The real cost per contact (and why most teams undercount it by 40%)
Most support orgs divide agent salaries by ticket volume and call it cost per contact. The fully loaded number is much larger, and it changes how you should evaluate AI pricing.
I spend a lot of time with support leaders and their finance partners, and almost every conversation involves a number called cost per contact. Almost every time, that number is wrong in the same direction.
The usual calculation is agent salaries divided by tickets. For a team of 100 agents at $55,000 each handling 1.2 million contacts a year, that gives $4.58.
The real number for that team is closer to $7.60. The difference isn't a rounding error. It decides whether an AI investment pays back in four months or never.
This post walks through the full calculation, shows how per-resolution pricing changes it, and explains why we price Pazau the way we do.
Fully loaded cost per contact
Here is a worked example for a mid-sized, US-based support team with realistic 2026 costs.
| Cost line | Annual | Notes |
|---|---|---|
| Agent base salaries | $5,500,000 | 100 FTE × $55,000 |
| Benefits and payroll taxes | $1,540,000 | 28% of base |
| Team leads and managers | $1,120,000 | 1 lead per 12 agents, 1 manager per 50, loaded |
| QA, WFM and training staff | $610,000 | 6 FTE, loaded |
| Recruiting and onboarding | $420,000 | 35% annual attrition × $12,000 per hire |
| Ramp productivity loss | $290,000 | New hires at ~55% productivity for 8 weeks |
| Helpdesk and telephony software | $380,000 | Seats, IVR, CCaaS minutes |
| Workspace, equipment and IT | $260,000 | Hybrid, blended |
| BPO overflow and seasonal surge | $0 – $900,000 | Many teams leave this out entirely |
| Total (excl. BPO) | $10,120,000 | |
| Contacts per year | 1,330,000 | Including repeat contacts |
| Fully loaded cost per contact | $7.61 |
Three lines are most often missing from the simple version.
Attrition. Contact center attrition is 30–45% a year across most industries. Each departure costs the recruiting fee plus two months of an employee who isn't yet fully productive. For our example team, that's $710,000 a year that never shows up in a salary line.
Management overhead. Leads, managers, QA reviewers and workforce planners scale with headcount. Their cost goes up and down with the team they support.
Repeat contacts. This one belongs in the denominator, and it's the most important. If 18% of your contacts are customers following up on the same unresolved issue, you aren't handling 1.33 million problems. You're handling about 1.09 million problems at a real cost of $9.28 per problem solved.
Cost per contact vs. cost per resolution
That last point is why we think cost per contact is the wrong metric. It rewards activity. A team can lower its cost per contact by handling more contacts, including contacts it caused by not solving the problem the first time.
The number that matters is cost per resolved problem:
cost per resolution = total support cost / problems resolved
problems resolved = contacts − repeat contacts − abandoned contacts
As we argued in Resolution, not deflection, an AI tool that "deflects" a conversation the customer comes back with tomorrow hasn't saved anything. It has moved the cost to a more expensive channel and lowered CSAT along the way.
How AI changes the math
Here's the same team a year after deploying Pazau Agent at 62% end-to-end resolution. That's a little below our 67% platform average, which is typical for a first year.
| Before | After (year one) | |
|---|---|---|
| Contacts per year | 1,330,000 | 1,190,000 |
| Repeat contact rate | 18% | 9% |
| Resolved by Pazau Agent | — | 738,000 (62%) |
| Handled by people | 1,330,000 | 452,000 |
| Agent FTE | 100 | 58 |
| People cost (fully loaded, incl. mgmt, attrition, workspace) | $9,740,000 | $5,650,000 |
| Software (Business plan, 58 seats) | $380,000 | $104,000 + $140,000 CCaaS |
| AI resolutions (738,000 × $0.75) | — | $553,500 |
| Total | $10,120,000 | $6,447,500 |
| Fully loaded cost per contact | $7.61 | $5.42 |
| Cost per resolved problem | $9.28 | $5.95 |
Total cost is down 36%, and cost per contact is down 29%. Contacts themselves also fell 11%, because repeat contacts halved. When problems get solved the first time, customers don't write back.
Across the Pazau customer base, the median reduction in cost per contact in year one is 41%. Our example comes in below that because we used conservative assumptions. Many teams also shift work from BPO overflow, which we didn't count here.
The FTE reduction in this example comes from not backfilling attrition over 12 months, not from layoffs. That's how most of our customers get there. At 35% attrition, a team can reach the new size within a year simply by not refilling roles.
Lumen & Oak: the exceptions are where the cost is
Lumen & Oak sells furniture online, and its most expensive contacts are delivery exceptions. A sofa arrives damaged. A freight carrier misses a two-hour delivery window. A dining table ships without its hardware kit. Each of these used to take an average of 3.4 contacts and 22 minutes of agent time across email, phone and follow-ups with the carrier.
With Pazau Agent handling delivery exceptions end-to-end through Actions into its order system and freight carriers, Lumen & Oak cut cost per contact by 38%. It handled 2.1 contacts per exception instead of 3.4, and 74% of exceptions were resolved without a person. More on that in the Lumen & Oak story.
Why we charge per resolution
Here's the part where the CFO explains the pricing.
Pazau charges a seat price for the people on your team ($89 per month on Growth, $149 on Business) plus a per-resolution fee for Pazau Agent: $0.90 on Growth, $0.75 on Business, and $0.99 for Agent-only. Handoffs to people are free. Conversations that come back within 7 days don't count as resolutions. Neither do abandoned conversations or conversations where the customer asks for a person.
We chose this model because every other one misaligns incentives somewhere.
| Pricing model | Vendor gets paid when… | What goes wrong |
|---|---|---|
| Per seat only | You hire more agents | The vendor benefits from your headcount |
| Per conversation / per message | The AI talks | Longer, chattier conversations cost you more |
| Per "deflection" | The customer stops replying | Customers giving up counts as success |
| Per resolution | The problem is actually solved | We only earn when you save |
Per-resolution pricing means our revenue follows your savings, and we pay for our own mistakes. If Pazau Agent gets something wrong and the customer comes back, we don't bill for the first attempt. That's a real cost to us. It's also why we invested in simulation before publish, the 7-day repeat window, and Insights alerts on repeat-contact spikes.
It also makes the return easy to calculate. If a person resolving a problem costs you $9.28 fully loaded, and Pazau Agent costs $0.75, each AI resolution saves $8.53 before software costs. You don't need a complicated ROI model to see that.
How to do this for your own team
If you want to rebuild your own number, here's the short version:
- Start with total cost, not salaries. Include benefits, management, QA, WFM, recruiting, attrition, software, BPO and facilities.
- Count repeat contacts honestly. Match on customer and topic within 7 days. Most helpdesks can export this. If yours can't, sample 500 tickets by hand.
- Compute cost per resolved problem, not just cost per contact.
- Model AI savings with a strict resolution definition. A vendor who quotes a "deflection" or "containment" rate is quoting a number that will look worse on your P&L.
- Build attrition into the timeline. Savings come from not backfilling, so they arrive over 9–15 months, not on day one.
Our 2026 AI Resolution Benchmark has cost per contact before and after deployment by industry, if you want to compare your number with peers. Our pricing page has a calculator that runs this same model with your inputs.
Support is one of the largest operating costs in most consumer businesses. It's worth measuring it correctly.