How many agents do you need for 200 calls an hour at 80/20? 17 on the phones and 25 scheduled at a 4-minute AHT. Erlang C math, tables and AI scenarios.
Hannah Owen
·
Updated
You need 17 agents on the phones to answer 200 calls an hour at an 80/20 service level (80% of calls answered within 20 seconds), assuming a 4-minute average handle time (AHT). With 30% shrinkage for breaks, training, meetings and leave, you schedule 25 agents to keep 17 available.
Those 17 answer 81.0% of calls within 20 seconds, so there is one point of margin: at 16 agents the service level falls to 69.0%. If an AI agent resolves 30% of the calls end to end, the 140 calls an hour left for people need 13 agents on the phones and 19 scheduled. Every number on this page comes from the standard Erlang C formula, and you can change any input in our free Erlang C calculator, which opens prefilled with this example.
Lorikeet publishes this guide and is one of the platforms compared: it makes two of the tools mentioned, the free Erlang C calculator at lorikeet.tools and Workforce Manager. We checked other tools on their own websites in November 2026; if something has changed, tell us and we will update the page.
Key takeaways
200 calls an hour at 4 minutes is 13.3 erlangs of work. Erlang C needs 17 agents for 80/20, with 78.4% occupancy and a 16.8 second average speed of answer.
Schedule 25, not 17. At 30% shrinkage, 17 / 0.7 = 24.3, rounded up.
Handle time moves the answer most. The same 200 calls need 14 agents at 3 minutes and 25 at 6 minutes.
Taking calls away does not cut agents in proportion. If AI resolves 30% of calls, you need 13 agents, not 12: smaller queues are less efficient.
The assumptions behind the answer
Input | Value used | Why |
|---|---|---|
Calls offered | 200 per hour | The question; 100 per 30 minutes gives the same answer |
Average handle time | 240 seconds (4 minutes) | Talk time plus wrap-up; use your own figure |
Service level target | 80% answered within 20 seconds | The common 80/20 target |
Shrinkage | 30% | An assumption; measure your own |
Abandonment | None | Erlang C assumes every caller waits |
How do you calculate agents for 200 calls an hour?
You convert calls into workload, then find the smallest number of agents that meets the service level, then add shrinkage. Four steps:
Workload. 200 calls x 240 seconds / 3,600 seconds = 13.3 erlangs. That is 13.3 hours of phone work arriving every hour, so you need more than 13 agents or the queue grows without limit.
Probability of waiting. The Erlang C formula gives the chance a caller has to queue. At 17 agents it is 25.7%, so about three in four calls are answered immediately.
Service level. Service level = 1 minus the probability of waiting x e^(-(agents minus workload) x 20 / 240). At 17 agents that is 81.0%, the first count to clear 80%.
Shrinkage. Divide by 1 minus shrinkage and round up: 17 / 0.7 = 24.3, so 25 scheduled.
Call Centre Helper's worked Erlang C example walks through the same formula step by step, including a maximum occupancy check, if you want to do it by hand.
What happens with one agent more or fewer?
Erlang C is a cliff, not a slope. Here is the same hour at 200 calls and a 4-minute AHT:
Agents on the phones | Answered within 20 seconds | Average speed of answer | Occupancy |
|---|---|---|---|
15 | 50.8% | 81.4 seconds | 88.9% |
16 | 69.0% | 34.8 seconds | 83.3% |
17 | 81.0% | 16.8 seconds | 78.4% |
18 | 88.8% | 8.5 seconds | 74.1% |
19 | 93.6% | 4.4 seconds | 70.2% |
One absence turns a passing hour into a failing one, which is why many teams staff an 18th agent in their riskiest hours.
What if your handle time or target is different?
Average handle time | 80/20: on the phones | 80/20: scheduled at 30% | 90/30: on the phones | 90/30: scheduled at 30% |
|---|---|---|---|---|
3 minutes | 14 | 20 | 14 | 20 |
4 minutes | 17 | 25 | 18 | 26 |
5 minutes | 21 | 30 | 22 | 32 |
6 minutes | 25 | 36 | 26 | 38 |
Shrinkage matters as much as the target: 17 agents on the phones means 23 scheduled at 25% shrinkage and 27 at 35%. Our guide to shrinkage, adherence and occupancy shows how to measure your own rate.
How many agents if an AI agent resolves some of the calls?
If an AI agent resolves part of the 200 calls end to end, only the rest reach your queue, but agents do not fall in proportion. The figures below are illustrative and keep the 4-minute AHT for the calls people still take:
Share of calls the AI agent resolves | Calls left for people | Agents on the phones (80/20) | Scheduled at 30% |
|---|---|---|---|
0% | 200 | 17 | 25 |
20% | 160 | 14 | 20 |
30% | 140 | 13 | 19 |
50% | 100 | 10 | 15 |
Two cautions. First, the calls an AI agent resolves are often the simpler ones, so the calls left for people can take longer: at 30% resolved and a 4.5-minute AHT on the rest, you need 14 on the phones and 20 scheduled. Second, plan the AI agent as a line in the roster with its own escalation path, not as a percentage taken off the forecast. Our guide to AI support staffing and volume planning works through this at a larger scale, and you can open the 140-call case in the calculator.
