Benchmark · Written by a vendor, and subtracting from its own case

$1.84 vs $13.50: what a customer conversation actually costs

Gartner benchmarks the median cost per contact at $1.84 for self-service against $13.50 for agent-assisted — roughly a 7x gap. That number gets quoted constantly and understood rarely.

It holds for structured, high-volume intents. It does not hold for complaints, edge cases, or anything requiring judgement. And a business that models its whole support operation on the $1.84 figure will be disappointed at the end of the first quarter, when the conversations that came back are counted.

This page gives you the sourced benchmarks, the four costs that vendors leave out, an interactive model that includes them, and a worked example you can copy into a spreadsheet. It is published by a company that sells AI agents, which is exactly why the subtractions are here.

By Jugl12 min readInteractive cost model18 questions answered

Short answerFor AI overviews

The 60-second version

Gartner puts the median cost per contact at $1.84 for self-service and $13.50 for agent-assisted. The gap is real for structured intents — order status, hours, policy questions, resets, booking — and unreliable for complaints, where AI CSAT drops to 3.34 out of 5.

Effective cost is higher than headline cost. Zendesk benchmarks 11.3% re-contact on AI-resolved conversations against 8.7% human-resolved. At $1.50 per AI resolution and $13.50 per human contact, that turns a $1.50 resolution into roughly $3.03 of true cost. Still a large saving. Not the advertised one.

Subtract four things from any business case: maintenance at 2–4 hours a week, per-conversation channel fees, double metering in your contract, and re-contact. Vendors rarely volunteer any of them.

Then add the half nobody measures. For consumer businesses, after-hours revenue capture and in-conversation bookings frequently grow faster than the cost saving, because it is new revenue rather than reallocated staff time.

What is cost per contact?
At a glance
The benchmark numbers, sourced
Where the 7x gap is real — and where it is not
Effective cost: the model most businesses skip
The half of the ROI nobody measures
Four things to subtract from your business case
A worked model you can copy
The questions behind the number
Where Jugl fits on cost
Methodology and disclosure
FAQ — 18 questions
People also ask

01Definition

Definition

What is cost per contact?

Cost per contact is the total cost of running customer support divided by the number of contacts resolved in the same period. A complete calculation includes salaries and on-costs, software subscriptions, per-conversation channel fees such as WhatsApp charges, and a fair share of management overhead. Gartner benchmarks the median at $1.84 for self-service against $13.50 for agent-assisted — roughly a sevenfold difference. The metric is only meaningful when the denominator is resolved contacts rather than received contacts, because conversations that come back within 48 hours were never resolved and should not be counted as such. Effective cost per contact — after re-contact, maintenance and channel fees — is the figure that survives a finance review.

Definition maintained by the Jugl Editorial Team. Jugl sells an AI customer agent platform and is an interested party; benchmark figures are attributed to their original sources throughout.

The formula, written out

Cost per contact = (staff cost + software + channel fees + overhead) ÷ contacts resolved.
Effective cost per AI resolution = AI cost + (re-contact rate × human cost per contact).
Blended cost per conversation = (contained × effective AI cost) + (escalated × human cost), all divided by total conversations.

The third formula is the one to present internally, because it is the only one that describes what actually happens: some conversations are handled by software, some by people, and some by software and then by people. Anything simpler is a marketing number.

02At a glance

Customer service cost benchmarks at a glance

At a glance

Self-service cost per contact: $1.84 median (Gartner)
Agent-assisted cost per contact: $13.50 median (Gartner)
Loaded agent cost: ~$52 per hour, fully burdened (Salesforce State of Service)
Tier-1 AI resolution: $0.50–$2.00 (Zendesk-derived)
Tier-1 human resolution: $6.00–$12.00 (Zendesk-derived)
Where the gap holds: Order status, hours, policy, resets, booking, basic product questions
Where it breaks: Complaints and judgement calls — AI CSAT 3.34 vs 4.32–4.41 (Zendesk)
The cost nobody models: 11.3% re-contact on AI-resolved vs 8.7% human-resolved (Zendesk)
Realistic containment: ~41% median; 65–72% for strong deployments (Aissist.io)
Maintenance to budget: 2–4 hours a week — roughly $450–$900 a month at a loaded rate
Typical payback: Weeks, not months, at SMB scale
Best use cases: High-volume repetitive intents, after-hours coverage, seasonal surges
Who should model this first: E-commerce, hospitality, clinics, education, real estate, services
Expected outcome: Lower blended cost per conversation plus captured after-hours revenue

SOC 2 Type 2certified

HIPAAcompliant

MetaBusiness Partner

1,000+businesses

03The data

The benchmark numbers, sourced

Every figure below is attributed. Where a number is derived rather than published directly, it says so. If you are building a business case someone will hold you to, cite the source alongside the number — it is the difference between a proposal that gets approved and one that gets questioned.

