500 calls an hour at 80/20: Erlang C says 39 on the phones and 56 scheduled, but 85.5% occupancy makes 40 and 58 the practical minimum. Burnout and WFM math.
Hannah Owen
·
Updated
Erlang C says 39 agents on the phones and 56 on the schedule for 500 calls an hour at an 80/20 service level, a 4-minute average handle time (AHT) and 30% shrinkage. But those 39 would be busy 85.5% of the time, just over the 85% default cap in our calculator, so 40 on the phones and 58 scheduled is the practical minimum.
The 40th agent looks like a rounding decision. It lifts the share of calls answered within 20 seconds from 84.2% to 89.1% and adds about 7 seconds of breathing room between calls. At this size the staffing math is the easy part; the harder job is keeping 40 on the phones when calls bunch, people call in sick and handle time creeps up, which is where workforce management (WFM) software starts to earn its cost. You can test your own inputs in our free Erlang C calculator, which opens prefilled with this case.
Lorikeet publishes this guide and is one of the platforms compared: it makes the free calculator at lorikeet.tools and Workforce Manager, both mentioned here. We checked competitor details on their own websites in November 2026; if something has changed, tell us and we will update the page.
Key takeaways
Plan for 40 and 58, not 39 and 56. The workload is 33.3 erlangs. Forty agents run at 83.3% occupancy, under the 85% cap.
The cap is a choice. An 80% cap means 42 on the phones and 60 scheduled; with a 90% cap, 39 and 56 are allowed.
Fatigue feeds back into the math. If tired agents drift to a 4.5-minute AHT, the same 40 answer only 50.7% within 20 seconds.
Large queues live close to the edge. With 40 agents, 10% more calls than forecast takes you to 63.1%.
Why 40 agents and not 39?
At this volume, service level stops being the limit and occupancy takes over. The workload is 500 x 240 / 3,600 = 33.3 erlangs. Thirty-nine agents clear 80/20 comfortably, but 33.3 / 39 = 85.5% occupancy. Holding 85% takes 33.3 / 0.85 = 39.2, which rounds up to 40. The last column shows the average idle time an agent gets per call.
Agents on the phones | Answered within 20 seconds | Average wait | Occupancy | Idle time per call |
|---|---|---|---|---|
37 | 68.2% | 28.3 seconds | 90.1% | 26 seconds |
38 | 77.4% | 17.1 seconds | 87.7% | 34 seconds |
39 | 84.2% | 10.8 seconds | 85.5% | 41 seconds |
40 | 89.1% | 6.9 seconds | 83.3% | 48 seconds |
41 | 92.5% | 4.4 seconds | 81.3% | 55 seconds |
42 | 95.0% | 2.8 seconds | 79.4% | 62 seconds |
Shrinkage then sets the schedule: 40 / 0.7 = 57.1, so 58 scheduled, where 39 / 0.7 = 55.7 gives 56. The calculator applies the cap for you and reports 40 raised from 39; to see the uncapped result, open it with no cap. Our Erlang C staffing guide covers the formula, and the assumptions here match our pages for 100 calls an hour and 200 calls an hour.
Input | Used here |
|---|---|
Volume | 500 calls an hour, spread evenly |
Handle time | 240 seconds, talk plus wrap-up |
Target | 80% answered within 20 seconds |
Shrinkage | 30%, an assumption; measure your own |
Occupancy cap | 85%, the calculator's default |
Hang-ups | None: Erlang C assumes every caller waits |
What does 85% occupancy mean for the people on the phones?
Occupancy is the share of logged-in time spent handling calls. At about 85%, an agent takes about 12.5 calls an hour with less than a minute between them on average, and some gaps will be far shorter. Compare that with the two minutes between calls in a 10-agent queue at 100 calls an hour.
The 85% line is a convention, not a law of physics. Call Centre Helper's worked Erlang C example uses 85% as its maximum and warns that above it agents will likely burn out and handle time may rise to cover it. Call Centre Tools says that running over 85% to 90% for long periods gets hidden in a longer AHT, and Voxjar links very high occupancy above about 85% to 90% with burnout and attrition. These are practitioner guidelines; none of those pages cites a controlled study, so set your own cap and watch attrition and handle time against it.
Occupancy cap | On the phones | Scheduled at 30% | Answered within 20 seconds | Occupancy reached |
|---|---|---|---|---|
80% | 42 | 60 | 95.0% | 79.4% |
85% | 40 | 58 | 89.1% | 83.3% |
90% or no cap | 39 | 56 | 84.2% | 85.5% |
Moving from a 90% cap to an 80% cap costs 4 seats on the schedule, about 7% more paid hours. Whether that is cheap depends on what attrition costs you, which is worth estimating before you choose.
What happens if fatigue pushes handle time up?