Does it matter if calls arrive unevenly within the hour?
Yes, a lot. The answer of 17 assumes the 200 calls arrive at a steady rate. Splitting the hour into two 30-minute intervals changes nothing if the calls split evenly, because 100 calls per 30 minutes is the same 13.3 erlangs. If 120 of the 200 arrive in the first half hour and 80 in the second, the first half needs 20 agents and the second needs 14. Staffed at a flat 17, the first half hour answers only 32.2% of calls within 20 seconds, while the hourly average still looks close to target. This is why the forecast that feeds the roster should be by interval, not by hour or by day.
Does the same math work for live chat?
Roughly, with one change: divide handle time by the number of chats an agent runs at once. 200 chats an hour with an 8-minute handle time and two chats at a time has the same 240-second effective handle time, so it also needs 17 agents for 80/20. Call Centre Helper's Erlang C for chat guide notes that handle time goes up as concurrency rises, so measure handle time at the concurrency your agents actually run.
What Erlang C does not tell you
It treats the hour as flat. If calls bunch after an outage notice, one 15-minute interval fails while the hourly figure passes. Forecast by 15 or 30 minute interval; our guide to forecasting support volume covers how.
It assumes nobody hangs up. With abandonment, Erlang C slightly overstates the agents you need; Erlang A models it.
It does not build the roster. Turning 25 scheduled agents into shifts that follow the call curve, respect rest rules and leave, and cover each channel is a separate job.
How we chose and checked these tools
We computed every figure on this page with a short Erlang C script (no abandonment, smallest whole number of agents that meets the target, shrinkage applied by dividing and rounding up) and cross-checked the main and AI cases against the lorikeet.tools calculator. The table lists free Erlang tools that answer this question, described from each tool's own page, checked on 6 November 2026.
Tool | What it does | Chat concurrency | Shrinkage | Abandonment (Erlang A) |
|---|---|---|---|---|
Agents, service level, speed of answer and occupancy, with the working shown, a day view and a free JSON API; no sign-up | Yes | Yes | No | |
Staff needed for a service level, with maximum occupancy and Excel downloads | Not published | Yes | Yes | |
Agents from call volume, handle time and service level target | Not published | Not published | Not published | |
Pre-worked answers for set call volumes, plus a calculator | Not published | Yes | Not published |
If abandonment is high in your queue, Call Centre Tools' Erlang A option is the better fit; for phone and chat with shrinkage and no sign-up, the lorikeet.tools calculator covers it.
Is Lorikeet the right fit?
You do not need Lorikeet to answer this question; the calculator is free. Lorikeet fits the next step: Workforce Manager builds the roster from your ticket and call volume by channel, day and hour and checks each draft week against contracted hours, rest between shifts, channel skills and approved leave, according to the launch post. Workforce Manager now "manages the day as it happens" (in-day management), per Lorikeet. It is not the right choice for everyone:
If you only need a staffing number, use a calculator and keep your current roster tool.
If you need a published Workforce Manager price or integrations list before a trial, neither is published.
If your team is under 20 agents, Workforce Manager is built for 20 or more, and a lighter tool will do.
Workforce Manager pricing is not published. For the AI concierge that resolves calls, Lorikeet publishes its pricing: the Start plan is $2,100 a month and Scale is $5,100 a month (both billed annually), with resolved chat, email and SMS conversations at $0.99 (Start) or $0.90 (Scale) and voice at $1.50 or $1.20. Unresolved tickets and handoffs to a person are not charged. Signature is custom, with a $75K platform fee. Full details are on the pricing page. To see both on your own volume, start a 30-day free trial.
Frequently asked questions
How many agents do I need for 200 calls an hour with an 80/20 service level?
17 agents on the phones at a 4-minute average handle time, answering 81.0% of calls within 20 seconds. With 30% shrinkage you schedule 25. At 3 minutes you need 14 on the phones, and at 5 minutes, 21.
Can 16 agents handle 200 calls an hour?
Not at 80/20 with a 4-minute handle time. Sixteen agents answer 69.0% of calls within 20 seconds and 15 answer 50.8%. Seventeen is the first count that clears 80%.
Why schedule 25 agents when Erlang C says 17?
Erlang C counts agents available to take calls. Shrinkage is paid time spent on breaks, training, meetings and leave. At 30% shrinkage, 17 / 0.7 = 24.3, which rounds up to 25 scheduled.
How many agents do I need if an AI agent resolves 30% of calls?
About 13 on the phones and 19 scheduled, if the 140 calls left keep a 4-minute handle time. If those calls take 4.5 minutes because the simpler ones went to the AI agent, you need 14 and 20. These figures are illustrative.
Does Erlang C work for live chat?
As an approximation. Divide handle time by the number of chats an agent runs at once: 200 chats an hour at 8 minutes with two at a time is a 240-second effective handle time, which also needs 17 agents for 80/20.
Is there a free Erlang C calculator for this?
Yes. The Lorikeet Erlang C calculator at lorikeet.tools is free with no sign-up, covers phone and chat with shrinkage, and opens prefilled with the 200 calls an hour example from this page.
