What Margin Can an Agency Actually Make Reselling AI

The honest math on services margins when delivery runs on a platform instead of on bodies.

Technology
By Mark Choudhari · Jun 7, 2026 · 6 min read

Thirteen percent is the average. The number you keep is the one to protect.
Made with Works

TL;DR

The average digital agency nets about 13% after tax, and margin falls as you grow because growth adds people, not leverage. Reselling AI by the hour drifts to the bottom of that range. A productized practice run on a shared operations layer can target 50 to 60% delivery margin, because the platform decouples revenue from headcount and the agency charges for expertise and results, not the eroding hour.

In this article

What margin can an agency actually make reselling AI

The advertised number and the kept number are not the same. White-label reseller margins are quoted at low 20 to 40% under price competition, 40 to 60% when an agency bundles services, and 60 to 80% or more on niche value-added work. Those are gross spreads before delivery cost. The audited net is far thinner.

The average digital agency netted about 13% after tax in 2025, down from 14% the prior year, and the figure falls as agencies scale: studios under 10 people clear roughly 19%, while shops past 50 FTE land around 8%. Reselling AI by the hour pulls you toward the bottom of that range, because the hour is exactly what AI is compressing.

The average digital agency netted just 13% after tax in 2025, and net margin falls as agencies scale, from about 19% at the smallest studios to about 8% past 50 people.
Promethean Research, 2026

What gross margin should an AI agency target after model, tooling, and support costs

A defensible target for an AI services line is 50 to 60% gross margin after technical salaries and AI tool and API costs, with 10 to 20% of revenue ring-fenced for R&D, and subscription or managed-service pricing carrying a higher net than project work. The cost trap specific to AI is variable usage: never absorb metered API cost inside a fixed fee without a buffer, or a heavy-use client quietly converts your margin into someone else’s compute bill.

The healthy cost shape underneath the target is roughly 55% delivery, 25% overhead, 20% profit, and the metric that drives it is utilization in the 65 to 80% sweet spot, with revenue per FTE around the $172k benchmark. Below those lines the model is leaking before any markup is added.

What margin can I hold after support, account management, and per-client maintenance

This is the question the reseller pitch skips. The spread you quote is gross. What survives is net, after the support tickets, the account management, and the per-client maintenance that grows with every logo you add. That maintenance is its own cost with its own dynamics, and it is the single biggest threat to the margin claim on this page. We cover the agency AI tax, the per-client cost that erodes the spread, in the agency AI tax breakdown.

The fundamental fix is to stop letting each new client add a proportional chunk of human maintenance. When the repeatable execution lives on a shared layer, the next account is mostly configuration, so the maintenance line grows far slower than the revenue line. That gap is the margin.

Markup vs margin, which number should I actually be protecting

Markup is what you add. Margin is what you keep. Agencies that chase markup can post eye-catching figures, including reported reseller margins above 100% on packaged presence products, with a guideline of 100 to 200% on core DIY products. But a 100% margin on small revenue can be worth less in real dollars than a 40% margin on large revenue, so protecting the percentage alone is the wrong instinct.

Protect delivery margin, not the eroding hour. Delivery margin is the percentage of fee left after the cost of delivering the work, and it is the number that compounds into a real business. A delivery-margin-first read of the P&L targets 50 to 60% on delivery and 60 to 70% on the project itself, with overhead at 20 to 30% of adjusted gross income.

Target 50 to 60% delivery margin on the P&L and 60 to 70% per project; the gap between project margin and net margin is overhead, utilization, and service mix.
Parakeeto, 2026

Should I charge a markup, a management retainer, or both

A pure markup ties your revenue to a tool price you do not control, and AI tool prices are falling. A management retainer ties revenue to the outcome you manage, which is the part clients actually value and the part that holds margin. The practical answer for most agencies is both, weighted toward the retainer: a thin, honest markup on the platform seat, plus a managed-service or productized-services retainer for the expertise, configuration, and results around it.

