What the agent is actually worth.
Not a smaller team — a bigger one, without hiring. Put in your volume; we'll price the hours the agent hands back against the plan your traffic lands on.
Every assumption is sourced and editable. Nothing asks for your email.
Across every channel. That's about 9,600 incoming messages — the unit Cloodot meters.
Closed without a person touching them. 30% is where an untuned agent lands.
Salary plus benefits and payroll taxes — the loaded cost, not the offer letter.
- $30an hour of loaded payroll
- 2.0 minof paid time per conversation
Assumptions
The constants behind the number. Your own analytics beats every one of them — overrule away.
At 3 at a time, that is 2.0 minutes of paid time per conversation.
Threads one person runs in parallel. A higher number flatters us, so it starts low.
Your own inbox timestamps beat every published figure for this one.
Most out-of-hours messages are existing customers, who wait happily until morning. This is the thin slice of new enquiries that goes elsewhere first. Only used if you price a customer.
What the year looks like.
Value returned each month against what the plan costs over the same period. Straight lines, because neither compounds — arithmetic, not a forecast.
- Team time returned$4,292at your payroll cost, over twelve months
- Growth plan$2,988billed monthly, twelve months
Time only — revenue you priced is your estimate, not ours, so it is not plotted. Annual billing takes the plan side down further; the pricing page carries the discount.
Four claims, and one we won't make.
- 01
Capacity, not headcount
Gartner asked 321 service leaders what AI had done to their staffing. One in five had cut headcount. More than half kept the same team and put more volume through it. That second outcome is the one this page prices.
- 02
The hours nobody is working
A third of what arrives lands when the inbox is closed. Today it waits for morning. The agent answers at the moment the customer asked, on every channel alike.
- 03
The questions that repeat
The agent does best exactly where a person's time is worth least: where is my order, are you open, do you have this in stock. Auto FAQ mines those from your own inbox.
- 04
What we have not counted
Nothing here credits the copilot speeding up the conversations your team still handles, the campaigns and workflows on every plan, or a faster reply keeping a customer. All real, none in the total — we cannot put a defensible number on them for your business.
The whole formula, on one screen.
Four lines of arithmetic. Disagree with a number? The sliders above are where you say so.
- 1
What a conversation costs a person
Handle time ÷ threads run at once. Six minutes across three is two minutes of paid attention.
- 2
How many of those the agent takes
Conversations a month × the share it handles end-to-end. Those minutes are the headline hours.
- 3
What an hour of your payroll costs
Loaded annual cost ÷ 2,080 working hours. At $62,000 that is about $30 an hour.
- 4
What Cloodot charges for the volume
Conversations × 8 incoming messages is the metered volume. The cheapest plan covering it sets the price, read live from the catalog checkout uses.
Where every number came from.
The defaults, and why
| Assumption | Starts at | Why | Confidence |
|---|---|---|---|
| What one person costs a year | $62,000 | Median US customer-service pay was $44,770 in May 2025. Benefits and payroll taxes add 29.7%, putting the loaded cost near $62,000 — the number finance uses, not the salary line.US Bureau of Labor Statistics — OOH (May 2025) and ECEC (June 2025) | High |
| Share the agent handles end-to-end | 30% | Deliberately low. Vendors publish 60–80%, but public case studies run 26–56% and platforms tell new customers to expect 30–50% before tuning. A tuned agent on narrow, well-documented work beats 30% — that's the upside, not the promise.Metrigy 2024–25 study; published Gorgias case studies; Intercom’s own guidance to new customers | Medium |
| Minutes to handle one conversation | 6 minutes | The middle of the 6–7 minute average handle time reported for chat and messaging. Your analytics has your real number; if it is higher, so is everything below.Cross-industry customer-service benchmarking, 2024–25 | Medium |
| Conversations handled at once | 3 | Two to three threads at a time is typical; high performers reach six. This number divides the handle time, so a higher one would flatter us.Live-chat concurrency benchmarks (HDI, Comm100, Call Centre Helper), 2022–25 | Medium |
| Share arriving out of hours | 30% | Published figures run 40–60% for local and home services, none of them traceable to a primary study — so we start below the whole range. Your inbox timestamps beat all of it.Vendor and vertical reports, 2024–26 — no primary study located | Illustrative |
Primary sources
- US Bureau of Labor Statistics — Occupational Outlook HandbookMedian pay for customer service representatives: $44,770 a year, $21.53 an hour (May 2025).
- US Bureau of Labor Statistics — Employer Costs for Employee CompensationWages are 70.3% of total compensation in private industry; benefits are the other 29.7% (June 2025).
- Gartner — AI and customer service staffingSurvey of 321 customer service leaders: 20% report AI-driven headcount reduction, 55% report stable staffing against higher volume (December 2025).
- Gartner — self-service resolutionSurvey of 5,728 customers: only 14% of customer service issues are fully resolved in self-service (August 2024). The reason this page starts the agent low.
- Metrigy — the evolving role of AI in customer experienceIndependent study: AI fully automates around 20% of interactions today, with leaders projecting about 37% by 2028.
- Meta — WhatsApp Business Platform pricingPer-message pricing since 1 July 2025. Billed by Meta, not part of the Cloodot plan price.
And five we threw out
Everywhere in this category, and each broke when we chased it to its source. None is in the arithmetic above.
“WhatsApp has a 98% open rate” (and the identical 98% for SMS)
Two unrelated channels, the same round number, no traceable study behind either. The best real measurement — one vendor’s aggregate across 382 brands — puts WhatsApp nearer 79%.
“AI deflects 45% of queries but only 14% are fully resolved”
Gartner's 14% is real, from a 2024 survey of 5,728 customers. The 45% is from the same survey but measures something else — people who felt misunderstood. Aggregators stitched them into a statistic Gartner never published.
“Reply in 5 minutes and you are 21× more likely to qualify a lead”
A vendor-funded 2007 study, never peer-reviewed, routinely misattributed to a separate 2011 Harvard Business Review article. Nineteen years old and about outbound phone calls.
“Poor customer service costs US businesses $75 billion a year”
A decade-old round number with no locatable source. Impressive, unusable.
“62% of small business calls go unanswered”
Three mutually exclusive attributions for one figure, one of them a survey of 85 businesses. The signature of a statistic that was copied rather than measured.
The awkward questions.
Why does this not promise a smaller team?
Why start at 30% automation when vendors advertise 80%?
What exactly counts as a conversation?
Is the revenue figure your estimate or mine?
Does the plan price here match what I would actually pay?
What is not counted here?
Find out on your
own inbox.
Point the agent at your own content, watch a week of real conversations, and measure its share against the 30% this page assumed.