Agent-native demand generation is live at real scale, and nobody has published a controlled customer-acquisition-cost comparison
The question
"What agent-native demand-generation tactics (SEO/content agents, outbound agents, agent-run ad optimization) are actually live in production today, and do they measurably cut customer acquisition cost versus human-run equivalents?"
Raised to adjudicate a live tension in [[01-projects/agentify-3000/opportunities]]: the project scored "find" (demand generation) as the weakest agent-shaped job at 1.9-2.0 of 3, and the founder pushed back on 2026-10-04 that agents can generate demand. That pushback is currently recorded as a caution on distribution rather than a ceiling.
What we already know (from the vault)
- The score and the pushback, verbatim. [[01-projects/agentify-3000/archetypes]] has find at 2.0 for business-to-business software-as-a-service (saas, n=513) and 1.9 for consumer apps (n=200), the lowest job score in each. [[01-projects/agentify-3000/opportunities]] records the resolution as: "For saas and consumer_app, find is the weakest job (1.9-2.0 of 3, still mostly agent-shaped: outreach, SEO); founder pushback 2026-10-04: agents can generate demand, so this is a caution on distribution, not a ceiling."
- The score is a model grading itself, not a measurement. archetypes.md states the scores come from a fixed-prompt model reading a 300-character public description per business, spot-checked by the producer rather than an independent reader. So the 1.9-2.0 figure carries no measured CAC behind it either. Neither side of this tension is currently backed by a measurement.
- archetypes.md already names the mechanism this brief tests. Its own conclusion on software and content: "Distribution is still bought with a face or a decade of SEO." The opportunities.md risk column prices the same thing per row: "backlinks are relationships" (Rankings.io / Hennessey), "law firms buy trust", "clients buy account managers" (Impression), "deliverability reputation takes years" (ConvertKit), "Google punishes AI lifestyle content" (Man of Many).
- RDCO's own demand-generation research has twice landed on the same split. [[2026-05-30-fde-capture-vs-create-demand]] found measured supply-side heat (forward-deployed-engineer job postings up ~729% year over year per Indeed data cited by The New Stack) with near-zero buyer-intent search footprint, and concluded the play is category creation, not search capture. [[2026-08-24-idle-capital-buying-demand-playbook]] concluded there is no documented six-figure buy-demand playbook, that renting attention is "a price list, not a playbook," and that it only works behind a conversion surface that already converts.
- The current play is credibility scored by pipeline, not subscribers. [[2026-09-29-nov30-credibility-targets]] sets six build-in-public pieces by Nov 30 (tally 0/6 as filed) and scores on phData pipeline, with the explicit frame "the score is phData pipeline, not followers."
What the web says
- Agent-run paid-ad optimization is the only tactic genuinely at production scale, and it is the one with the most adverse evidence. Meta Advantage+ automation reportedly "touches 82% of Meta advertisers and runs 62% of ecommerce ad spend on the platform" (adbeacon.com figure, cited secondhand by playad.ai; the primary dataset is not shown). Label: secondhand vendor-blog citation.
- The automation CPA claims all fail the baseline-plus-N-plus-window test. Meta's own internal data claims a 22% return-on-ad-spend (ROAS) lift for simplified Advantage+ structures; jetfuel.agency publishes a 17% lower cost-per-acquisition (CPA) "industry benchmark" versus manual campaigns. Neither states a sample size or a time window, and the Meta figure is the platform grading itself. playad.ai, which reproduces both, says plainly that it reproduces none of the underlying datasets. Label: vendor and platform marketing claims, not measurements.
- The most-cited aggregate counter-finding could not be verified at its primary source. Multiple secondhand summaries report that Haus analyzed 640 Meta incrementality tests and found 58% of brands saw higher incremental ROAS on manual campaigns than on Advantage+, with Advantage+ overreporting by roughly 12 percentage points. I went to haus.io directly and could not find that report in its published research, blog listings, or case studies. Label: secondhand, unverified at source. Do not cite the 640 or the 58% as established.
- Haus does publish measured, holdout-controlled single-brand results, and they are not agent-versus-human comparisons. Dr. Squatch +6.51% new revenue lift; Grüns roughly 25% lower cost per incremental acquisition; FanDuel +11% sportsbook activations. Each is one brand, measured against a geo holdout. Label: measured results with controls, but the treatment is a channel, not an agent.
- One secondhand CAC series points the wrong way for the automation thesis. An analysis of 55,000 Meta campaigns reportedly found Advantage+ new-customer acquisition cost rising from $257 (May 2024) to $528 (May 2025) while reported ROAS held near $4.52. Label: secondhand, unverified. Also confounded: auction prices rose industry-wide over that window, so a doubling is not attributable to automation.