Pricing model What it ties revenue to Margin behavior
Hourly Time, which AI is compressing Erodes as delivery speeds up
Tool markup A price you do not control Thin, falls with tool prices
Management retainer The result you manage Holds, recurs, compounds
Productized service A fixed scope you own Highest, scales with leverage

Retainer-heavy agencies already report roughly 8 percentage points more net margin than project-based ones, and niche specialists running this model land in the 40 to 75% range.

How do I price this without racing competitors to the bottom

The race to the bottom is the per-hour reseller’s outcome, not a law of the category. You avoid it by not selling the commodity. Specialize, because specialist agencies clear 25 to 40% net against 15 to 20% for generalists. Productize, so the buyer compares your result against their problem, not your rate against the next vendor’s rate. And run the operation tightly, because cutting overhead from 30% to 25% of revenue lifts profit by roughly 25% on its own.

Specialist agencies run 25 to 40% net margin versus 15 to 20% for generalists, and utilization in the 65 to 80% band is the profitability sweet spot.
TMetric, 2025

Where the margin math changes

Every number above points at one bar any real answer has to clear: break the link between growth and headcount, because that link is what caps the services model at 13%. The agencies that hold a better margin do it by building leverage into delivery, and a shared operations layer is how you build it without hiring for it.

That is what JynAI built Works to be for an agency practice. The agency runs client delivery on the platform, and the repeatable execution stops being a person.

  • The plays come built. 500-plus workflows on the methodologies an agency already runs, EOS, MEDDIC, ABM, PLG, so a new account starts from a configured practice, not a blank page or a new hire.
  • Existing automations carry over. The Make and n8n automations an agency already built import in rather than getting rebuilt, so the setup cost per client drops.
  • The work actually ships across the client’s stack. 3,000-plus apps reachable, with the agency setting how much runs hands-off, so delivery scales without proportional labor.
  • The result is provable. Every run and outcome is logged and exports to a client-ready report, which is what justifies a retainer instead of a markup.

The price proof is the part that makes the margin honest. The cost base the agency’s fee sits on is a per-seat subscription, with the Pro tier at $49 per seat, not a salary. A markup on a seat is a different business from a markup on a hire, and that is the whole point of running delivery on a platform.

This sits alongside the rest of the agency picture: the white-label and partner economics behind the spread, the fractional multiplier that raises per-account margin, and the agency AI tax that erodes it. For the full pillar, see AI for agencies.

Build a Works practice. Get early access, or ask us for the agency economics breakdown.

Common Questions

What margin can an agency actually make reselling AI?

Reseller gross spreads are quoted at 20 to 80 percent, but the number that matters is the audited net: about 13 percent for the average digital agency, falling to roughly 8 percent past 50 staff. A productized practice on a platform can hold 50 to 60 percent delivery margin by keeping repeatable execution off human headcount, so the 13 percent figure is a floor for hour-sellers, not a ceiling for the model. Full breakdown in the margin section above.

Is reselling AI a thin-margin race to the bottom?

For commodity hour-sellers, yes. For specialists and productized-service agencies, no. The difference is what you are pricing: an hourly rate compresses as AI speeds up delivery, while a retainer prices the outcome and the expertise, which do not compress. Niche specialists who productize hold 40 to 75 percent net against the 13 percent average. More in pricing without the race.

What margin can I hold after per-client maintenance?

Less than the gross spread, because support and per-client maintenance grow with every logo. The fix is keeping maintenance off human headcount; the cost side is covered in the agency AI tax.

Markup or retainer?

Both, weighted to the retainer. A thin markup on the platform seat plus a managed-service retainer for expertise and results, because the retainer is the part that holds margin and recurs.

Does running delivery on a platform actually change the margin?

Yes, by breaking the growth-equals-headcount link. When repeatable execution lives on a shared layer, the next account is mostly configuration, so revenue grows faster than cost. That gap is the margin.

Get Started With AI

Are You Ready to Make AI Work for You?

Simplify your AI journey with solutions that integrate seamlessly, empower your teams, and deliver real results. Jyn turns complexity into a clear path to success.

