What's Actually in a $15K AI Marketing Agency Retainer (2026 SOW Breakdown)

TLDR: AI marketing agency retainer scope, the 2026 SOW transparency frame

Founders on AI marketing agency intro calls keep asking the same question: "so what do I actually get for $15K?" Most agencies cannot answer cleanly because their internal SOW is hours-and-FTEs while their pitch is outcomes-and-vibes. The gap creates buyer-side risk: founders sign a retainer they cannot independently audit, and the agency burns goodwill the first time the bill comes in.

This post publishes the FORKOFF default retainer SOW verbatim. Eight line items, named owners, pre-priced add-on triggers, outcome anchors per cycle. Numbers come from the FORKOFF AI Agency Engagement Ledger 2026: n=23 active retainers across AI/SaaS, fintech, web3, dev tools, and healthcare verticals. The median retainer is $14,800/mo. The point of this breakdown is not to pitch FORKOFF. The point is to give every founder running an AI agency intro call a defensible scope grid to compare any proposal against, including FORKOFF's.

What an AI marketing agency retainer scope actually means in 2026

A retainer scope of work is the contractual definition of what the agency owes you per billing cycle. In legacy agencies, the unit was hours: 80 hours per month at $200/hour gets you $16,000 of "agency time." That model collapsed once AI-drafted content and operator-orchestrated tooling made hours a misleading proxy for output.

The 2026 AI marketing agency retainer scope is built on deliverables, not hours. A deliverable is a countable, named, attributable shipped asset: a podcast appearance, a published blog post, a launched X campaign, a closed pipeline source. The agency commits to ship N deliverables per cycle, each tied to one or more outcome KPIs (reply rate, qualified inbound, pipeline source attribution). Hours become an internal capacity-planning concern for the agency; they stop being the operator's billing language.

The shift matters because it changes the incentive structure. Under the hours model, the agency is rewarded for slow execution: more hours = more billing. Under the deliverable model, the agency is rewarded for ship speed + asset quality: faster correct delivery = better renewal odds. The AI agency pricing unit economics breakdown covers the WHY of this shift in detail; this post covers the WHAT.

The 8 line items in a defensible AI marketing agency retainer SOW

The FORKOFF default retainer ships 8 line items per cycle. Every item has a named owner, a numerical commit, a pre-priced add-on trigger, and an outcome anchor. The full grid below; the prose walkthrough follows.

Line item 1: founder-voice content assets

Per cycle: 4 to 6 assets. Owner: agency drafts, founder approves. Add-on trigger: 7th+ asset = $1,200 per additional asset. Outcome anchor: 1.6x reply-rate uplift on outbound sequences citing the asset.

A founder-voice asset is a long-form thinking piece (X thread, LinkedIn long-form, blog post, podcast clip) that ships with the founder's name and POV. The agency drafts in the founder's voice based on a recorded prep call; the founder approves with light edits. Token cost averages $30-$60 per asset; verification time (the founder's read-and-edit) averages 25-40 minutes per asset.

The reason 4-6 is the right number, not 10+, is calibration: in the FORKOFF Outbound Ledger 2026 (n=10,847 sequences), the founder-voice asset reply-rate uplift saturates around 5 assets per month. Adding asset 6, 7, 8 doesn't compound; it dilutes the founder's voice across too many surfaces. Operators who want more volume should re-allocate cycle budget to distribution, not asset production.

Line item 2: distribution campaigns

Per cycle: 2 to 4 campaigns. Owner: agency runs, operator approves channel mix. Add-on trigger: 5th+ campaign = $2,400/campaign. Outcome anchor: 3.2x qualified inbound versus founder-solo cadence.

A distribution campaign is the orchestration layer that takes the founder-voice asset and ships it through 3-7 channels (X, LinkedIn, podcast guesting, Reddit, newsletter swaps, conference speaker placements, KOL co-signs). One campaign typically wraps around one asset: the asset is the seed, the campaign is the bloom.

The agency owns campaign execution. The operator approves which channels fit which assets. The 3.2x lift number is grounded against a control: in the FORKOFF cohort, founders running founder-voice content WITHOUT agency-side distribution ship the asset and stop. Pipeline from those assets reaches 34% of the agency-distributed equivalent at 90-day attribution windows, validated through the FORKOFF Founder-Funnel Cohort.

