For most software, AI cuts your cost and you keep the difference. For a BPO, AI threatens the line item your whole contract sits on. If you bill clients per agent or per hour and an AI agent now does the work of several people, you have quietly priced yourself toward selling less of your own core product. That is the puzzle every outsourcer and reseller is working through in 2026.
The way out runs through the pricing. Understand how AI support is priced underneath you, then choose a resale model that survives the moment automation actually works. This piece walks through the economics from the reseller's side: how vendors price the layer you would resell, how that collides with per-agent contracts, and where the durable margin lives.
It sits under our hub on how AI customer service pricing works, which covers the buyer-side models. This is the same map read from the supply side.
What you are actually reselling, and what it costs you
Traditional outsourcing prices labor. AI support prices outcomes, and the gap between those two units is the whole strategic problem.
The labor side is well understood. Outsourced support runs roughly $8 to $15 an hour offshore, $20 to $30 nearshore, and $40 to $60 onshore, or about $1,200 to $4,000 per agent per month, often quoted to clients as a per-FTE or per-hour rate. Your margin is the spread between what you pay an agent and what you bill for them.
The AI layer you would resell is priced per resolution: roughly $0.50 to $2.00 per resolved ticket across the major vendors in 2026. Intercom's Fin runs $0.99 per outcome and Zendesk $1.50 for committed volume up to $2.00 pay-as-you-go, while HubSpot moved its Customer Agent to $0.50 per resolved conversation. Your cost of goods is no longer an hourly wage. It is a per-event fee, and your margin is the spread between the vendor's resolution rate and what you charge your client for the same resolved ticket.
Those two cost structures do not blend cleanly, which is exactly why the resale model you choose matters more than the rate you negotiate.
Two cost units, one resale spread: labor vs. resolution (2026)
Outsourced labor rates from text.com; AI per-resolution rates from each vendor’s own pricing page or announcement. Cost-of-goods, not sell price.
Your margin = sell rate − buy rate, in whichever unit you bill.
The margin trap in per-agent contracts
If your client contract bills per agent or per hour, AI puts you in a bind that better automation only tightens.
Suppose you staff a client with ten agents and bill per seat. You deploy AI, it resolves 60% of tickets, and now you need four agents instead of ten. Under a per-agent contract you have three options, and two of them are bad. You can keep billing for ten agents you no longer need, which works until the client notices. You can drop to four agents and watch your contract value fall by more than half. Or you can renegotiate the basis of the deal before the client does it for you.
The structural issue is that your revenue is tied to headcount while your delivery is moving to outcomes. The better your AI performs, the faster a per-agent contract erodes. This is the same misalignment buyers face with per-seat AI tools, described in the pricing pillar: the price stops describing the value when one agent's output stops describing the work.
The resale models that actually survive automation
There are three ways to price AI support to a client. Only some of them hold up once the AI is doing real work.
Per-agent, AI bundled in. You keep billing per seat and quietly use AI to need fewer of them, pocketing the labor you no longer pay for. It protects margin briefly and erodes trust permanently, because the client eventually asks why they are paying for ten people when four are working. It also caps your upside: you can never bill for more value than headcount implies.
Per-resolution markup. You buy resolutions from the vendor and resell them with a margin. Your cost and revenue now use the same unit, so automation works in your favor instead of against you: more resolutions at a stable spread means more margin. The discipline it demands is real, because your cost is now variable and a client spike flows straight through to your vendor bill unless your contract passes it along.
Outcome-based managed service. You bill the client for a result, resolution rate, CSAT, hours saved, and own the mix of AI and humans behind it. This is the highest-margin model and the highest-skill one, because you are now pricing your own judgment about how to hit the outcome rather than the inputs you use to get there.
The trend in the wider category favors the outcome end of that spectrum. Bessemer pegs AI product gross margins at 50 to 60 percent against 80 to 90 percent for classic software because compute is a real variable cost, which means the resellers who win are the ones who price for value rather than reselling raw resolutions at a thin markup.
Why AI-era margins reward the outcome end
Gross-margin economics from Bessemer’s AI Pricing & Monetization Playbook; resolution rates from vendor pricing pages.
AI product gross margin (Bessemer)
Classic software gross margin (Bessemer)
Per-resolution cost of goods, 2026
The reseller-specific terms to nail down
Reselling adds clauses a direct buyer never thinks about. Get these settled before you sign with any underlying vendor.
- Wholesale rate and volume tiers. Your buy rate across your whole book rather than per client, since aggregated volume is your main pricing advantage as a reseller.
- Pass-through on spikes. Whether a client's volume spike hits your margin or theirs. Without a pass-through clause, an outage at one client erodes the spread on all of them. This is the reseller version of the surprise-invoice problem of keeping AI support costs predictable.
- The escalation rule. Whether human handoffs are billable changes your blended cost of delivery, since your agents are handling exactly those escalated tickets. A vendor that does not charge for handoffs, as Open.cx does not, keeps your delivery cost legible when the AI passes work back to your team.
- White-label and data terms. Whether you can present the AI as your own service, and how client data is handled across the stack, which your clients will ask about directly. The data-handling story matters enough to settle with your vendor up front.
The math, from your side of the table
The reseller calculation has one extra layer the buyer's does not: you are pricing a spread between your buy rate and your sell rate, where a buyer prices only a single cost.
Start with the resolutions a client generates, times a realistic automation rate, times your wholesale rate. That is your cost of goods. Then set your client price by the value you deliver, the labor you replace or the result you guarantee, rather than by marking up the resolution rate a fixed percentage. The spread between the two is your margin, and it widens as automation rises only if your client price is tied to value rather than to headcount.
Run the same exercise against your current per-agent economics. The honest comparison is not "AI resolution rate versus agent hourly wage." It is "my margin per client under a per-agent contract versus my margin per client under an outcome contract, at the automation rate I can actually deliver." Our team size calculator and ROI calculator help model the headcount and margin shift; for the per-hour comparison specifically, the AI vs human call center cost breakdown lays out the labor side in detail.
The BPOs that come out ahead in 2026 are the ones that sell resolved outcomes rather than hours, because that is the unit AI is denominated in. The ones that cling to per-agent billing are betting their margin against the steady improvement of the very technology they are deploying, which is not a bet that ages well.