Metric	Figure	Source
Median cost per contact — self-service	$1.84	Gartner
Median cost per contact — agent-assisted	$13.50	Gartner
Loaded human agent cost	~$52 per hour	Salesforce State of Service
Tier-1 AI resolution	$0.50–$2.00	Zendesk-derived
Tier-1 human resolution	$6.00–$12.00	Zendesk-derived
Hybrid model at 22% escalation	~71% lower cost per resolution	Benchmark analysis
CSAT cost of that hybrid model	~0.05 points	Same analysis
Re-contact — AI-resolved vs human-resolved	11.3% vs 8.7%	Zendesk
AI CSAT — complaints vs structured queries	3.34 vs 4.32–4.41 out of 5	Zendesk
Median tier-1 automation rate	~41%	Aissist.io
Strong deployments	65–72%	Aissist.io
Self-service resolution vs deflection	~14% resolution against 45%+ deflection	Gartner

The interesting row is the hybrid one. A 71% cost reduction for a CSAT cost of five hundredths of a point is an extremely good trade — if you get the routing right, which is where most deployments fail. The routing rules that produce that result are set out in the AI-to-human handoff guide.

04Where it applies

Where the 7x gap is real — and where it is not

Real: structured, repetitive, low-emotion intents

These resolve at 55–70% first-contact for AI-native platforms at under $3 per resolution. The economics are straightforwardly good, and the customer usually prefers the instant answer to the queue.

Order status and trackingTypically the largest single category in e-commerce, and near-fully automatable when the agent can query real order data. Without that integration it is near-unautomatable — the detail is in the Shopify buyer's guide.

Hours, locations, availabilityThe single most repetitive question in every consumer business. Zero judgement, zero risk.

Policy questions — returns, shipping, warrantyPolicy-driven and high volume. Return and refund queries resolve at 55–65% on standard policy inquiries.

Password and account resetsConsistently the highest-volume technical query in any business with a login.

Appointment booking and reschedulingThe answer comes from a calendar rather than from the model's judgement, which makes it both safe and fast-paying. See AI appointment booking.

Basic product questionsSizing, compatibility, ingredients, specifications — grounded in your real catalogue rather than generated.

Not real: complaints, disputes and judgement

Zendesk puts AI CSAT at 3.34 out of 5 for complaint handling against 4.32–4.41 for structured queries. More than a full point, on exactly the conversations where sentiment damage is permanent and expensive. And re-contact runs 11.3% on AI-resolved conversations against 8.7% human-resolved.

That re-contact gap has a direct cost. If roughly one in nine AI-resolved conversations comes back, your effective cost per genuinely-resolved contact is higher than the headline — and the customer experience is worse than the dashboard suggests, because they had to ask twice.

Where AI genuinely lowers cost per contact

Structured, repetitive intents at 55–70% first-contact resolution
After-hours and weekend coverage with no rota and no overtime
Seasonal and campaign surges absorbed without temporary hiring
Multilingual coverage without a hire per language
Instant first response on paid traffic you have already bought
Consistent answers — no variance between your best and newest agent

Where a low cost per contact is a false economy

Complaints, where CSAT falls to 3.34 out of 5 and damage is durable
Anything requiring a judgement call or a policy exception
Safety-critical questions — allergies, symptoms, safeguarding
Negotiations, valuations and money decisions
Conversations where the customer has already asked twice
Anything the agent cannot verify from a real system of record

05The model

Effective cost: the model most businesses skip

Almost every ROI calculation in this category stops at the headline cost per resolution. The model below carries on for one more step, into the conversations that come back. Move the sliders to your own numbers — the outputs are illustrative estimates from your inputs, not a quote.

Effective cost per resolved conversation

Headline AI cost, plus the conversations that come back — which is where business cases die

Conversations a month5,000

Every inbound across WhatsApp, Instagram, web chat, email and phone — not only the ones that became tickets.