Line item 3: weekly sync (60 minutes)

Per cycle: 4 syncs. Owner: agency lead, attended by operator + founder. Add-on trigger: $0 included up to 6 syncs/mo (allows for vacation drift). Outcome anchor: cadence-velocity index.

The weekly sync is the operating heartbeat of the retainer. Two purposes: (1) review the prior week's shipped assets + campaign data, (2) approve the upcoming week's drafts. Sixty minutes is the right length: longer drifts into status-theater, shorter doesn't surface the decisions that matter.

Format is fixed: 10 minutes shipped-week recap, 25 minutes drafts review and approval, 20 minutes upcoming-week plan, 5 minutes blockers. The agency owns the agenda. Founders who cannot consistently attend the sync need to assign a delegate; absent founder-voice means the agency starts writing in a generic voice and the entire pipeline degrades.

Line item 4: monthly attribution report

Per cycle: 1 report. Owner: agency owns the dashboard + the prose narrative. Add-on trigger: custom-cut analysis = $800. Outcome anchor: qualified-pipeline source-of-truth.

The attribution report is what separates a real retainer from a vibes retainer. It lists every qualified pipeline source over the prior 30 days, attributes each to a specific asset or campaign (or marks it untrackable + explains why), and shows the trend against the prior 90 days. It also shows where the attribution model is uncertain (touchpoint X happened, pipeline source Y closed, can we draw the line cleanly).

The report is not a vanity dashboard. It is the document the founder takes to the board to justify continued spend. In the FORKOFF cohort, founders who actively use the monthly attribution report in their internal reporting renew at 89% vs the cohort baseline of 71%, because the report converts agency spend from an operating-expense line into a measured pipeline asset.

Line item 5: outcome gate per cycle

Per cycle: 1 gate. Owner: joint operator + agency review. Add-on trigger: gate-miss triggers renegotiation, not an extra-charge. Outcome anchor: releases the next cycle.

This is the line item that makes the retainer outcome-priced rather than time-priced. Each cycle has one named outcome gate, defined in the SOW at signature. The gate is operator-specific: for an early-stage AI startup it might be "10 qualified inbound conversations from named-account targets this cycle"; for a Series A SaaS it might be "1 podcast appearance on a tier-1 show + 2 spoke posts ranking on page 1 for the target KW."

If the gate is hit, the next cycle releases at the same price. If the gate is missed by a small margin (say, 80% of target), the agency typically eats the gap and ships an extra deliverable next cycle. If the gate is missed by a large margin (under 50%), the contract is renegotiated, scope or price adjusts, both parties pause and re-align. The mechanism is not punitive; it is the mechanism that prevents the retainer from drifting into hours-billed undefined-output theater.

Line item 6: model-token pass-through

Per cycle: $180-$420 actual cost on a $15K retainer (1-3% of revenue). Owner: agency procures, operator sees the invoice. Add-on trigger: token spike above 150% of monthly average triggers a scope review. Outcome anchor: cost-per-output trend over time.

Model-token costs (Claude, GPT, Gemini, internal fine-tunes used for drafting + research) are passed through as a separate invoice line. Bundling them into the retainer creates a hidden margin and incentivizes the agency to under-use models. Passing them through with a small handling fee (5-10%) keeps the agency honest and gives the operator visibility into how AI compute scales with output volume.

If token costs spike above 150% of the monthly average, both parties review the cycle: it usually signals either a scope expansion (new campaign types) or an inefficiency (an agency operator over-prompting due to under-trained workflow). The cost transparency is a defense against the "AI is expensive, we billed you" pattern that some early AI agencies pulled in 2024-2025.

Line item 7: crisis or incident response

Per cycle: on call, not retainer. Owner: agency lead, founder approves resolution path. Add-on trigger: flat $4,000-$12,000 per incident depending on severity. Outcome anchor: resolution + RCA + recovery plan.

Crisis response is explicitly out of the standard retainer scope. Reason: incidents are unpredictable, high-stakes, and require a different staffing posture than steady-state content production. A model-failure trust crisis, a competitor leak, a product-launch backlash, a regulatory call-out, an HR-side public incident, these are not retainer-cycle events.