- Outbound is where the one usable measurement lives. Digital Applied published a 100,000-cold-email comparison over six months (October 2025 to April 2026): 50,000 artificial-intelligence-generated, 50,000 human-written, each AI email paired with a human email on persona, ideal-customer-profile firmographic, sequence stage, sender domain age and sender domain-authority score. Deliverability came from Gmail Postmaster Tools and Microsoft SNDS. Results: reply rate 4.1% AI versus 5.2% human; positive reply 1.4% versus 2.1%; meetings booked 0.7% versus 1.1%; spam-flag rate 8% versus 3%; bounce 6% versus 6%; inbox placement 71% versus 86%. The study reports no cost per meeting, no cost per opportunity and no CAC. Label: measured, observational (matched, not randomized); publisher sells AI sales-development-representative services.
- Named outbound vendors have published nothing comparable. The only Artisan figure surfaced was a 3.6% positive reply rate in a SellScale/NewtonX roundup of vendor case studies, with no human baseline; that roundup itself says controlled A/B data and full-funnel revenue proof remain scarce. Searches returned no published results at all for 11x or Regie. A widely repeated "AI-only booked 847 meetings at 11% conversion versus hybrid 312 at 38%" comparison has unverified methodology. My inference, labeled as such: that comparison favors hybrid on absolute outcome (847 x 0.11 = 93 closed-equivalents versus 312 x 0.38 = 119), the opposite of how it is usually quoted.
- SEO and content agents show quality parity and a large cost-per-output gap, with no credible CAC link between the two. Semrush, as reported by grafit.agency, found 57% of AI articles and 58% of human articles reached Google's top 10: near parity on ranking. On cost, wedoworldwide.com reports AI posts at $35 versus $400 human with organic traffic 12% lower (1,623 versus 1,847 visits per post), giving no source or method. Single-site anecdotes: a Seer Interactive agent moved one phrase from position 12 to 6 in seven days with clicks up 28% (n=1, no control); an SEOProfy workflow surfaced 34 cannibalization issues with primary pages gaining about 5.2 positions in 45 days (small sample, no control). Labels: Semrush figure is secondhand report of a vendor study; the cost figures and anecdotes are unsourced vendor claims.
- The one explicit CAC percentage I found has no baseline, no N and no window. thestacc.com reports AI content lowering customer acquisition cost by 37% and costing up to 4.7x less than human content, both attributed to Loopex Digital. There is no control group, no sample and no time window. This is not evidence of a CAC cut and must not be quoted as one.
- A category the question did not anticipate: agents are measurably strongest at diagnosing demand, not creating it. quickseo.ai's aggregation reports SEO-audit agents returning 11.4x median return on investment against 2.9x for content-brief agents, with a losing quartile at 0.7x, and 49% of respondents naming "proving results" as the binding constraint. The whole incrementality-measurement category (Haus, geo holdouts, marketing-mix modeling run at a cadence humans cannot sustain) is the same shape. Label: aggregator figures, methodology unstated; the directional pattern is consistent across sources.
- Skipped and flagged: no paywalls blocked this run. The one source I actively failed to reach was the Haus aggregate report, which is absent from haus.io's public research index rather than gated.
Convergences and contradictions
- Strong convergence on the mechanism, from opposite directions. archetypes.md's "distribution is still bought with a face or a decade of SEO" and the 100,000-email study's inbox placement of 71% versus 86% are the same claim in different units. Every measured gap the web produced sits at a trust or deliverability chokepoint (spam-flag 8% versus 3%, meetings booked 0.7% versus 1.1%, inbox placement 71% versus 86%), not at a capability chokepoint. On ranking quality, where trust is not the gate, AI reaches parity (57% versus 58%).
- Direct contradiction between vendor CPA claims and the only aggregate test evidence anyone cites. Platform and agency sources claim automated campaigns cut CPA 17% to 22%; the incrementality-testing community's position is that platform-reported numbers overstate incremental contribution. I could not verify the specific aggregate that is always cited for this, which leaves the contradiction live rather than resolved.
- The vault and the web agree that the binding constraint is downstream of production, and the vault got there first. [[2026-08-24-idle-capital-buying-demand-playbook]] already concluded that spending on attention without a conversion surface "buys traffic, not demand." Nothing on the web contradicts that, and the SEO cost figures ($35 versus $400 per post) make it sharper: the thing agents cut is the cost of the unit, and the unit was never the bottleneck.
