Customer Support Agent Productivity Statistics 2026

---

Agent productivity data: why the numbers matter

Workforce costs run 60 to 75 percent of most support operating budgets. How much each agent gets done in a shift -- and at what quality -- is what drives headcount, outsourcing, and automation decisions. Get the benchmark wrong and you staff incorrectly in either direction.

The data below covers tickets resolved per day, handle time by channel, occupancy and utilization targets, AI assist impact, knowledge base productivity effects, and fully loaded cost per agent. Sources are Zendesk, Gartner, Forrester, SQM Group, Salesforce, and NICE CXone. Where sources diverge, that range is noted with attribution rather than averaged into a single number that obscures the disagreement.

For what each ticket costs to process, see customer support cost per ticket benchmarks for 2026. For context on the volume growth behind those workloads, see customer support ticket volume statistics for 2026.

---

Tickets resolved per agent per day

Tickets per shift is the most cited productivity benchmark, and it is also the most easily misread. Channel mix alone swings the number by a factor of five -- which is why a single blended average tells you almost nothing about whether a specific team is performing well or badly.

The chat range of 40 to 80 is what makes aggregate comparisons slippery. Agents running three to five concurrent chat sessions close far more contacts than phone agents, but the ceiling has less to do with individual skill than with the platform's concurrency settings and session complexity. Gartner's 2025 customer service survey notes that the high end of chat productivity figures generally reflects teams where first-contact automation has already filtered routine queries -- so the contacts reaching agents are pre-triaged but also harder than average.

Phone teams face a physical ceiling: one agent, one call, with handle time determining how many contacts fit in a shift. At a 7-minute average handle time and 80 percent occupancy across 8 hours, a phone agent can theoretically complete 54 calls. After accounting for after-call work, breaks, and wrap-up, 10 to 15 resolved tickets is what most operations actually see.

---

Average handle time benchmarks

Average handle time (AHT) combines talk time or response time, hold time, and after-call work. Along with occupancy, it sets the ceiling on how many contacts a given headcount can process.

SQM Group's 2025 benchmarking study covers more than 500 North American contact centers. The voice AHT median lands at 7.5 minutes for routine contacts. What makes the SQM data worth reading carefully: the top quartile of centers by first-call resolution (FCR) ran AHTs 12 to 18 percent above the industry average. That cuts against the assumption that faster is better.

SQM Group also found that centers reducing AHT more than 20 percent below benchmark -- without corresponding FCR gains -- saw CSAT scores drop by an average of 8.3 points. The speed comes from agents wrapping conversations before issues are fully resolved. The customer calls back.

Gartner's 2025 survey data shows chat AHT has been climbing: average session length grew 11 percent between 2023 and 2025. The reason is the same one that drives the ticket volume growth -- automation absorbed simple queries, so the contacts reaching agents are harder. The workload per interaction is going up even as total volume growth continues.

---

Agent occupancy and utilization rates

Occupancy rate is the share of logged-in time agents spend on active work -- handling contacts, completing after-call work, or doing assigned tasks. Utilization is different: it captures the ratio of hours worked to hours scheduled. Both matter, and conflating them tends to produce bad staffing models.

Gartner's 2025 workforce data is direct on this: teams running occupancy above 90 percent consistently show higher error rates, longer handle times on complex contacts, and accelerated burnout. The short-term efficiency gain from pushing past 85 percent is smaller than it looks once you account for quality degradation and the downstream rework it generates.

At 80 percent occupancy, an agent logged in for 8 hours spends about 6.4 hours on active work and 1.6 hours available but idle. That idle time is not waste. It is the buffer that absorbs volume spikes, system lag, and contacts that run long. Teams that eliminate all idle time find that a single unusual day turns manageable workload into a queue that takes two days to clear.