The agency commits to be on-call (response within 4 hours during business days, 12 hours otherwise) but bills incidents as flat-fee engagements. Severity tiers: minor incident $4,000 (single-channel containment, 48-hour resolution); major incident $8,000 (multi-channel coordination, 72-hour resolution); structural incident $12,000+ (long-running trust recovery, 1-2 week engagement). The AI agent blast radius marketing playbook covers the upstream containment patterns that limit how often incidents fire in the first place.

Line item 8: quarterly outcome review

Per cycle: 0 within-cycle, 1 quarterly. Owner: agency presents, operator gives the renewal scorecard. Add-on trigger: a negative quarterly review makes the next renewal contingent on a scope renegotiation. Outcome anchor: 6-month renewal trigger.

Every 3 months, the agency presents a quarterly review: the 12 shipped assets, the 8 campaigns, the qualified pipeline attributed, the budget actuals, the model-token cost curve, the gate-pass rate. The operator scores the agency on 5 dimensions: ship velocity (did assets land on cadence), output quality (founder-voice fidelity), outcome attribution (pipeline causally traced), partner posture (responsiveness + decision-making), forecasting accuracy (did the agency call shots that worked).

A score below 3.5 of 5 on any dimension triggers a structural conversation: scope adjustment, fee adjustment, or off-board. In the FORKOFF cohort, the median quarterly score is 4.2, and the off-board rate is 8% per year. The review is mechanically embedded; it does not depend on the founder remembering to ask.

Who owns what: the operator vs agency surface

A common SOW failure pattern is unclear ownership: both parties think the other side is doing X, X doesn't ship, blame loops fire. The FORKOFF default retainer SOW assigns ownership explicitly per deliverable, with three lanes:

• Operator-owned: founder voice, calendar availability for sync + recordings, named-account list, internal brand guardrails, board reporting, legal review, final approvals.
• Agency-owned: asset drafts, campaign orchestration, distribution-channel execution, attribution dashboard, model-token procurement, weekly cadence facilitation.
• Joint: outcome gate definition, quarterly review, scope renegotiation, crisis response.

Joint items default to operator-decides if there is a tie. The agency cannot unilaterally redefine the outcome gate mid-cycle. The operator cannot unilaterally cut a deliverable mid-cycle without renegotiating fee. Both protections.

Add-on triggers: pre-priced, not surprise-priced

Every add-on in the FORKOFF retainer SOW has a named trigger and a pre-set rate. This prevents the surprise-invoice pattern that founders complain about across the AI agency category. Three trigger categories:

Volume add-ons (deliverable count exceeds cycle base): 41% of FORKOFF retainers fire one per quarter. Rates: $1,200 per founder-voice asset, $2,400 per distribution campaign, $800 per custom analytics cut. Volume triggers are usually founder-driven: the founder asks for an extra X thread when a competitor announces.

Scope add-ons (new channel or vertical): 22% of retainers fire one per quarter. Rates: $3,500 for a new channel pilot (e.g. adding YouTube as a 4th distribution surface), $5,500 for a new vertical adaptation (existing AI SaaS retainer extends to a fintech sub-product). Scope triggers are usually operator-strategy-driven.

Incident add-ons (crisis response): 8% of retainers fire one per year. Rates documented in line item 7.

If an add-on trigger fires, the agency invoices the line item on the next monthly invoice with a one-line description. No mid-cycle billing surprises. The operator approves the trigger before the work begins; if the operator declines, the work doesn't ship and the trigger doesn't bill. The 3-tier verification audit covers the verification spec depth that prevents trigger-creep into ambiguous "is this in scope" territory.

Per-vertical retainer scope variations

The FORKOFF default retainer SOW above applies to most AI/SaaS verticals as published. Three vertical adaptations are worth calling out:

AI/SaaS (12 of 23 cohort retainers): the default ships. Founder-voice content tilts toward technical depth (notebooks, benchmarks, head-to-head comparisons). Distribution leans X and HN.

Fintech (4 of 23): compliance review adds a fixed weekly line item ($1,800/mo). Distribution skips Reddit and weights LinkedIn + earned-media higher. Outcome gate often includes regulatory-narrative milestones, not just pipeline.

Web3 (3 of 23): KOL coordination becomes a named line item ($2,200/mo for tier-2 KOL ops). Distribution adds Farcaster + Telegram + crypto Twitter as primary surfaces. Token-token costs are tracked separately due to higher model-research density.