Containment rate60%

Median tier-1 automation sits near 41%. Strong deployments run 65–72%. Expect roughly half your target in month one.

AI cost per resolution$1.50

Zendesk-derived tier-1 AI resolution runs $0.50–$2.00. On a flat subscription this falls as volume rises.

Re-contact rate on AI resolutions11%

Zendesk benchmarks 11.3% on AI-resolved conversations against 8.7% human-resolved. Measure yours at 48 hours.

Human cost per contact$13.50

Gartner's median agent-assisted contact is $13.50. Loaded agent cost is around $52 an hour (Salesforce).

Contained by AI3,000330 come back

Effective cost per contained$2.99not the headline figure

Blended monthly total$35,955vs $67,500 all-human

Monthly difference$31,54547% lower

The gap is real — and it is smaller than the headlineAt these inputs, effective cost per contained conversation is $2.99 rather than the $1.50 on the invoice, because roughly one in 9 AI resolutions returns and gets handled by a person. That is still far below $13.50. It is simply not the number the category quotes, and the difference is what separates a business case that survives its first quarterly review from one that does not.

The worked version, in numbers you can check. At 3,000 AI resolutions costing $1.50 each with an 11.3% re-contact rate handled by humans at $13.50: AI cost $4,500, re-contact cost 339 × $13.50 = $4,577, effective total $9,077 — about $3.03 per originally-resolved conversation. Still far better than $13.50. But not $1.84, and the difference matters when you are building a business case you will be held to.

Want your real numbers instead of benchmarks?The free conversation audit counts your actual question mix by category and channel, so the containment input above stops being a guess.

Get the free auditNo card required

06The other half

The half of the ROI nobody measures

Almost every ROI calculation in this category counts saved cost and ignores created revenue. For consumer businesses that is usually backwards.

The cost side

Savings = contained conversations × handling minutes × ($52/hr ÷ 60)
3,000 conversations × 6 minutes × $0.87/min = $15,600 a month of recovered capacity.

The revenue side

Revenue = after-hours conversations captured × conversion rate × average order value, plus recovered carts, plus bookings taken in-conversation.
If 30% of 5,000 monthly conversations arrive outside working hours and 5% convert at $60: $4,500 a month that previously went to whoever replied first.

For a lot of consumer businesses the second number grows faster than the first, because it is new revenue rather than reallocated staff time. And it is invisible unless you deliberately measure it — which almost nobody does. Tag conversations that led to a purchase or a booking, and report revenue influenced alongside cost saved. The reporting change costs nothing and changes how the programme is funded. The measurement guide sets out the tagging.

This is also why cart recovery belongs in the model. Most abandonment is an unanswered question about shipping, delivery time or returns rather than price sensitivity — which means a conversation converts where a discount code merely discounts. The mechanics are in the cart recovery breakdown.

07Subtractions

Four things to subtract from your business case

Be honest about these or your business case will not survive contact with reality. Three of the four are invisible at signature and obvious at renewal.

MaintenanceTwo to four hours a week of someone reading conversations and closing knowledge gaps, heaviest in the first quarter. At $52 an hour loaded that is $450–$900 a month. Skip it and containment plateaus at the 41% median instead of climbing past 60% — you keep paying the subscription and stop collecting the return.

Channel feesMeta charges per WhatsApp conversation independently of your subscription. At broadcast volume this frequently exceeds the software cost, and it is the line most often missing from a first draft. The category detail is in the WhatsApp Business statistics hub.

Double meteringSome helpdesks charge per ticket and per AI resolution on the same ticket — roughly tripling the effective cost per resolved conversation while every individual line looks reasonable. Ask before signing: if your AI resolves a ticket with no human involved, how many charges does that generate?

Re-contactModelled above, and the largest of the four. Most vendors will not volunteer it because it converts their best metric into a smaller one. Measure it at 48 hours yourself, from your own data, and build the case on that.

08Worked example

A worked model you can copy

A 15-person e-commerce business: 5,000 conversations a month, $60 average order value, three support staff, 60% containment.