Forrester's 2025 analysis is useful here because it separates the two metrics explicitly. Occupancy answers "are agents busy when they are logged in?" Utilization answers "are we getting the scheduled hours we are paying for?" You can have 85 percent occupancy and low utilization if scheduling leaves large gaps between shifts. Managing one while ignoring the other generates costs that do not show up until turnover or quality audits surface them.

After-call work above 35 percent of total AHT warrants investigation. It usually points to inadequate case management tools, unclear escalation paths, or agents manually completing steps that should be automated -- logging contact details into a CRM that does not integrate with the contact platform being the most common example.

---

Impact of AI assist on agent productivity

AI assist tools have moved from pilot programs to standard deployment in a meaningful share of contact centers over the past two years. The category covers response suggestion copilots, auto-populated case fields, real-time knowledge base surfacing, post-call summarization, and sentiment analysis.

Zendesk's 2025 CX Trends report found agents using AI copilot tools closed tickets 35 percent faster than agents without them, measured across 100,000+ agent-hours. That headline number applies to routine, low-complexity contacts where the AI suggestion is high confidence and the agent accepts it with minimal editing. Zendesk's footnotes are clear on this.

For complex contacts -- billing disputes, technical troubleshooting, emotionally charged calls -- the productivity gain narrows to 10 to 18 percent. The suggestion is more likely to get rejected or substantially rewritten, and reviewing irrelevant suggestions takes time that partially cancels out the time saved when the suggestion is actually useful.

Post-call summarization numbers are more consistent. Manual after-call notes run 4 to 6 minutes per voice contact. Automated summarization with human review cuts that to under 60 seconds in centers with good CRM integration. On a team handling 200 calls per day, that alone adds hours of effective capacity back without changing headcount.

Forrester's 2025 technology index found 47 percent of enterprise contact centers have deployed some form of AI agent assist, up from 29 percent in 2023. Among those deployers, 76 percent rated the tools as meeting or exceeding productivity expectations. The performance gap between deployments was largely explained by change management: teams that trained agents on how to work with AI suggestions saw 40 percent better outcomes than teams that rolled out the tools without structured training.

---

Impact of knowledge base quality on agent productivity

How fast an agent can find the right answer affects handle time on nearly every contact involving a non-trivial question. The quality of the knowledge base determines whether that search helps or slows things down.

SQM Group's 2025 data shows a step-function relationship between knowledge base maturity and first-call resolution. FCR is the metric most tightly correlated with both customer satisfaction and total support cost: contacts requiring a callback or follow-up ticket cost roughly 2x what a first-contact resolution costs.

Gartner puts the ceiling for AI-integrated knowledge bases at 30 to 40 percent AHT reduction relative to no structured KB. Against a mature but non-AI KB, AI surfacing adds another 10 to 15 percent.

The standard failure mode is content decay. Forrester's 2025 survey found the median enterprise knowledge base has a 34 percent article staleness rate -- more than one in three articles outdated, incorrect, or irrelevant to the current product version. When staleness passes 25 percent, usage drops sharply. Agents learn through repeated bad experiences that the articles cannot be trusted, so they stop looking and default to memory or asking colleagues.

Maintaining a usable knowledge base requires someone to own it. Forrester found that organizations with an assigned knowledge base manager had 22 percentage points lower staleness rates and 18 percent higher FCR than organizations where upkeep was treated as everyone's responsibility and therefore no one's.

---

First-contact resolution rates

SQM Group treats FCR as the most predictive productivity metric for both customer satisfaction and long-run cost structure. The case is simple: contacts that require a callback or escalation cost more and produce worse outcomes than contacts resolved on the first attempt.

SQM Group's 2025 data finds that each 1 percent FCR improvement corresponds to roughly a 1 percent reduction in total contact center operating costs, because repeat contacts require re-authentication, full issue re-review, and often supervisor involvement.

At a 72 percent industry-average FCR, 28 percent of contacts need at least one follow-up. In a center handling 10,000 contacts per month, that is 2,800 repeat contacts -- each costing 40 to 60 percent more than first-contact resolution.

