dispositions. A human agent types the outcome after every call. Topcalls writes the disposition, transcript and recording to the CRM through Integrations while the next call is already dialing, which is where the brokerage call outcome automation post picks up.

6. Where do human agents still win on cost?

Human agents win on cost when the value of one conversation is high enough that a lower per-attempt price stops mattering. A trader with a six-figure balance who wants to talk through a hedging setup, a client disputing a margin call, or a professional client asking about a bespoke spread all justify a senior account manager at any hourly rate.

The right model for most brokerages is not AI or humans but AI first, humans second. The AI handles the 20,000 attempts, qualifies who actually wants to talk, and hands the live ones to an account manager. That way the $30.80 hour is spent on conversations that were already warm. The human handoff guide for AI calls covers the trigger rules, and the AI voice agent vs forex call center comparison looks at the same trade-off from the staffing side.

Topcalls customers typically see 3 to 10 times the call volume of a human-only team, and the point of that volume is to feed the humans better conversations, not to remove them. The customer reactivation solution page shows what the handoff looks like in a live campaign.

7. When doesn't the AI cost case fit?

The AI cost case is weakest when the list is small, the consent data is missing, or the desk is already paid for and idle. A broker with 300 dormant accounts saves a few hundred dollars by switching, which is not worth a procurement cycle. A broker with no marketing-consent flags on the list can't run either channel until the data is fixed.

Idle in-house desk. If a retention team is on payroll with capacity to spare, the marginal cost of their next call is close to zero, and the AI comparison only holds for volume the desk can't reach. The number to use is the cost of the extra attempts, not the whole payroll.

Advice conversations. Calls that would cross into investment advice under FCA, CySEC or ASIC rules belong with a licensed human, and the AI script should be written so it never gets there. The cost comparison doesn't apply to conversations one side isn't allowed to have.

Broken data. An export from the trading platform with wrong country codes sends both the human and the AI campaign into dead numbers, and the AI bill for calling dead numbers is small but not zero. Clean the list first; the reactivation list data checklist covers the fields that matter.

The math above is the version to take into the budget meeting: $0.53 an attempt against $2.57, $10,500 against $51,300 on a 10,000-trader list, and the humans redeployed to the conversations that were worth their hour. If you'd rather see it built on your own book, book a 30-minute call and Topcalls sends a proposal within 48 hours.