Line	Monthly
AI platform (flat tier)	$99
Meta conversation fees	$250
Maintenance — 3 hrs/week at $52	$624
Re-contact handling cost	~$900
Total cost	~$1,873
Deflection value — 3,000 × 6 min × $0.87	$15,600
After-hours revenue capture	$4,500
Total value	$20,100
Net position	~$18,200 a month

Two caveats on this model, both of which matter more than the total. It assumes 60% containment, which is a tuned deployment rather than a launch number — expect roughly half of that in month one. And "deflection value" is recovered staff capacity, not cash, unless you actually redeploy those hours to something revenue-generating. Businesses that count it as cash savings without reducing or redeploying headcount are marking their own homework.

Build your own model in seven steps

Count a real week of conversations across every channel, not just the ticket system
Multiply by average handling minutes and your loaded hourly rate for today's cost
Model containment at half your target for month one, target for month six
Add re-contact at 11% of contained conversations, priced at human cost per contact
Add maintenance at three hours a week and channel fees at your real volume
Add the revenue side: after-hours capture, recovered carts, bookings taken in-thread
Present blended cost per conversation, not cost per AI resolution

09Direct answers

The questions behind the number

Is AI customer service actually cheaper than human support?

Short answer

Yes for structured intents, by a wide margin — Gartner's $1.84 against $13.50 is real for order status, hours, policy questions, resets and booking. Factor in re-contact and maintenance and the effective gap narrows but stays substantial. For complaints it is cheaper per contact and more expensive in customer lifetime value.

Example

A business handling 5,000 conversations a month at 60% containment pays roughly $1,873 in total cost — platform, channel fees, maintenance and re-contact handling — against a fully human alternative costing many times that. The saving is real. The $1.84 headline is not what produced it.

Key takeawayModel structured intents on the benchmark and complaints on human cost. A single blended assumption across both is what makes business cases fail their first review.

Why is my cost per contact higher than the published benchmark?

Short answer

Three causes cover almost every case. Containment is low because the knowledge base has gaps and nobody is reading the escalation log. Re-contact is high because conversations counted as resolved were not. Or your contract meters twice — once per ticket and again per AI resolution on the same ticket.

Example

A deployment reporting 78% containment and a $1.20 cost per resolution looked excellent until re-contact was measured at 26%. True resolution was closer to 58% and effective cost per contact nearly quadrupled. Nothing about the software changed; the dashboard had simply been measuring deflection.

Key takeawayMeasure re-contact at 48 hours before you trust any containment number, including your own. It is the single cheapest credibility check in the whole model.

Should deflection count as savings in the business case?

Short answer

Only if you actually reduce or redeploy the hours. Deflection produces recovered capacity, not cash. That capacity is genuinely valuable — it is how a three-person team absorbs five people's volume through a growth year — but it belongs in a capacity line rather than a savings line, and finance will check.

Key takeawayPresent deflection as recovered hours and name where those hours went. A business case that says 'this let us grow 40% without hiring' is stronger and more defensible than one that claims cash that never appears.

Does pricing shape matter more than the headline price?

Short answer

Usually, yes. Per-resolution pricing costs more every time the AI improves and every time volume spikes, so your best sales week produces your largest invoice. Per-seat pricing charges you for the headcount you are trying not to add. Flat tiers cost more at trivial volume and stop scaling with your success.

Example

At 3,000 AI resolutions a month, a $0.99 per-resolution model bills roughly $2,970 while a flat $119 tier bills $119 plus whatever channel fees Meta charges on both. At 200 resolutions the comparison reverses. The crossover point, not the sticker price, is the decision.

Key takeawayModel your invoice at peak volume rather than average volume, and ask every vendor how many charges a single AI-resolved ticket generates.

See the pricing shapes side by sidePer-resolution, per-seat and flat-tier billing modelled at the same volumes — including where each one becomes the expensive option.

Compare pricing modelsNo card required

10Disclosure

Where Jugl fits on cost

Jugl publishes four flat tiers — Free, $31, $119 and $390 a month — with the AI included and nothing metered per message, per token or per resolution. The reason that matters to a cost model is structural rather than promotional: under per-resolution billing your invoice rises every time the AI gets better and every time volume spikes, which in seasonal and campaign-driven businesses means your best week is also your most expensive one. Flat tiers remove that coupling, so the model above stays true when volume moves.