SQM Group points to three FCR levers with the strongest evidence: knowledge base maturity, agent training quality, and escalation authority. Agents with explicit authority to resolve issues at their tier without needing supervisor approval consistently post FCR rates 8 to 12 percentage points above agents in centers with restrictive escalation policies.

---

Cost per agent: all-in figures for 2026

Salary alone understates what a support agent actually costs. The fully loaded figure includes payroll taxes, benefits, tools, training, and a share of management overhead.

Forrester's 2025 workforce cost analysis puts the all-in median for a US support agent at about $63,000. The $52,000 to $78,000 spread reflects geography (coastal metros pull the top of the range), benefits levels, and tool spend.

Software costs have grown. Zendesk's 2025 benchmark data shows spend per agent increased 22 percent between 2022 and 2025, led by AI assist licensing, contact center platform upgrades, and CRM fees. Centers with a full AI assist deployment can exceed $7,000 per agent annually in software costs -- though that figure is usually evaluated against the reduction in contacts handled per agent required as AI absorbs volume.

For offshore comparison: Forrester puts fully loaded annual cost for Philippines-based agents at $14,000 to $22,000 and India-based agents at $12,000 to $18,000. Those figures cover local benefits norms and tools but exclude the coordination, QA, and management overhead that running an offshore team from a US or European headquarters adds. When organizations account for that overhead, the effective offshore cost typically runs 25 to 35 percent higher than the per-agent figure alone.

For context on how turnover inflates these cost figures through repeat onboarding and training cycles, see customer support agent turnover statistics for 2026.

---

Agent productivity by team size and structure

Gartner's 2025 survey includes a segment breakdown that most productivity benchmarks skip: how does team size and management ratio affect what individual agents actually produce?

The supervisor-to-agent ratio finding is worth taking seriously. Teams above a 1:20 ratio show lower FCR and slower performance improvement over time, because coaching conversations become infrequent. What Gartner also found: the gap between high-performing and low-performing agents grows wider in under-supervised teams. Top performers self-manage. Lower performers stay stuck without feedback.

On remote teams: Gartner's 2025 data shows no statistically significant productivity difference between well-managed remote and in-office teams when tooling is equivalent. Where a performance gap appears, it traces to tooling or management quality, not to the location itself.

---

What the data actually means for staffing decisions

A few conclusions hold across the sources:

Occupancy above 85 percent costs more than it saves. The math on 90-plus percent looks attractive on a headcount model and usually looks different in the actual cost analysis once error rates, repeat contacts, and turnover are factored in.

AI assist is a capacity multiplier, not a simple headcount replacement. The 25 to 35 percent handle time reduction documented by Zendesk means each agent can handle more contacts at the same quality level. Organizations that redeploy that capacity toward harder contacts or extended coverage get better outcomes than those that immediately cut headcount after deployment, because complexity per contact tends to rise as automation deflects simpler requests.

FCR has more leverage on cost than AHT. A contact center optimizing for tickets per agent is optimizing throughput. One optimizing for FCR is reducing total contacts while also improving satisfaction. SQM Group's 2025 data finds that a 5 percent FCR improvement produces more operating cost reduction than a 10 percent AHT reduction, because FCR eliminates contacts entirely rather than just making each one slightly cheaper.

For context on how ticket volumes are shaping the workloads behind these productivity numbers, see customer support ticket volume statistics for 2026. For the cost-per-ticket picture that connects these benchmarks to unit economics, see customer support cost per ticket benchmarks for 2026.

---

Sources

• Zendesk Benchmark Report 2025
• Zendesk CX Trends 2025
• Gartner Customer Service Survey 2025
• Forrester Customer Service Index 2025
• Forrester Total Workforce Cost Analysis 2025
• SQM Group Contact Center Benchmarking Study 2025
• Salesforce State of Service 2025
• NICE CXone Industry Report 2025