See AI for Real Business Impact in Action →

ai that powers your team 226d8ee5db

What Margin Can an Agency Actually Make Reselling AI

The honest math on services margins when delivery runs on a platform instead of on bodies.

Technology
By Mark Choudhari · Jun 7, 2026 · 6 min read

Thirteen percent is the average. The number you keep is the one to protect.
Made with Works

TL;DR

The average digital agency nets about 13% after tax, and margin falls as you grow because growth adds people, not leverage. Reselling AI by the hour drifts to the bottom of that range. A productized practice run on a shared operations layer can target 50 to 60% delivery margin, because the platform decouples revenue from headcount and the agency charges for expertise and results, not the eroding hour.

In this article

What margin can an agency actually make reselling AI

The advertised number and the kept number are not the same. White-label reseller margins are quoted at low 20 to 40% under price competition, 40 to 60% when an agency bundles services, and 60 to 80% or more on niche value-added work. Those are gross spreads before delivery cost. The audited net is far thinner.

The average digital agency netted about 13% after tax in 2025, down from 14% the prior year, and the figure falls as agencies scale: studios under 10 people clear roughly 19%, while shops past 50 FTE land around 8%. Reselling AI by the hour pulls you toward the bottom of that range, because the hour is exactly what AI is compressing.

The average digital agency netted just 13% after tax in 2025, and net margin falls as agencies scale, from about 19% at the smallest studios to about 8% past 50 people.
Promethean Research, 2026

What gross margin should an AI agency target after model, tooling, and support costs

A defensible target for an AI services line is 50 to 60% gross margin after technical salaries and AI tool and API costs, with 10 to 20% of revenue ring-fenced for R&D, and subscription or managed-service pricing carrying a higher net than project work. The cost trap specific to AI is variable usage: never absorb metered API cost inside a fixed fee without a buffer, or a heavy-use client quietly converts your margin into someone else’s compute bill.

The healthy cost shape underneath the target is roughly 55% delivery, 25% overhead, 20% profit, and the metric that drives it is utilization in the 65 to 80% sweet spot, with revenue per FTE around the $172k benchmark. Below those lines the model is leaking before any markup is added.

What margin can I hold after support, account management, and per-client maintenance

This is the question the reseller pitch skips. The spread you quote is gross. What survives is net, after the support tickets, the account management, and the per-client maintenance that grows with every logo you add. That maintenance is its own cost with its own dynamics, and it is the single biggest threat to the margin claim on this page. We cover the agency AI tax, the per-client cost that erodes the spread, in the agency AI tax breakdown.

The fundamental fix is to stop letting each new client add a proportional chunk of human maintenance. When the repeatable execution lives on a shared layer, the next account is mostly configuration, so the maintenance line grows far slower than the revenue line. That gap is the margin.

Markup vs margin, which number should I actually be protecting

Markup is what you add. Margin is what you keep. Agencies that chase markup can post eye-catching figures, including reported reseller margins above 100% on packaged presence products, with a guideline of 100 to 200% on core DIY products. But a 100% margin on small revenue can be worth less in real dollars than a 40% margin on large revenue, so protecting the percentage alone is the wrong instinct.

Protect delivery margin, not the eroding hour. Delivery margin is the percentage of fee left after the cost of delivering the work, and it is the number that compounds into a real business. A delivery-margin-first read of the P&L targets 50 to 60% on delivery and 60 to 70% on the project itself, with overhead at 20 to 30% of adjusted gross income.

Target 50 to 60% delivery margin on the P&L and 60 to 70% per project; the gap between project margin and net margin is overhead, utilization, and service mix.
Parakeeto, 2026

Should I charge a markup, a management retainer, or both

A pure markup ties your revenue to a tool price you do not control, and AI tool prices are falling. A management retainer ties revenue to the outcome you manage, which is the part clients actually value and the part that holds margin. The practical answer for most agencies is both, weighted toward the retainer: a thin, honest markup on the platform seat, plus a managed-service or productized-services retainer for the expertise, configuration, and results around it.