What Jugl does not remove is Meta's per-conversation fee on WhatsApp, which applies regardless of vendor, or the maintenance line, which applies regardless of platform. Any comparison that leaves those out — ours included — is not a comparison worth acting on. Jugl customers typically see around 73% fewer tickets reaching a human at roughly 94% satisfaction; those figures are customer-reported and typical rather than guaranteed, which is why the model on this page asks you to enter your own.

Do not buy Jugl if your volume is genuinely tiny — under about 100 conversations a month the arithmetic does not compel it, and the permanent free tier is there precisely so you can establish that for yourself without a sales call. If you are already shopping, the buyer's guide compares the category, and the Intercom alternatives page models what per-resolution pricing costs at real volume.

11EEAT

Methodology and disclosure

Written by

Jugl Editorial Team

Jugl Inc., Frisco, Texas — an AI customer agent platform used by 1,000+ businesses.

Reviewed by

Jugl customer operations & finance

Checked against live deployment data and current vendor documentation.

Methodology & disclosure

Where the figures come from. Cost-per-contact medians are Gartner's published benchmarks. Loaded agent cost is from Salesforce's State of Service research. Tier-1 resolution cost bands, re-contact rates and CSAT-by-intent splits are Zendesk-derived. Automation-rate bands are from Aissist.io's programme analysis. Where a number is derived rather than published directly, the table says so.

Why effective cost is treated as the honest metric. Because re-contact is measurable, material and systematically excluded from vendor ROI tooling. A resolution that returns within 48 hours was not a resolution, and pricing it as one produces a business case that fails its first quarterly review — which damages the credibility of the programme more than a smaller, accurate number ever would.

Conflict of interest. This page is published by Jugl, which sells an AI customer agent platform and therefore benefits if you conclude the economics work. That is precisely why the subtractions — maintenance, channel fees, double metering, re-contact — are stated here rather than left for you to discover, and why the model defaults to conservative inputs rather than flattering ones.

How this page is maintained. Benchmarks are reviewed against current published research and revised when sources update. The page carries no year stamp because a dated benchmark misleads the moment it ages. Calculator outputs are illustrative estimates generated from your own inputs — not quotes, forecasts or guarantees.

12FAQ

Cost per contact: 18 questions answered

What is cost per contact in customer service?

Cost per contact is the total cost of your support operation divided by the number of contacts it resolves in the same period. It includes salaries and on-costs, software, channel fees such as per-conversation WhatsApp charges, and a fair share of management overhead. Gartner benchmarks the median at $1.84 for self-service and $13.50 for agent-assisted — roughly a 7x gap. The metric is only useful when the denominator is resolved contacts rather than received contacts, because counting received contacts rewards you for conversations that were never actually finished.

Is the $1.84 vs $13.50 gap real?

Yes, for the conversations it describes. Structured, repetitive, low-emotion intents — order status, opening hours, policy questions, password resets, appointment booking — resolve at 55–70% first-contact for AI-native platforms at under $3 per resolution. The gap is not real for complaints, disputes and anything requiring judgement, where AI CSAT falls to 3.34 out of 5 against 4.32–4.41 for structured queries (Zendesk). A business that models its whole operation on $1.84 will be disappointed; one that models its structured intents on it will be roughly right.

How do I calculate the effective cost per resolved conversation?

Take AI cost per resolution multiplied by contained conversations, then add re-contact rate multiplied by contained conversations multiplied by your human cost per contact. At 3,000 AI resolutions costing $1.50 each with an 11.3% re-contact rate handled by humans at $13.50, that is $4,500 of AI cost plus $4,577 of re-contact cost — about $3.03 per originally-resolved conversation rather than $1.50. Still far better than $13.50, but not $1.84, and the difference is what a finance review will find.

What is the payback period on AI customer service?

For most small and mid-sized businesses, weeks rather than months, because the software cost is small relative to the capacity it recovers. A flat tier at $99 a month against 3,000 contained conversations does not need much containment to pay for itself. The longer and more interesting question is whether you redeploy the recovered hours to something that generates revenue. Businesses that count deflection as cash savings while keeping the same headcount and the same workload are marking their own homework.

Should I count deflection as savings in my business case?

Only if you actually reduce or redeploy the hours. Deflection produces recovered staff capacity, not cash. That capacity is genuinely valuable — it is how a three-person team handles the volume of a five-person team through a growth year — but it belongs in a capacity line, not a savings line. Presenting it as cash to a finance director who later checks payroll is the fastest way to lose credibility for the whole programme, including the parts that were working.