This is the feedback loop those warnings describe. If 40 agents drift from 4 to 4.5 minutes, the workload becomes 37.5 erlangs, occupancy hits 93.8%, the service level falls to 50.7% and the average wait goes from 6.9 to 64.0 seconds. At 5 minutes the workload is 41.7 erlangs, more than 40 agents can carry, and the queue never clears. The table shows what each handle time really needs:
Average handle time | 80/20 before the cap | 80/20 with the 85% cap | Scheduled at 30% | 90/30 with the 85% cap | Scheduled at 30% |
|---|---|---|---|---|---|
3 minutes | 30 | 30 | 43 | 31 | 45 |
4 minutes | 39 | 40 | 58 | 40 | 58 |
4.5 minutes | 43 | 45 | 65 | 45 | 65 |
5 minutes | 48 | 50 | 72 | 50 | 72 |
6 minutes | 57 | 59 | 85 | 59 | 85 |
Shrinkage moves the schedule just as much: 40 on the phones means 54 scheduled at 25%, 62 at 35% and 67 at 40%. Our guide to shrinkage, adherence and occupancy shows how to track all three together, so a handle time rise gets noticed before it reaches the service level.
How many agents if an AI agent resolves part of the calls?
At this size the savings are close to proportional, unlike a small queue. These figures are illustrative, keep the 4-minute AHT on the remaining calls and apply the 85% cap:
Share the AI agent resolves | Calls left for people | On the phones | Scheduled at 30% | Occupancy |
|---|---|---|---|---|
0% | 500 | 40 | 58 | 83.3% |
20% | 400 | 32 | 46 | 83.3% |
30% | 350 | 28 | 40 | 83.3% |
50% | 250 | 21 | 30 | 79.4% |
70% | 150 | 14 | 20 | 71.4% |
Resolving 30% removes 12 of 40 phone seats, where at 100 calls an hour it removes 2 of 10. But notice the occupancy column: at 20% and 30% resolved, the people left are exactly as busy as before. If the simpler calls went to the AI agent and the rest take 4.5 minutes, 30% resolved needs 31 on the phones and 45 scheduled at 84.7% occupancy, so agents face a harder mix with no extra recovery time. Consider spending part of the saving on a lower cap. You can open the 350-call case, and our guide to AI support staffing and volume planning covers planning the two together.
Should you split 500 calls an hour into separate queues?
Only if the skills truly differ, because you pay for it. Two queues of 250 calls an hour need 21 agents each, so 42 instead of 40. Five queues of 100 need 10 each, so 50 instead of 40, a quarter more people for the same calls. Where you must split, cross-train a group that can move between queues, so you keep some of the pooled efficiency.
How fragile is 40 agents when the forecast misses?
More fragile than a small queue. High occupancy leaves little room, so a modest miss does large damage:
Calls in the hour | Against forecast | Answered within 20 seconds with 40 agents | Occupancy | Agents you needed (85% cap) |
|---|---|---|---|---|
500 | On forecast | 89.1% | 83.3% | 40 |
525 | 5% over | 79.3% | 87.5% | 42 |
550 | 10% over | 63.1% | 91.7% | 44 |
575 | 15% over | 37.9% | 95.8% | 46 |
600 | 20% over | Queue never clears | 100.0% | 48 |
Bunching has the same effect. If 300 of the 500 calls arrive in the first half hour, that half needs 48 agents with the cap and the second half needs 32, a 16-seat swing inside one hour, and a flat 40 has no spare capacity at all during the rush. Absences are the one area where size helps: 58 scheduled gives 40.6 on the phones on average, and it takes three unplanned absences (55 x 0.7 = 38.5) to drop below the 39 that clear 80/20. Our guide to forecasting support volume covers forecasting by interval.
Where does WFM software pay off at this size?
The answer is a count for one hour; the operation is a week of them. If you held 500 calls an hour across 60 open hours, 58 x 60 / 40 = 87 people on 40-hour contracts, as an upper limit because demand usually falls away from the peak. Four jobs then get too big to do reliably by hand:
Following the curve. A 16-seat swing inside an hour means shifts and breaks have to follow the forecast by interval, not sit flat across the day.
Checking the rules. Contracted hours, rest between shifts, skills and leave have to hold for every person after every swap, which is slow to check in a spreadsheet at this headcount.
Reacting on the day. A 5% miss costs two agents on the phones and takes you under 80/20, so someone has to spot it and move people within the hour.
Planning hires. Every 5% of volume growth adds about two agents on the phones and about three on the schedule, and recruiting takes time. Seeing the shortfall months ahead is worth more here than at 15 people.