Synthesis for RDCO
The primary result is the absence. Across three tactic families, three web searches, three source fetches and a direct attempt at the primary source of the one aggregate dataset everyone cites, I found zero published comparison where the endpoint is an acquired customer, there is a human-run control, and the sample size and time window are both stated. Every CAC percentage in this space traces to a vendor blog citing another vendor blog. That absence is not a gap in my search; it is the state of the evidence, and it answers the question that was asked. Nobody has demonstrated that agent-native demand generation cuts customer acquisition cost, and nobody has demonstrated it does not. What has been demonstrated, repeatedly and with real numbers, is that agents cut the cost per unit of output.
The evidence supports the reframing, and inside the reframing it lands closer to agentify-3000's low score than to the founder's pushback. I am flagging that explicitly per the standing rule on conviction asymmetry, and I am not going to soften it: on production evidence as of today, "find" is weak, and it is weak for the reason opportunities.md already guessed. But the score itself is the wrong instrument. "Find" is not one job. It is produce (write the post, draft the email, generate the creative variant) and earn-attention (land in the inbox, hold the ranking, be trusted enough to get a reply). Agents score near the top of the scale on produce and near the bottom on earn-attention. A mean of 1.9-2.0 is an average over a bimodal split, and averaging a 3 with a 1 to report a 2 hides the only distinction that matters operationally. The defect in the scoring model is the taxonomy, not the number. Splitting find into produce and earn-attention is the change this brief recommends, and it would simultaneously vindicate the founder (agents really can generate demand-generation activity at volume he would be right to call transformative) and preserve the project's finding (the activity does not convert into customers at a lower cost, because the conversion step is gated on trust).
One trend line argues the founder is right on a longer horizon. In the same 100,000-email dataset, AI reply rate moved from 2.8% (2024) to 4.1% (2026) while human moved 4.8% to 5.2%. The gap narrowed from 2.0 points to 1.1 points in two years. If that convergence rate holds, reply-rate parity arrives before 2028, and the "caution, not a ceiling" language in opportunities.md is the correct read on a two-to-three-year horizon even though it is not yet the correct read on today's numbers. That single series is the most load-bearing number in this brief and it deserves re-checking when the study is repeated.
What would have to be true for the low score to flip. Stated falsifiably, any one of these: (1) a published geo-holdout or randomized experiment in which an agent-run acquisition channel delivers lower incremental cost per acquired customer than a matched human-run channel, with at least 1,000 acquisitions, a window of at least eight weeks, and a control run by a party that does not sell the agent; (2) trust-metric parity with a surviving cost gap, concretely agent cold-email inbox placement at or above 85% and AI articles matching human on both top-10 rate and traffic per post, while unit cost stays at least 5x cheaper, since parity on the trust metric plus a surviving cost gap is the exact mechanism by which cost-per-output becomes CAC; (3) a repeat of the 100,000-email study showing AI meetings-booked at or above the human rate (currently 0.7% versus 1.1%). Condition (3) is the cheapest tripwire and the one to watch.
For RDCO specifically, three concrete things follow. First, the current play is correctly specified and the measurement choice is the part the whole industry gets wrong. [[2026-09-29-nov30-credibility-targets]] scores on phData pipeline rather than followers; every CAC claim surveyed here is a traffic, reply or subscriber claim wearing a CAC costume. Keep the pipeline denominator and do not let a cheaper-per-post finding tempt a switch to volume metrics. Second, do not buy agent-run outbound for RDCO at any price. The only matched dataset in existence says agent cold email runs 2.7x the spam-flag rate and 15 points lower inbox placement, and RDCO's entire asset is a named person's credibility attached to a real employer. opportunities.md itself prices deliverability reputation as taking years. A burned sending domain would cost more than the pipeline it could plausibly produce. Third, and this is where the evidence is actually useful rather than cautionary: the thing agents measurably cut is production cost, and the Nov-30 sheet shows 0 of 6 build-in-public pieces shipped. The bottleneck on the credibility play is founder hours converted into published artifacts, which is precisely the cost agents do cut. Agent-native demand generation pays for RDCO as a production subsidy on the credibility play, not as an acquisition channel. That is the same conclusion [[2026-08-24-idle-capital-buying-demand-playbook]] reached about cash, arrived at independently about agents.
One Ray-selected opportunity, awaiting his read. The highest measured-ROI agent category in demand generation is diagnosis, not creation: SEO-audit agents at 11.4x against content-brief agents at 2.9x, and the entire incrementality-testing stack. In agentify-3000's taxonomy that is "judge," not "find," and judge is the job the project did not flag as weak. It also sits directly on the founder's stated domain of agents in production over data infrastructure. An agent-run incrementality or attribution harness, built in public on Snowflake, would be both in-domain and on the one part of demand generation where the evidence says agents win. Flagged as a Ray-selected candidate for the build-in-public queue, not a founder decision he has made.
