Marketing
The Amplifier
Sees the pattern in a crowd and turns a message into momentum.
Motif · Concentric rings, a beacon
Figure — Reported time saved
2–3 hrs
per content piece, and around 2.5 hrs a day for marketers who use it daily
Handles variations, resizing, first drafts and A/B production, so people are freed to set strategy, define the brand voice, and build genuine audience connection.
Per content piece, and around 2.5 hrs a day for marketers who use it daily. Every figure on this page is traced to a named source below, with its method and its limits.
Read the evidence01
What these agents do
Marketing agents plan campaigns, segment audiences, draft and adapt copy, schedule delivery across channels, and report on what happened. The more autonomous ones also move budget between channels while a campaign runs.
02
Genuinely good at
- Producing many variants of a message quickly and consistently
- Holding a brand system steady across channels and languages
- Reallocating spend faster than a weekly human review cycle
- Summarising campaign performance without waiting for an analyst
03
Genuinely bad at
- Judging whether a claim is defensible in a regulated market
- Knowing when a cultural moment makes a scheduled message wrong
- Explaining why a specific creative choice was made
- Working from thin or poorly labelled first-party data
04
The risks that matter
- Reputational: an unreviewed message published at scale is difficult to recall.
- Compliance: advertising claims and personalisation both fall under existing law before any AI rule applies.
- Data: audience building often means processing personal data — check the lawful basis, not just the feature list.
- Oversight: 'approval required' is only meaningful if approving is realistic at the volume the agent produces.
05
How to evaluate one responsibly
- Ask what happens between the agent deciding and the public seeing it.
- Ask for the error rate, not the speed gain. Speed is easy to demonstrate.
- Test it on your worst brief, not your cleanest one.
- Check whether spend authority can be capped in the product, not just in policy.
Now compare what is actually declared.
The registry holds each agent’s stated facts — autonomy, oversight, compliance, residency, sustainability disclosure — with provenance on every field.
Compare marketing agents06
The evidence
- Independent research
- No commercial interest in the outcome. Method published.
- Analyst research
- Structured method, but the publisher sells advice in this market.
- Practitioner survey
- Self-reported. Good for direction, unreliable for magnitude.
- Vendor or customer material
- A claim about the publisher's own product. Not a finding.
Figures below are reproduced as published. We label who paid for the work and what the method can and cannot show. Where a number is self-reported or vendor-published, we say so next to the number rather than in a footnote.
4 of 4 sources
- 01Independent research
A field experiment with 758 consultants found those given GPT-4 completed 12.2% more tasks, 25.1% faster, with results rated 40% higher in quality — on tasks inside the tool's capability.
- Source
- Harvard Business School / BCG — Navigating the Jagged Technological Frontier (Dell'Acqua et al., working paper 24-013)September 2023
- Method
- Pre-registered randomised controlled trial, 758 consultants, two task sets inside and outside the model's frontier.
- What it does not show
- On tasks outside the model's capability, assisted participants were 19 percentage points less likely to be correct. Consulting tasks, not marketing execution.
- 02Independent research
A randomised trial of 453 writers on mid-level professional writing tasks found time on task fell 40% and output quality rose 18%.
- Source
- Science (Noy & Zhang, MIT) — Experimental evidence on the productivity effects of generative AIJuly 2023
- Method
- Randomised controlled trial, 453 college-educated professionals, blind grading.
- What it does not show
- Short, self-contained tasks. Says nothing about campaign-level outcomes, brand consistency, or the review time a regulated market adds.
- 03Practitioner survey
Marketers report generative AI saving roughly three hours per content piece, and around 2.5 hours a day overall.
- Source
- Marketing practitioner surveys — Aggregated practitioner survey reporting2024–2025
- Method
- Self-reported time savings, opt-in respondents, no control group.
- What it does not show
- Respondents who adopted AI are more likely to answer. No measurement of rework, review, or output that was discarded.
- 04Vendor or customer material
Around 80% of marketers use AI somewhere in content production, with image tooling accounting for most of the reported design-time reduction.
- Source
- Industry surveys — Adoption reporting across marketing-technology vendors2024–2025
- Method
- Vendor-run adoption surveys of their own user bases and mailing lists.
- What it does not show
- Adoption is not effectiveness. Sample is drawn from people already buying the tooling.
We take no payment from any organisation named on this page, and no source is listed or omitted on commercial grounds. If a figure here is wrong or out of date, tell us and we will correct it with the date of the change.
Ideas in this domain
Continue