Pricing model What it ties revenue to Margin behavior
Hourly Time, which AI is compressing Erodes as delivery speeds up
Tool markup A price you do not control Thin, falls with tool prices
Management retainer The result you manage Holds, recurs, compounds
Productized service A fixed scope you own Highest, scales with leverage

Retainer-heavy agencies already report roughly 8 percentage points more net margin than project-based ones, and niche specialists running this model land in the 40 to 75% range.

How do I price this without racing competitors to the bottom

The race to the bottom is the per-hour reseller’s outcome, not a law of the category. You avoid it by not selling the commodity. Specialize, because specialist agencies clear 25 to 40% net against 15 to 20% for generalists. Productize, so the buyer compares your result against their problem, not your rate against the next vendor’s rate. And run the operation tightly, because cutting overhead from 30% to 25% of revenue lifts profit by roughly 25% on its own.

Specialist agencies run 25 to 40% net margin versus 15 to 20% for generalists, and utilization in the 65 to 80% band is the profitability sweet spot.
TMetric, 2025

Where the margin math changes

Every number above points at one bar any real answer has to clear: break the link between growth and headcount, because that link is what caps the services model at 13%. The agencies that hold a better margin do it by building leverage into delivery, and a shared operations layer is how you build it without hiring for it.

That is what JynAI built Works to be for an agency practice. The agency runs client delivery on the platform, and the repeatable execution stops being a person.

  • The plays come built. 500-plus workflows on the methodologies an agency already runs, EOS, MEDDIC, ABM, PLG, so a new account starts from a configured practice, not a blank page or a new hire.
  • Existing automations carry over. The Make and n8n automations an agency already built import in rather than getting rebuilt, so the setup cost per client drops.
  • The work actually ships across the client’s stack. 3,000-plus apps reachable, with the agency setting how much runs hands-off, so delivery scales without proportional labor.
  • The result is provable. Every run and outcome is logged and exports to a client-ready report, which is what justifies a retainer instead of a markup.

The price proof is the part that makes the margin honest. The cost base the agency’s fee sits on is a per-seat subscription, with the Pro tier at $49 per seat, not a salary. A markup on a seat is a different business from a markup on a hire, and that is the whole point of running delivery on a platform.

This sits alongside the rest of the agency picture: the white-label and partner economics behind the spread, the fractional multiplier that raises per-account margin, and the agency AI tax that erodes it. For the full pillar, see AI for agencies.

Build a Works practice. Get early access, or ask us for the agency economics breakdown.

Common Questions

What margin can an agency actually make reselling AI?

Reseller gross spreads are quoted at 20 to 80 percent, but the number that matters is the audited net: about 13 percent for the average digital agency, falling to roughly 8 percent past 50 staff. A productized practice on a platform can hold 50 to 60 percent delivery margin by keeping repeatable execution off human headcount, so the 13 percent figure is a floor for hour-sellers, not a ceiling for the model. Full breakdown in the margin section above.

Is reselling AI a thin-margin race to the bottom?

For commodity hour-sellers, yes. For specialists and productized-service agencies, no. The difference is what you are pricing: an hourly rate compresses as AI speeds up delivery, while a retainer prices the outcome and the expertise, which do not compress. Niche specialists who productize hold 40 to 75 percent net against the 13 percent average. More in pricing without the race.

What margin can I hold after per-client maintenance?

Less than the gross spread, because support and per-client maintenance grow with every logo. The fix is keeping maintenance off human headcount; the cost side is covered in the agency AI tax.

Markup or retainer?

Both, weighted to the retainer. A thin markup on the platform seat plus a managed-service retainer for expertise and results, because the retainer is the part that holds margin and recurs.

Does running delivery on a platform actually change the margin?

Yes, by breaking the growth-equals-headcount link. When repeatable execution lives on a shared layer, the next account is mostly configuration, so revenue grows faster than cost. That gap is the margin.

Get Started With AI

Are You Ready to Make AI Work for You?

Simplify your AI journey with solutions that integrate seamlessly, empower your teams, and deliver real results. Jyn turns complexity into a clear path to success.

See AI for Real Business Impact in Action →

ai that powers your team 226d8ee5db