Why this is in the vault
It adjudicates the open 2026-10-04 tension in [[01-projects/agentify-3000/opportunities]] between the 1.9-2.0 "find" score and the founder's pushback, and it recommends a specific change to the project's scoring taxonomy (split find into produce and earn-attention) rather than a change to the number. It also sets the measurement guardrail for the Nov-30 credibility play in [[2026-09-29-nov30-credibility-targets]]: keep pipeline as the denominator, treat agents as a production subsidy, and do not route RDCO outreach through an agent sender.
Open follow-ups
- Does the Haus 640-test aggregate on Advantage+ versus manual incremental ROAS exist as a published primary document, or is the figure a citation chain with no source? It is the most-repeated counter-finding in agent-run advertising and I could not locate it at haus.io.
- Has anyone published a geo-holdout or randomized experiment where the treatment is an agent-run acquisition channel and the endpoint is an acquired customer, in any industry? This brief searched marketing sources; the answer may sit in academic marketing-science literature rather than vendor blogs.
- What is the actual trend in cold-email inbox placement for agent senders across 2024-2026, and do the mailbox providers' own published policy changes explain the 71%-versus-86% gap better than message content does? If the gap is policy rather than content, it will not close with better models.
- Does the produce-versus-earn-attention split hold for the other four agentify-3000 jobs, i.e. is "deliver" at 1.5 for agencies also a bimodal average hiding a high production score and a near-zero trust score?
Related
- [[01-projects/agentify-3000/opportunities]]
- [[01-projects/agentify-3000/archetypes]]
- [[2026-08-24-idle-capital-buying-demand-playbook]]
- [[2026-05-30-fde-capture-vs-create-demand]]
- [[2026-09-29-nov30-credibility-targets]]
- [[2026-10-04-starter-story-pat-walls-assessment]]
- [[2026-03-05-not-boring-costless-sacrifice]]
- [[2026-01-15-every-social-dandelions-trust-spreads]]
- [[2026-05-11-practitioner-newsletter-acquisition-tactics]]
Sources
Vault
01-projects/agentify-3000/opportunities.md(2026-10-04) — the 1.9-2.0 find score, the founder pushback line, the per-row risk column01-projects/agentify-3000/archetypes.md(2026-10-04) — per-archetype job means, "distribution is still bought with a face or a decade of SEO", the 300-character-blurb scoring caveat01-projects/personal-brand/drafts/2026-10-04-fifth-job-01-four-of-five-v2.md— the published-facing framing of the five-job taxonomy06-reference/research/2026-08-24-idle-capital-buying-demand-playbook.md— no documented buy-demand playbook; "buys traffic, not demand"06-reference/research/2026-05-30-fde-capture-vs-create-demand.md— supply-side heat, no buyer-intent search footprint01-projects/personal-brand/2026-09-29-nov30-credibility-targets.md— pipeline-not-followers scoring, 0/6 pieces
Web (opened or returned in search results this run)
- https://www.haus.io/ — published research index and single-brand holdout results (Dr. Squatch +6.51%, Grüns ~25% lower cost per incremental acquisition, FanDuel +11%); the 640-test aggregate is absent here
- https://playad.ai/blog/marketing-insights/measure-black-box-campaigns — Advantage+ 82%/62% penetration (adbeacon.com), Meta's 22% ROAS-lift claim, jetfuel.agency 17% lower CPA; no underlying datasets reproduced
- https://www.digitalapplied.com/blog/ai-sdr-real-performance-100k-email-analysis-2026 — the 100,000-email matched comparison; publisher sells AI sales-development services
- https://www.grafit.agency/blog/ai-content-vs-human-content — Semrush 57%/58% top-10 parity (secondhand)
- https://thestacc.com/blog/ai-content-marketing-statistics/ — 37% CAC-reduction and 4.7x cost claims attributed to Loopex Digital; no baseline, N or window
- https://wedoworldwide.com/blog/40-ai-agents-every-growth-team-should-have — $35-versus-$400 per post, 1,623-versus-1,847 traffic, $142 cost per qualified lead; no sources given
- https://quickseo.ai/blog/ai-agents-for-seo-statistics-2026 — SEO-audit agents 11.4x versus content-brief 2.9x, losing quartile 0.7x, 49% name proving results as the constraint
- https://cited.so/blog/ai-seo-agent — Seer Interactive position 12-to-6 anecdote, SEOProfy 5.2-position gain
- https://www.skalestrategy.com/blog/meta-ads-incrementality-testing and https://pixis.ai/blog/advantage-vs-performance-max-head-to-head-2026 — secondhand carriers of the Haus 640-test and 55,000-campaign CAC figures
- X/Twitter recent search (
mcp__xmcp__searchPostsRecent) for operator claims on agent-run CAC and incrementality returned two posts in window, both marketing assertions with no measurement attached. No operator-level data found.