On July 28, Knak —the marketing production platform used by teams at Google, Amazon, Uber, Meta, and OpenAI— published a report that puts numbers to something we've been watching closely at Geek Vibes: AI isn't the bottleneck in marketing. Production is. The report, titled Marketing Production in the Age of AI, surveyed over 300 enterprise marketing leaders, and the central finding dismantles much of the narrative being sold about "AI that frees up time for strategy".
The number that matters: 70% adopted AI, but 88% still edits by hand
According to the report, 70% of teams already have AI deployed in production. But 88% say the output still needs moderate to substantial editing before it can be used. In other words: AI is getting teams to a first draft, not to launch. Jennifer Delevante, CMO of Knak, sums it up this way in the press release: "Marketers were told AI would give them time back for strategy. Our data says that hasn't happened yet. It's not a talent or creativity problem. It's the production layer —the gap between a good idea and a launched campaign."
That diagnosis holds up with a second hard number: 85% of marketing teams missed at least one planned launch date in the past 12 months, and 1 in 10 misses more than five times a year. The causes they report aren't about strategy or creativity —they're operational: approvals and sign-offs (47%), design and creative production (38%), and cross-team coordination (36%).
A single email can cost over $300 in internal hours
The report breaks down what it actually costs to produce a single campaign email in an enterprise organization:
- 60% of teams involve four or more people to produce a single email.
- 69% need between two and three review rounds before publishing.
- That translates to more than $300 in internal labor cost per send, calculated using compensation data from the Bureau of Labor Statistics.
- 54% use between three and five different tools to produce a single email.
- 51% still rely on email threads, Slack, or Teams as the main mechanism for workflow and approval.
This is the context that explains why generative AI, as impressive as it looks in a demo, isn't moving the needle where it hurts most: in the operational friction that happens after the idea has already been approved.
Where AI is being used (and where it isn't)
The report shows that AI use in marketing is concentrated in the most visible and least costly parts of the process, not the ones that actually consume time:
- 64% use it to generate first drafts of copy.
- 56% use it for image generation.
- Only 25% use it to build or code emails and landing pages, which is exactly where the biggest production load lives.
It's a pattern we also see with clients: AI gets adopted for what's easy to show in a demo (text, images), but the heavy lifting —layout, cross-client QA, ESP integration, legal approvals— still runs through the same old manual processes. And email remains the channel where most is invested: 88% of teams produce campaign assets for email (almost 50% more than the second channel, paid social, at 60%), and 52% plan to increase email investment over the next 12 months.
Why this hits even harder for LATAM agencies
The report surveyed enterprise marketers, many with large internal teams and budget for dedicated production platforms. At a typical Mexican or Latin American agency, the same problem exists but with fewer people absorbing the same number of review rounds: it's not four people per email, it's one person doing the work of four different roles —copy, design, QA, and client coordination— sequentially. That means the approval and coordination friction that the report measures in team hours translates directly into late nights and weekends at a small agency. Generative AI helps with the first draft, but if the real bottleneck is waiting for client sign-off or reconciling three rounds of changes, no language model solves it alone. What does help is redesigning the approval flow itself —fewer back-and-forths, fewer parallel communication channels— and that's where the conversation with the client is worth more than any new subscription.
What this means for agencies and marketing teams
If your agency or internal team is evaluating where to put AI first, this report sends a clear signal: the ROI isn't in generating more drafts faster, it's in tackling the production layer.
- Audit the process before buying another generation tool. If the real bottleneck is approvals (47% report it as the main cause), no copy or image tool will solve it. What's needed is a shorter approval flow, not a faster model.
- Count how many tools touch a single deliverable. If your team uses three to five different platforms to produce an email, each integration is a friction point and source of error. Consolidating the stack usually yields more than adding one more AI.
- Measure the real cost of production, not just the cost of the AI subscription. $300 in internal hours per email is a number most teams don't have calculated. Without that number, it's impossible to make the business case for investing in solving the root problem.
- Don't sell "AI that saves time" without specifying at which stage. Generating a copy draft in seconds doesn't change the full lifecycle of a campaign if there are then two or three manual review rounds. The value promise needs to be specific about which friction it eliminates.
The bottom line: the next wave of AI value in marketing won't come from models that write better copy —that's largely solved. It'll come from whoever figures out how to embed AI into the approval, review, and publishing flow, which is where time really gets lost.
Has your team measured how much it costs to produce a single campaign, from idea to launch? If you want us to review it together, write to us at comercial@geekvibes.agency.



