By the Onsites AI team · Last updated · 5-minute read
Strip the feature grids and every help desk on the market bills one of three ways. Per seat: a monthly rate per licensed human, usually inside tiers that bundle features. Per resolution: a fee each time a conversation closes — most visibly when AI closes it. The flat hybrid: flat seats plus a usage line priced as inputs (prepaid AI credits at published rates) and storage by the gigabyte — Onsites' shape. The choice matters more than any demo, because the meter compounds for years under the team, and each model rewards a different shape of company: per seat favors stable desks that use bundles deeply; per resolution favors tiny or spiky-lite desks with low AI share; the flat meter favors growth and transparency. This guide walks all three models at three desk sizes — 3 seats, 10 seats, 25 seats — with worked numbers and the failure modes of each, so the decision can made once, correctly, and then left alone for years.
The industry default. Mechanics: every licensed human pays a monthly rate; the rate is set by the tier the team's feature list forces it into. Behavior at scale: linear in headcount, steppy in features — and the steps, not the slope, are where budgets break. A ten-person desk on a small-team tier pays modestly; the day it needs advanced routing or an SLA engine it jumps a whole tier for everyone. Who it fits: static-desk organizations that will genuinely use most of a bundle's features, already inside enterprise agreements, or with hiring curves so flat the seat meter never surprises. Where it strains: growth (every hire reprices at the tier rate), occasional participants (warehouse, finance, brokers — license them fully or ration them), and the AI era, where the copilot lives a tier or add-on above the seat the desk already pays. The small-team comparisons document the tier tax desk by desk.
The AI-era meter. Mechanics: pay each time a conversation resolves — published AI rates around $0.99/resolution, often inside monthly packs. Behavior: the line rises with volume and with AI adoption simultaneously, so success bills twice. At 4,000 conversations and 20% AI share, expect ≈$794/mo before seats; the seasonal peak reprices worse. Who it fits: small desks (200 conversations, 10% share ≈ $20/mo beats any seat price) and teams that genuinely prefer $0 invoices in dead months. Where it strains: any desk whose volume curve has a peak (most desks), and any desk with an AI-adoption goal — because the meter taxes the exact KPI leadership is asking it to grow. The dedicated guide works the full math; the shape to remember is that outcome meters budget badly precisely when business is good.
Mechanics: seats flat at $15/mo beyond three free (prorated daily, billed annually); storage 100 MB included, $1/GB beyond; AI as prepaid credits — draft 3, suggested reply 6, translate 3, summary 6, roughly a cent each, packs that never expire. Behavior: the seat line follows headcount, the credit line follows usage, nothing follows outcomes; a flood month costs credits at cent-rates, and a dead month costs almost nothing without vanishing prepaid packs. Who it fits: growing teams (the meter stays legible from hire 4 to hire 40), multi-channel desks (all seven channels native), teams wanting the commercial loop in the same tool (quotes → orders → invoices in the thread), and anyone who wants finance to pre-approve the whole year in one line each. Where it strains: very low-volume desks (three free seats already cover it, so the strain is nil) and enterprise procurement built around per-seat negotiations with the majors. The free-tier guide covers the first-year shape.
Same workload family, three sizes, both meters. 3 seats, 500 conversations, light AI: flat meter — all seats free, credits ≈$15/mo, storage $0 → ≈$15/mo. Per-seat suite — three licenses at small-team rates → ≈$180–$300/mo. 10 seats, 4,000 conversations, 20% AI share: flat meter — $105 seats + ≈$100 credits + $0 storage → ≈$255/mo (the derivation lives in the ten-person math). Seat-plus-resolution stack: ≈$1,200 seats + ≈$794 AI + tier add-ons → ≈$2,300–$3,100/mo. 25 seats, 9,000 conversations, 30% AI share: flat meter — $330 seats + ≈$330 credits → ≈$700/mo; resolution-metered stack: ≈$6,500–$8,000/mo, and the gap widens with every step because volume multiplies the outcome line while seats merely add. The pattern holds at every size: the flat meter's lines are few, published, and scale gently; the metered lines are many, and at least one compounds with success.
Because meters compound, extend every comparison to the decade the tool will actually serve. Per seat: the decade's pattern is a staircase — tiers step up, headcount steps up, and each step retro-prices the whole team at renewal; a desk that grows 5→25 people over ten years will have crossed three or four tier boundaries, each repriced at everyone's next-monthly-rate. Per resolution: the decade's pattern tracks your volume curve exactly — cheap years when the business is slow, punishing ones when it is hot, and at year five or six the AI line typically becomes the desk's largest single cost. The flat hybrid: the decade's pattern is two gentle slopes — hire twenty more people and the bill adds $300/mo; double your AI usage and the credits line doubles from a small number to a small number. The honest question for the decade is not "which month was cheapest?" but "which meter's worst year could you have predicted in year one?" The staircase surprises, the outcome meter spikes, the published cents persist. Price the decade, sign the meter, and revisit only when your curve's shape changes.
Four questions separate the models for any real desk. One — what is your curve? Flat volume favors per-seat bundles; spikes punish per-resolution meters and ride cheaply on credits. Two — are you hiring? Growth multiplies seat bills and tips tiers; the flat meter's $15 stays $15. Three — does AI have a target share? A rising AI goal raises resolution meters and lowers (yes, lowers) credit bills per resolution achieved. Four — does support carry commerce? Quotes, orders, invoices in the thread beat the suite-and-integration stack at any price, and are native on exactly one of the three models. Run the numbers at your real curve in the calculator, and re-audit annually — models drift, and the meter signed in good faith at five seats will still be compounding at fifty. The model that publishes its rate card and bills inputs is the one whose decade you can forecast.
What are the main help desk pricing models?
Three: per-seat (a monthly rate per person, inside feature tiers), per-resolution (a fee each time a conversation closes — around $0.99 for AI), and the flat hybrid (flat seats + storage + prepaid AI credits at published cents). Nearly every product on the market is one of these shapes.
Which pricing model is cheapest at scale?
For growing teams with volume, the flat hybrid: 10 seats ≈$255/mo and 25 seats ≈$700/mo in the worked examples, versus $3,100–$8,000 on seat-plus-resolution stacks. Per-seat wins only for static desks deep inside negotiated bundles; per-resolution wins only at tiny volumes.
What does a 10-person team pay per model?
Worked example at 4,000 conversations, 20% AI share: flat meter ≈$255/mo (seven paid seats = $105, ~$100 credits); per-seat suite ≈$1,200–1,500; adding the resolution meter and tiers lands ≈$2,300–$3,100.
Does per-resolution pricing punish success?
Structurally yes: the line rises with conversation volume and with AI adoption — both are measures of the desk succeeding, not choices it makes. Input meters (credits) move with usage at about a cent per assist instead.
When is a per-seat suite the right choice?
Static headcounts that genuinely use most of a bundle's features, enterprise agreements already in place, or organizations where the workflow engine's depth beats meter economics. Re-audit annually — the steps are where bills jump.