As I said to a client: I don't need 18 years of experience to write a text. I need it to know when a good text is the wrong text.
That one sentence is the honest thesis behind every AI tool I use. Not "AI replaces experience." Not "AI makes anyone a marketer." The actual argument is more interesting and more uncomfortable: AI eliminates the hours spent on technically correct but cognitively routine work, which surfaces a harder question, do you have the judgment to know what correct even means for this specific situation?
Most AI marketing content skips that part. It tells you which tools to use, gives you the time savings, and stops there. I want to be more honest about what this actually means, because the tools are the easy part.
The Industry Context Worth Knowing
AI adoption in marketing is accelerating faster than governance is following. Salesforce's 2024 State of Marketing report found that 75% of high-performing marketing teams are now using AI in their work. A separate finding from the same study: 62% say AI tools help them deliver personalized experiences at scale, but fewer than 40% say they have formal frameworks for quality control on AI-generated content.
That gap is where the risks live. More volume, less oversight. Faster output, fewer checks. The companies winning with AI marketing are not the ones with the most tools. They're the ones that have figured out which decisions still require human judgment and structured their workflows accordingly.
I run as a Fractional CMO for multiple clients simultaneously. That means I'm managing campaigns, writing strategies, interpreting data, and producing content, for different companies, in different markets, at the same time. The only reason that's possible without a team of 10 is AI. But it works because I've been explicit about what AI is actually for in my workflow.
I've tracked it: 8 to 12 hours saved per client per week, depending on the workload. Here's where those hours come from.
I haven't written a first draft from scratch in over a year.
Content and Copy: Claude + ChatGPT
For long-form writing and structural copy, blog posts, landing pages, email sequences, strategy documents, I use Claude. It handles complexity well, holds context across long documents, and produces output that's closer to what I'd actually publish.
For brainstorming and generating variations, I use ChatGPT. When I need ten different angles on a headline or five ways to open a LinkedIn post, ChatGPT generates variety faster.
I haven't written a first draft from scratch in over a year.
The way I work: I brief the AI with my voice, the target audience, the goal of the piece, and any constraints (tone, word count, key claims). I edit the output. The whole process, brief, generate, edit, polish, takes about 20 minutes instead of 90. For a 1,000-word post, that's over an hour back in my day, every single time.
The one thing that made this work was building a proper prompt template that captures my writing style. Without that, AI-generated copy sounds like AI-generated copy. With it, the output is close enough that most people who know my writing style can't tell the difference without looking carefully.
Research and Competitor Intelligence: Perplexity + Claude
Perplexity handles real-time market information, recent news, competitor moves, industry developments. Claude handles synthesis, turning raw information into structured analysis.
Before AI, a proper competitor analysis was half a day of work. Open 20 tabs, read through everything, take notes, pull it into a document, make sense of it. Now it's this: Perplexity for the landscape overview, Claude for the structured analysis, 15 minutes of review. Total time under an hour.
I also use this for client call preparation. I upload call transcripts into Claude and ask it to identify patterns: What objections come up repeatedly? What outcomes matter most to this client? What's the gap between what they think the problem is and what the data suggests? That analysis goes directly into the copy brief, so the messaging is grounded in actual client language, not my assumptions about what they want to hear.
Paid Advertising: Google Performance Max + Meta Advantage+
Both platforms have shifted heavily toward AI-driven distribution and bidding. I've stopped fighting that.
What that means practically: I spend more time on inputs and less time on daily optimization. Copy variants, creative assets, audience signals, landing page alignment, that's where my attention goes. The AI handles distribution and bid management. I handle strategy and creativity.
This only works if the inputs are good. Performance Max running on weak creative and generic copy will find an audience, just not the right one. The quality of the human input determines the ceiling of what the AI optimization can achieve.