Dev tools (2 of 23) and healthcare (2 of 23): default ships with minor adaptations. The vertical AI agency pricing case studies covers the per-vertical pricing-and-scope deltas at depth.

Common scope traps founders fall into when signing a retainer

After auditing 75 pre-contract founder calls across H1 2026, four scope traps account for ~80% of the buyer-side regret cases:

Trap 1: hours bundled into deliverables. The SOW says "4 assets per month" but the contract underneath says "80 hours of agency time." When the operator asks for asset 5, the agency says "no extra cost, just extra hours from your bucket." Sounds reasonable. Fails in month 3 when the hours bucket is empty and asset 4 stalls because there's no time left. Fix: SOW must commit to N deliverables independent of hours.

Trap 2: vanity-metric outcome gate. The SOW says "outcome anchor: 50,000 impressions on X." Impressions are not pipeline. Agency optimizes for impressions, ships boosted-low-quality reach, founder gets zero qualified inbound. Fix: outcome gate must be a measurable buyer-side action (qualified inbound, scheduled call, attributed pipeline source).

Trap 3: unbounded "additional work". The SOW says "scope changes billed at standard rate." Standard rate is undefined. By month 2 the operator has 6 ambiguous trigger-events on the invoice. Fix: every add-on category has a named trigger AND a named rate, pre-signed.

Trap 4: model-token markup bundled in retainer. Agency bills $14,800/mo and quietly absorbs $1,200/mo of model token cost as margin (8% of retainer). When tokens spike to $2,400 in a month due to a new campaign, the agency under-uses models to protect margin. Fix: tokens pass through at cost + handling fee.

The credibility vs user-acquisition campaigns analysis covers the upstream lane-pick decision (which trap-prone scope to even sign in the first place).

The 3-question pre-contract scope audit

Before signing any AI marketing agency retainer in 2026, ask the agency these three questions. Get the answers in writing as an addendum to the SOW.

Question 1: List every deliverable you will ship per month, with a number next to it. (Not "4 to 6 assets", not "marketing support", but: 5 founder-voice X posts, 1 LinkedIn long-form, 1 podcast pitch sequence to 8 shows, 1 monthly attribution report.) If the agency cannot list deliverables countably, the SOW is not real.

Question 2: What is the single outcome metric I can check at end-of-month to know whether this cycle worked? (Not "engagement up", not "good content shipped", but: 8 qualified inbound conversations from named-account targets, OR 1 podcast tier-1 placement, OR specific named milestone.) If the agency cannot name a single check, the SOW is not measurable.

Question 3: What triggers an add-on charge, and what is the rate? (Not "if scope changes, we'll let you know", but: an 8th content asset = $1,200, a new channel = $3,500, a crisis = $4,000-$12,000.) If the agency cannot pre-price triggers, every monthly invoice is a negotiation.

Agencies that answer all three questions cleanly have done the SOW homework. Agencies that hedge on any of the three are either not yet operationally mature, or they are deliberately leaving themselves room to scope-creep your budget. The fractional CMO vs AI agency buying shift analysis covers when scope ambiguity is a feature (fractional CMO model) vs a bug (AI agency model).

The 90-day onboarding cycle

For the founder-side accountability manual that runs on top of this SOW, see the B2B SaaS first 90 days with growth agency operating manual.

The FORKOFF retainer ships in three phases over the first 90 days. Each phase has named deliverables and a release gate.

Phase 1: Voice calibration (days 1-30). The agency studies the founder's existing content, runs 2 voice-recording sessions (60 min each), drafts 3 sample assets, gets founder approval on style + voice. Deliverables ship slower in phase 1 (3 assets vs the steady-state 4-6) because the agency is still learning. Distribution is paused until phase 2.

Phase 2: Distribution wiring (days 31-60). Voice is calibrated. The agency runs the first 2 distribution campaigns, wires attribution, ships the first monthly report. Founder approves channel mix + KOL outreach list + targeted publication list. Steady-state cadence emerges by day 60.

Phase 3: Steady-state + outcome gate (days 61-90). Full cadence. First outcome gate fires at day 90: did the founder receive N qualified inbound conversations from named-account targets attributable to retainer assets? Gate-hit releases the next cycle at the same price. Gate-miss triggers structured renegotiation.

