Six months ago, a small marketing team with a growing client list made a decision: AI goes first on everything.
Not "try AI when it seems useful." Not "use AI for the boring stuff." Every deliverable. Every workflow. AI first, human review after.
They wanted to stop guessing and find out what actually held up under real conditions.
Here's what they learned.
What You Can Actually Trust
Outreach automation
They built a workflow that takes their segmented prospect lists, generates tailored outreach briefs, and sends the emails — using Tasklet, Claude, and a Gmail Apps Script. The system knows which types of sites match which clients and industries. It runs without hand-holding.
This works. Not because AI is writing brilliant cold emails, but because the system handles volume and personalization at a level no human on a small team could sustain. The quality is good enough, the reach is wide enough, and it frees the humans for relationship work.
First drafts for high-volume content
Listicles, review posts, comparison articles — if you can hand the AI a detailed product brief, it can produce a keyword-rich first draft that's roughly 70% of the way there.
The key word is "first." The 30% humans add is what makes the difference between content that ranks and content that sounds like every other content. But 70% done is 70% done.
Proposals, SOPs, and client-facing decks
This is where they built their most impressive workflow. Meeting notes go into the CRM. Account data and KPIs pull from dashboards. A template feeds Claude to assemble a first draft. That draft goes to a tool called Alai to become a branded deck.
Weekly client reviews. Sales proposals. Onboarding SOPs. All running through this pipeline.
The output isn't perfect. But it's consistent, it's fast, and it frees senior team members from document assembly — which is time they now spend on strategy.
What You Can't Trust
Final content pieces
This is where they push back hard on the "full funnel AI content" crowd that's all over LinkedIn and Reddit right now.
Their finding: "The content feels like every other content out there almost 80% of the time."
That's not a small problem. In a world where your audience is increasingly alert to AI-generated writing — where people are developing instincts for "AI slop" the same way they learned to spot spam — content that sounds like everyone else's content is content that doesn't get read.
AI can do the scaffolding. It can't do the soul.
Strategic thinking
"AI is good at looking at data and forming opinions based on them — for takeaways or learnings, that's good. But actually brainstorming and thinking next steps is still done by my team via their expertise."
This is the distinction that gets blurred in the hype: AI can analyze. It can summarize patterns. It can reflect data back to you in organized ways.
It cannot replace the judgment that comes from knowing a specific client, a specific market, a specific moment. That judgment is still human. It's still what you're selling.
The Real Lesson
The team that ran this experiment didn't conclude that AI is overhyped. They concluded something more useful: AI has a precise jurisdiction.
Inside that jurisdiction, it's reliable, scalable, and freeing. Outside it, it quietly degrades your work — often without anyone noticing until the results stop coming.
Most businesses haven't mapped that jurisdiction for themselves. They're either using AI for everything (and wondering why the output feels flat) or avoiding it for everything (and falling behind on volume and speed).
The six-month AI-first experiment isn't a methodology most teams can copy directly. But the output — a real list of what holds up and what doesn't — is worth building for your own operation.
Not based on what the LinkedIn gurus say. Based on what you actually test, in your actual workflow, with your actual clients.
That map is the most valuable AI asset you can build right now.
Want the workflow breakdown for proposals and SOPs? Reply to this email and I'll dig into the specifics.