Analytics: GA4 + Claude
I export GA4 data and feed it into Claude for pattern interpretation. My standard prompt is something like: "These are 8 weeks of session and conversion data across three markets. What are the 3 most notable patterns, things that are meaningfully different from what you'd expect?"
That question alone returns a direction for analysis in under 5 minutes. I then dig into the specifics. Before, I'd spend an hour staring at dashboards before I found something worth acting on. Now I find the signal first and validate it second.
For board reports, the process is: raw GA4 data → Claude builds a structured narrative → I adjust for the specific management team audience. A report that used to take 3 hours takes 45 minutes, and the output is more focused because I'm not spending cognitive energy on structure, only on judgment.
SEO: Semrush + Claude
Semrush does the data work: keyword gaps, ranking positions, competitor analysis. Claude does the content strategy work: clustering themes, identifying the angle that fits both search intent and my client's positioning, drafting a content calendar with rationale.
On-page SEO used to take me close to an hour per page, reading the content, identifying gaps, rewriting titles and meta descriptions, checking heading structure, making sure the copy actually addressed the search intent. Now it's about 20 minutes. The structural stuff is fast; the judgment calls are still mine.
How It Fits Together: A Typical Week
Monday: Perplexity + Claude market update for each active client. 30 minutes total. I know what's relevant in their industry before any meetings happen.
Tuesday: GA4 + Claude narrative for the clients with reporting due. 45 minutes instead of 3 hours.
Wednesday: Semrush data + Claude content brief. An hour instead of two.
Thursday: Claude drafts an article based on the Wednesday brief. I edit and prepare for publishing.
Friday: Performance Max and Meta campaigns reviewed. Inputs adjusted if needed. 30 minutes per client.
Across a full client engagement, I'm saving approximately 10 hours per week on work that used to require manual effort and mental energy. That's not 10 hours of busywork, it's 10 hours of strategic work I can now do for other clients, or not do at all.
What AI Can't Do, and Why That's the Whole Point
Relationships. Judgment. Assessing whether someone is the right cultural fit for a leadership role. Navigating a disagreement in a management team meeting. Knowing when the data says one thing and the context says another, and having the experience to trust the context.
I don't use AI for any of those things. I use AI for the work that doesn't require 18 years of experience to produce competently.
That distinction matters more than it might seem. There's a version of this where you use AI to scale mediocrity, where you produce more content, more reports, more analysis, all of it fast and none of it particularly good. That's not what I'm describing.
The version I'm describing is this: I use AI to eliminate the hours I used to spend doing technically correct but cognitively routine work, so that my actual experience, knowing which insight matters, which message will land, which strategic direction is right for this specific market, gets applied where it actually creates value.
Which brings me back to where I started: I don't need 18 years of experience to write a text. I need it to know when a good text is the wrong text.
A text can be well-structured, on-brand, and grammatically clean, and still be the wrong message for this market, this buyer, this moment in the deal cycle. AI will produce the text. The judgment about whether it should exist at all, whether it's the right angle, whether the framing actually serves the business goal, that's still a human job.
That's where experience compounds. Not in producing the output. In knowing what the output should be for.
If you want to understand what a structured AI marketing workflow could look like for your company, not in theory, but in operational detail, book a 45-minute Marketing Scan.
Sources & References
- Salesforce, State of Marketing Report, 8th Edition (2024), AI adoption rates: 75% of high-performing marketing teams use AI. Personalization at scale: 62%. Formal AI governance frameworks: under 40%. https://www.salesforce.com/resources/research-reports/state-of-marketing/
- Google, Performance Max Best Practices Guide (2024), On the role of human creative inputs as the determinant of AI bidding performance ceiling. https://support.google.com/google-ads/answer/10724817
- Own workflow data (2025–2026): Time savings per client per week, content (70 min saved per 1,000-word post), competitor analysis (3.5 hours → under 1 hour), board reporting (3 hours → 45 min), SEO on-page (60 min → 20 min per page). Average: 8–12 hours/client/week.