This phasing is why 90 days is the FORKOFF minimum retainer term. Shorter terms under-rotate the voice-calibration investment + don't give attribution enough time to stabilize. The SaaS 2026 distribution gated founder funnel reset covers the broader funnel implications of this phasing.

Why the FORKOFF retainer SOW is published verbatim, not gated

Most AI marketing agencies treat the SOW as confidential and only show it after a signed NDA + a discovery call. The FORKOFF position is that publishing the SOW publicly forces three things:

Forcing function 1: pricing discipline. If every founder can read what's in a $15K retainer, FORKOFF cannot quote $25K for the same scope. The market disciplines us. Disclosure is a commitment device.

Forcing function 2: pre-call qualification. Founders who land on this post and decide the FORKOFF SOW doesn't fit their stage (too early, too late, wrong vertical) self-disqualify before booking a call. The intro-call show-rate at FORKOFF is 92% post-publish (n=23 cohort) vs ~70% industry average, because the SOW is the qualification.

Forcing function 3: competitive transparency benchmarking. Other AI marketing agencies who land on this post can see what FORKOFF ships per cycle and either match the transparency or accept that their SOW is the weaker negotiation surface. Either outcome is fine; the category gets cleaner.

Quarterly refresh commitment

This post documents the FORKOFF default retainer SOW as of 2026-Q2. The line items, cadence, and add-on rates will shift as the AI agency space matures + as model costs change + as the FORKOFF cohort grows. This post is refreshed quarterly with the prior quarter's cohort data; the lastUpdated frontmatter field always reflects the most recent refresh.

If you're reading this on a page where lastUpdated is more than 90 days old, the SOW you'll receive at intro-call time supersedes this page. Otherwise, what you read here is what you get.

What's NOT in this scope

To make the SOW comprehensive in both directions, here's what FORKOFF retainers explicitly DO NOT include:

• Performance media buying (Meta ads, Google ads, paid LinkedIn). FORKOFF runs organic + earned distribution; paid is a separate engagement (or a different vendor).
• Brand identity work (logo, visual system, brand book). Marketing-foundation engagement (separate FORKOFF service line) handles this.
• Sales enablement collateral (decks, one-pagers for sales team use). Founder-voice assets serve the marketing-driven pipeline; sales-team decks are out of scope.
• Hiring or fractional CMO services. FORKOFF runs the engagement; we don't headhunt or co-lead an operator's marketing function.
• Vendor evaluation for non-AI tools. We will recommend AI tools we use internally; vendor evaluation for CRM, CDP, MA platforms is out of scope.

If any of these are required, FORKOFF either declines the engagement (clean fit-failure) or proposes a hybrid arrangement (rare; requires operator + agency principal sign-off).

The retainer-vs-project decision

Not every founder should sign a retainer. Three patterns where a single project beats a retainer:

Pattern 1: pre-PMF founders running a 1-month launch sprint. A project-priced launch package (typically $8,000-$18,000 for 30 days) beats a $14,800/mo retainer because the founder cannot commit to the 90-day calibration phase.

Pattern 2: post-PMF founders with one specific weakness. If the operator has marketing operations in-house and just needs FORKOFF for podcast guesting, a $6,000-$10,000 quarterly project beats a full retainer.

Pattern 3: incident-only engagement. If the only need is crisis response on a specific event (model launch backfire, regulatory call-out), the flat-fee incident engagement is right. Retainer is overkill.

For everything else, the retainer compounds value over 6-12 months at higher unit economics than project-priced. The credibility vs user-acquisition campaigns analysis has the time-horizon math.

Operator-note: how to use this post in practice

If you're a founder evaluating AI marketing agency retainer proposals (FORKOFF or otherwise), do these three things:

1. Print the dataTable above. Bring it to every intro call. Ask the agency to fill in their version next to FORKOFF's.
2. Run the 3-question pre-contract scope audit. Get written answers as a contract addendum.
3. Force the agency to commit to one outcome gate per cycle. If they cannot name one, the SOW is not real.

The post is yours. The FORKOFF SOW is yours. Run a clean procurement.

Want a personalized scope for your domain? Book a 15-minute call. Want the form-based path with attached briefs? Use the contact form.