Most "how we use AI" posts from agencies are either marketing fluff or quiet panic dressed up as innovation. Here's a plainer version: an honest map of where the tools genuinely help, and where they don't get near the decision.
This is the actual shape of the workflow at Creative Hero, a small studio, not a framework built for a case study.
The adoption is real, and it's fast
Studios that pretend they aren't using AI daily are either behind or not being straight with you. Salesforce's State of Marketing research found that 87% of marketers now use generative AI in at least one workflow, up from 51% in 2024, according to reporting summarized in AI marketing adoption research for 2026. That's not a niche behavior anymore, it's the baseline. The honest question isn't "do you use AI," it's "where exactly, and who's checking it."
Where AI accelerates the work
Research and first drafts. Competitor scans, keyword groundwork, first-pass copy structure, meeting notes turned into action items. This is where speed gains are real and low-risk: the output gets checked before it goes anywhere near a client.
Variation and exploration. Generating five headline directions, ten layout variants, or a batch of social caption drafts to react to. Nobody ships the first output. It's raw material for a human decision, the same way a mood board is raw material, not a final design.
Code and site scaffolding. First-pass component structure, boilerplate, repetitive CSS, test data. This is where AI genuinely returns hours, because the pattern is well understood and the risk of a subtle error is caught in review, not shipped blind.
Internal process. Meeting summaries, project status drafts, first-pass client emails that get edited before sending. Useful, low-stakes, and it frees up time for the parts of the job that need a human brain at full capacity.
Where the human stays fully in charge
Brand voice and positioning. The tone rules, the specific words a brand does and doesn't use, the actual strategic bet on who a client is talking to: that's judgment built from knowing the client, not something delegated to a prompt.
Anything client-facing before it's reviewed. No AI output goes to a client without a human reading every line first. Not a policy for show, a hard rule.
Final visual and copy decisions. AI can generate ten directions. It doesn't get a vote on which one ships. That decision sits with the person who understands the brand, the market and the client relationship.
Fact-checking and claims. Every statistic, every claim about performance, every number in a proposal or blog post gets verified against a real source before publication. This is treated as a hard gate, not a nice-to-have.
Client relationships and strategy calls. The conversation where a client is nervous about a redesign, or a pricing decision needs to account for a relationship history: that's a human conversation, always.
Where AI sits in the process
| Stage | AI role | Human role |
|---|---|---|
| Research / competitor scan | Does the first pass, fast | Verifies findings, spots what's missing |
| Copy first draft | Generates structure and raw text | Edits for voice, cuts, verifies claims |
| Design exploration | Produces variation quickly | Picks direction, refines to brand standard |
| Code / site build | Scaffolds components, boilerplate | Reviews logic, tests, owns final quality |
| Client communication | Drafts a first-pass reply | Reads, edits, sends |
| Strategy and pricing | No role | Full human decision |
| Final sign-off, every deliverable | No role | Full human decision, always |
A working checklist for other small studios
- Decide in advance which decisions AI is never allowed to make alone (brand voice, final sign-off, pricing, strategy)
- Require a human read of every line before anything reaches a client
- Verify every number and claim against a real source before it's published
- Use AI for volume and speed (drafts, variations, scaffolding), not for judgment
- Say plainly, to clients and publicly, where AI is used in your process. Vague claims about "human-crafted" work that quietly used AI everywhere erode the same trust this whole approach is meant to protect
FAQ
Doesn't using AI this much make the work less "handmade"? The parts a client is actually paying for, the strategic call, the specific line, the final say, are still fully human. AI speeds up the scaffolding around that decision, it doesn't replace it.
How do you keep quality consistent if AI is involved in drafts? Every single deliverable gets a human review pass before it ships. That's the actual quality control, not a claim about how the draft was produced.
Will this balance change as the tools get better? Probably at the edges: more of the scaffolding work will get faster. But the decision layer, taste, judgment, relationship, verification, is not a speed problem AI is positioned to solve, and it's the part clients are actually hiring a studio for.
Curious what this looks like on an actual project? Creative Hero is a small studio in Antwerp running exactly this process on branding, web and marketing work.