What costs do vendors leave out of their ROI calculators?

Four, consistently. Maintenance: two to four hours a week of someone reading conversations and closing knowledge gaps, heaviest in the first quarter, roughly $450–$900 a month at a loaded rate. Channel fees: Meta charges per WhatsApp conversation independently of your subscription, and at broadcast volume this frequently exceeds the software cost. Double metering: some helpdesks charge per ticket and per AI resolution on the same ticket. And re-contact, which is the largest of the four and the one nobody volunteers.

What is double metering and how do I avoid it?

Double metering is being charged twice for one conversation — a per-ticket or per-seat fee for the helpdesk, plus a per-resolution fee for the AI that resolved it. It can roughly triple the effective cost per resolved conversation while every individual line item looks reasonable. Ask one direct question before signing: "if your AI resolves a ticket with no human involved, how many charges does that generate?" A vendor that cannot answer plainly is telling you the answer.

Why does re-contact rate matter more than containment?

Because containment without re-contact is a vanity metric. If one in nine AI-resolved conversations comes back, your true resolution number is materially lower than your dashboard and your effective cost is materially higher. Zendesk benchmarks 11.3% re-contact on AI-resolved against 8.7% human-resolved. Measure yours at 48 hours. If it is above 20%, your containment number is fiction and the cost model built on it is fiction too.

How much does maintenance really cost?

Two to four hours a week of someone competent reading escalations and writing the missing answers, heaviest in the first quarter and never zero afterwards. At a $52 loaded hourly rate that is roughly $450–$900 a month. It is also the highest-return activity in the entire system: deployments that do it climb from 40–50% containment past 60% over six to twelve months, and deployments that skip it sit at the 41% median indefinitely while paying the same subscription.

Does AI reduce headcount or absorb growth?

Overwhelmingly the second, in smaller businesses. The common pattern is not three support people becoming two — it is three support people handling the volume that would otherwise have required five, through a period when hiring was neither affordable nor fast enough. That is a real economic gain, and it is worth stating honestly in a business case rather than promising redundancies that will not happen and would damage the service if they did.

What is the revenue side of the AI support business case?

Almost every ROI calculation counts saved cost and ignores created revenue, which for consumer businesses is usually backwards. If 30% of 5,000 monthly conversations arrive outside working hours and 5% of those convert at a $60 average order value, that is $4,500 a month that previously went to whichever competitor replied first. Add recovered carts and bookings taken in-conversation. For many consumer businesses this number grows faster than the cost saving, because it is new revenue rather than reallocated staff time.

How do WhatsApp conversation fees affect cost per contact?

Meta charges per conversation on WhatsApp independently of what your platform charges, and the category matters enormously: marketing conversations cost the most, utility conversations much less, and service replies to customer-initiated conversations are the cheapest and often free within the service window. At broadcast volume these fees frequently exceed the software subscription. Two actions pay immediately — audit your template categories, and shift volume from outbound broadcast to inbound conversation.

What containment rate should I budget for in year one?

Budget for roughly half your target in month one, and 60%+ by month six if someone is doing the weekly review. Median tier-1 automation across programmes is about 41%, while strong deployments run 65–72%. The gap is not a technology gap — the same models are available to everyone. It is deployment quality: integrations connected, knowledge base purpose-written, escalation log actually read.

Is per-resolution pricing cheaper than a flat subscription?

It depends entirely on volume and on how well the AI performs. Per-resolution pricing looks cheap at low volume and becomes expensive precisely as the AI gets better, because every improvement generates another billable event. Flat tiers cost more at trivial volume and stop scaling with success. The structural problem with per-resolution billing in seasonal businesses is that your busiest sales week is also your largest invoice — the full breakdown is in our pricing guide.

How do I model cost per contact for a business my size?

Start with a real week of conversation counts across every channel, not the ticket system alone — the ticket system usually misses Instagram and WhatsApp entirely. Multiply by your loaded hourly rate and average handling minutes to get today's cost. Then model containment at half your target, add re-contact at 11%, add maintenance at three hours a week, and add channel fees. If it still pays at those conservative inputs, it will pay at real ones.

What is the cheapest way to test the economics before committing?

