Let’s get something out of the way up front: we use AI every day.
If you’re reading a series of blog posts from a marketing agency about why clients still need agencies in the age of AI, it would be reasonable to assume we’re anti-AI. We’re not. We’re anti–lazy AI. We’re anti-replacing-judgment-with-a-prompt. And we’re deeply, specifically for using these tools the way they should be used — as leverage for experienced humans, not a substitute for them.
So here, we’re pulling back the curtain a little. This is how we actually use AI inside a boutique marketing and communications firm today.
The Core Principle
We have one rule that governs every AI decision we make for our clients’ work. It’s short:
AI is allowed to help us produce faster. It is not allowed to decide what to produce.
That single sentence does most of the heavy lifting. Let us explain what it means in practice.
“Decide what to produce” is strategy. It’s judgment. It’s the call about what story to tell, which audience to tell it to, when to tell it, and what outcome we’re trying to drive. That’s what a senior human does, based on everything they know about the client, the market, and their own accumulated experience.
“Produce faster” is execution. It’s turning a finished strategy into drafts, variants, transcripts, briefs, analyses, and first-pass deliverables that a human then edits, improves, and signs off on. That’s where AI shines. That’s where we lean on it hard.
When you get that line right, AI makes an experienced agency faster, sharper, and more valuable per dollar. When you get it wrong — when you let AI drive strategy — you get generic, forgettable, and sometimes actively harmful marketing.
Where We Use AI (And Where We Don’t)
Here’s the short version of our internal playbook.
Where we use AI heavily:
- First-draft copy for emails, posts, and landing pages (always edited by a senior hand)
- Research synthesis — turning ten articles into a briefing doc
- Brainstorming angles, headlines, and creative directions
- Restructuring and formatting (turning notes into outlines, turning outlines into drafts)
- Summarizing transcripts of interviews, calls, and meetings
- Code and data work for marketing automation, reporting, and analysis
- Translating and proofreading
- Generating variants of ad copy for A/B testing
Where we deliberately don’t:
- The strategic decision about what the campaign is actually for
- The selection of the specific journalist, influencer, or partner for an outreach effort
- The final quality check on anything going to a client or the public
- Crisis communications drafts that will be used in real time
- Anything that requires reading a specific person or specific room
- The client relationship itself — we pick up the phone
That second list is where the value is. That’s what the “agency premium” pays for. AI doesn’t touch any of it.
An Example From Last Week
We were working on a campaign for a B2B client targeting finance leaders at mid-sized companies. The brief called for a thought leadership article.
Step 1 (human): We had a forty-five-minute conversation with the client’s CEO about what he actually believed and what was frustrating him about his industry. No AI. Just a conversation, a notepad, and twenty-nine years of knowing which questions to ask.
Step 2 (AI-assisted): We transcribed the conversation and used AI to pull out the three or four genuinely contrarian arguments he was making — the ones that would actually make a finance leader stop scrolling. This took about ten minutes. In 2019, this would have taken half a day.
Step 3 (human): We picked the one argument that was most differentiated, most true, and most relevant to what was happening in the market right now. That’s a judgment call. AI was not in the room for that call.
Step 4 (AI-assisted): AI drafted three structural approaches to the article based on the chosen argument. Again, faster than human-only drafting would be.
Step 5 (human): A senior writer took the best structural approach, rewrote it almost entirely in the client’s voice, added examples AI would never have known about because they’re specific to the client’s career, and tightened the argument until it was sharp.
Step 6 (human, again): Before we pitched it anywhere, a different senior person read it cold and flagged two sentences that were going to land wrong with the specific audience we were targeting. We changed them.
That’s AI-assisted agency work. Total elapsed time: about a third of what the same project would have taken five years ago. Quality: higher than it would have been five years ago, because we spent the saved time on judgment instead of typing.
Why This Is Hard to Do Well
You might read the above and think, “Okay, anyone could do this.” You’d be half right.
Anyone with a subscription could do this. Very few non-professionals actually do.
What trips people up isn’t the tooling. It’s knowing where the human judgment is supposed to go. Without experience, the temptation is to let AI do the whole thing — including the parts where a seasoned professional would have intervened. The output still looks fine. It just doesn’t land. And the person producing it often has no idea why.
This is the trap we see in-house teams fall into constantly. The AI gets smarter. The outputs look more polished. The results get quietly, steadily worse — because nobody is left in the loop who knows which outputs are actually worth sending.
What This Means for Clients
The “agency vs. AI” framing is the wrong one. The right framing is:
- Untrained AI producing generic content
- Experienced humans not using AI at all
- Experienced humans using AI well
Option 1 gets you average work fast and cheap. Option 2 gets you great work slow and expensive. Option 3 gets you great work fast and at a reasonable price — because we can spend our hours on strategy and quality instead of typing.
That’s the option we’re selling. That’s what a modern boutique agency should actually be.
The Bottom Line
We use AI. Every day. We use it well. And we use it specifically because it lets us spend more of our time on the things AI can’t do — the judgment calls, the relationships, the accountability, the quality control.
Anyone who tells you their agency “doesn’t use AI” these days is either behind the curve or not being straight with you. And anyone who tells you AI replaces the need for experienced humans has never sat in the chair when something actually went wrong.
The right answer is the combination. We’re pretty sure it’s going to stay the right answer for a long time.
Up next in the series: why your brand voice can’t come from a prompt. In the meantime, if you want an agency that’s honest about how it uses AI — and what it doesn’t use AI for — come say hi.
Frequently Asked Questions
Does TSN use AI in client work?
Yes, every day — for first drafts, research synthesis, brainstorming, formatting, transcription, and variant testing. The rule is simple: AI helps us produce faster, but it never decides what to produce.
Where do you deliberately not use AI?
On the strategic call about what a campaign is for, the choice of a specific journalist or partner, real-time crisis drafts, the final quality check, and the client relationship itself. That's where the value lives.
Why can't an in-house team just do the same thing?
The tooling isn't the hard part — knowing where human judgment belongs is. Without experience, teams let AI do the whole thing, the output looks fine, and results quietly get worse because no one knows which outputs are worth sending.
Isn't an agency that uses AI just cutting corners?
Used well, it's the opposite. The time saved on production gets reinvested in strategy, relationships, and quality control — so you get great work faster and at a reasonable price.
Last updated August 20, 2026