A rough break-even test: Assembled publishes its Pro WFM plan at $45 per agent per month plus a platform fee, so 87 agents cost $3,915 a month before that fee. At an illustrative loaded cost of $25 an hour, that is about 157 paid hours, less than one full-time agent's month of about 173. Use your own rate, but if software stops you over-staffing by one seat, or prevents a resignation, it has likely covered that bill. It pays off less if you run one steady queue with a stable weekly pattern, where a well-kept spreadsheet may last longer than vendors suggest.
How we chose and checked these tools
We computed every staffing figure with a short Erlang C script (no abandonment, smallest whole number of agents that meets the target and the cap, shrinkage applied by dividing and rounding up) and matched the main, forecast-miss, handle-time and AI cases against the live lorikeet.tools calculator API on 6 November 2026. The table describes WFM tools that suit a team this size, from each vendor's own pages on the same date. It is not a ranking; our WFM software guide ranks them, and our mid-size scheduling guide looks at smaller teams.
Tool | Fits a team this size if | Published WFM price | In-day management |
|---|---|---|---|
You run digital-first support, including BPO users | $25, $45 or $75 per agent per month plus a platform fee | Real-time adherence and analytics | |
You have 20 or more agents on chat or phones | Not published; 30-day free trial | "Manages the day as it happens" | |
You already run on Zendesk | US$50 per agent per month (bundle, paid yearly) on top of a Zendesk plan | Real-time activity tracking and adherence | |
Calls are most of your volume | Not published | Intraday scheduling tools | |
You are moving off spreadsheets | $8 or $16 per user per month; Enterprise custom | Intraday chart and activity stats | |
You run a large center with a planning team | CXone suites $110 to $249 per agent per month | Yes | |
You are already on Genesys | $155 per user per month billed annually, includes WFM | Intraday monitoring and real-time adherence |
Is Lorikeet the right fit?
A team this size is in range for both things Lorikeet sells. Its AI concierge can take part of the calls, and Workforce Manager, built for 20 or more agents on chat or phones, builds the roster from ticket and call volume, checks each draft week against contracted hours, rest, skills and leave, and shows capacity shortfalls up to 12 months ahead, according to the launch post. It now also "manages the day as it happens", per Lorikeet. It is not the right choice for everyone:
If you run a multi-site center with WFM analysts and years of history in an incumbent system, NiCE or Verint fit better.
If you need a published per-agent price before you trial anything, Assembled and Soon publish theirs; Workforce Manager does not publish a price or an integrations list.
If hang-ups are a large part of your queue and you only need the number, Call Centre Tools' Erlang A option suits you better.
If most of your calls are simple FAQs, a lighter AI tool will be cheaper and faster to switch on.
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 a roster built from your own volume, start a 30-day free trial.
Frequently asked questions
How many agents do I need for 500 calls an hour with an 80/20 service level?
Erlang C says 39 on the phones at a 4-minute average handle time, answering 84.2% of calls within 20 seconds. Occupancy is then 85.5%, above an 85% cap, so 40 on the phones is the practical minimum. With 30% shrinkage you schedule 58.
Should I staff 39 or 40 agents for 500 calls an hour?
Use 40 if your occupancy cap is 85%. Forty agents answer 89.1% of calls within 20 seconds at 83.3% occupancy. Thirty-nine meet the 80/20 target but run at 85.5%. With a 90% cap, 39 and 56 scheduled are allowed.
What occupancy is too high for call center agents?
Common guidance puts the limit at about 85% to 90% for sustained periods, with burnout and rising handle time above it. It is a practitioner rule of thumb, not a tested threshold, so track attrition and handle time against your own cap.
How many agents should I schedule for 500 calls an hour?
58 at 30% shrinkage to keep 40 on the phones, since 40 / 0.7 = 57.1. Without the occupancy cap, 39 / 0.7 = 55.7 gives 56. At 25% shrinkage, 40 on the phones means 54 scheduled, and at 35% it means 62.
How many agents do I need if an AI agent resolves 30% of 500 calls an hour?
About 28 on the phones and 40 scheduled if the 350 calls left keep a 4-minute handle time. If those calls take 4.5 minutes, you need 31 and 45. These figures are illustrative and apply an 85% occupancy cap.
Is workforce management software worth it at 500 calls an hour?
Usually, if volume moves during the day. A 10% forecast miss takes 40 agents to 63.1% within 20 seconds, and 60/40 bunching within an hour is a 16-seat swing. It matters less for one steady queue with a stable weekly pattern.
Does splitting 500 calls an hour into two queues need more agents?
Yes. Two queues of 250 calls an hour need 21 agents each, so 42 in total instead of 40. Five queues of 100 need 50 in total. Split only when the skills truly differ.
Is there a free calculator for 500 calls an hour?
Yes. The Lorikeet Erlang C calculator at lorikeet.tools is free with no sign-up, applies an 85% occupancy cap by default, and opens prefilled with the 500 calls an hour case from this page.