Run a permanent free tier against your actual questions for two weeks and count what it resolves. This costs an afternoon of setup and produces evidence about your business rather than a benchmark about someone else's. Ask your top twenty real questions, count how many resolve without a person, then check at 48 hours how many of those customers came back. That single test tells you more than any vendor ROI calculator, including ours.

Does lower cost per contact hurt customer satisfaction?

Not automatically, and the data is more encouraging than the debate suggests. A hybrid model at 22% escalation runs roughly 71% lower cost per resolution for a CSAT cost of about 0.05 points — an extremely good trade. But it depends entirely on routing. Push containment up by holding conversations the AI should not handle and CSAT falls sharply, because complaint handling scores 3.34 out of 5 against 4.32–4.41 for structured queries. Cheaper and worse is easy; cheaper and equivalent takes routing discipline.

How does Jugl price against per-resolution vendors?

Jugl publishes four flat tiers — Free, $31, $119 and $390 a month — with the AI included and nothing metered per message, per token or per resolution. The practical difference is not the headline number but the shape: a per-resolution vendor bills you more every time the AI improves and every time your volume spikes, which in seasonal or campaign-driven businesses means your best week generates your worst invoice. Meta conversation fees still apply on WhatsApp regardless of vendor, and any honest comparison should include them on both sides.

13People also ask

The cheapest way to settle this is your own data

Every benchmark on this page describes someone else's business. Yours has a different question mix, a different channel split and a different share of conversations arriving at 11pm — and those three variables decide whether your cost per contact lands nearer $3 or nearer $13.

Point an agent at your real questions for two weeks and count what it resolves. The free tier is permanent and needs no card, so the test costs an afternoon rather than a procurement cycle. Every week you defer it, the conversations arrive anyway — they just get answered by whoever is free, or by nobody.

Flat tiers — nothing metered per message, token or resolutionWhatsApp, Instagram, Messenger, web chat, email and SMSContainment, CSAT and re-contact reported honestlyLive order, booking and CRM lookups inside the conversationFull-context handover to a human, by designPermanent free tier — evidence before procurement

Start free at JuglNo card · live the same day Book a demoWith the co-founders

The conversations you are not answering still cost you. They just do not appear on any invoice.

SOC 2 Type 2 · HIPAA compliant · Meta Business Partner · NVIDIA Inception · 1000+ businesses

Keep reading

AI customer service pricingPer-resolution, per-seat and flat tiers modelled at the same volume.AI customer service statisticsThe sourced numbers behind every claim on this page.AI agent benchmarksWhat good looks like after launch, and what to measure.Measuring agent performanceContainment, CSAT split and re-contact, defined properly.11 AI support mistakesThe eleven fixable errors behind the 41% containment median.AI-to-human handoffThe routing rule that keeps CSAT at 4.3 instead of 3.34.WhatsApp Business statisticsWhere per-conversation channel fees change the whole model.AI chatbots for ShopifyWhy order lookup decides your largest ticket category.Intercom alternativesWhat per-resolution pricing costs at real volume.What is Jugl?Capabilities, fit, pricing, and who should walk away.Jugl pricingFour published flat tiers with the AI included. Free forever, no card.Free conversation auditYour real question mix, counted before you model anything.

Sources: Gartner customer service research (median cost per contact for self-service and agent-assisted; self-service resolution against deflection); Salesforce State of Service (fully burdened agent cost); Zendesk customer experience benchmarks (tier-1 resolution cost bands, re-contact rates on AI-resolved and human-resolved conversations, CSAT by intent type); Aissist.io programme analysis (median and strong-deployment automation rates); Lorikeet (first-contact resolution for AI-native platforms); published benchmark analysis of hybrid escalation models; and Jugl's own published pricing and customer-reported deployment data. Figures described as derived are calculated from published bands rather than quoted directly. This page is published by Jugl, which sells an AI customer agent platform and is therefore an interested party; Jugl's outcome figures are customer-reported and typical rather than guaranteed. Calculator outputs are illustrative estimates generated from your own inputs, not quotes, forecasts or guarantees, and nothing here is financial advice. Meta, WhatsApp, Messenger, Instagram and Facebook are trademarks of Meta Platforms, Inc.; Jugl is a Meta Business Partner and this page is published by Jugl and is not endorsed by or affiliated with Meta Platforms, Inc. All other product names are trademarks of their respective owners.

Start free at Jugl · No card required · Permanent free tier