There's a strange dance happening in marketing right now. Half the industry is using AI for everything and pretending they're not. The other half is pretending they use AI when really a junior copywriter is still doing it all manually. Both camps are lying to their clients, and it's getting tiresome.
So here's what AA2 actually does with AI. No hedging. No vague references to "proprietary technology." Just the truth.
AI is involved in roughly 60% of the operational work that happens here. That's a number I've never shared publicly before, and I'm sharing it because I think clients deserve honesty about what they're paying for.
Data research and enrichment. When we build a prospect database through the Lead Engine, AI handles the initial research: identifying companies that match the ideal customer profile, finding the right contacts within those companies, cross-referencing information across multiple sources. A data build that would take a human researcher three weeks takes the automated pipeline four hours. The output still gets reviewed by me. Every list, every contact, every enrichment batch gets human eyes before it goes anywhere.
Email draft generation. First drafts of email campaigns are AI-generated. I feed in the client brief, their brand voice guidelines, the campaign objective, the target audience profile, and performance data from previous campaigns. The AI produces a draft. I then rewrite it. Sometimes I rewrite 20%. Sometimes 60%. It depends on how well the AI captured the nuance. The AI is good at structure and flow. It's less good at the specific turn of phrase that makes an email sound like a human being wrote it for another human being.
Content outlining. Blog posts, articles, white papers: the outline is often AI-assisted. The AI identifies relevant angles, suggests structure, pulls in data points. The actual writing is mine or, for client content, written to their specific voice. AI doesn't write the jokes. AI doesn't write the opinions. AI definitely didn't write this sentence.
Report compilation. Monthly campaign reports involve pulling data from multiple sources, calculating performance metrics, identifying trends, and presenting them clearly. AI automates the data pulling and initial analysis. I write the narrative. The bit that says "here's what this means for your business and here's what we should do about it" comes from a person who knows your business, not an algorithm processing numbers.
Dashboard automation. Live client dashboards update automatically. The workflows that pull data, calculate metrics, and flag anomalies run without human intervention. When something needs attention, I get an alert. The Intelligence Layer is largely automated by design.
Workflow building. The automation workflows themselves, the email sequences, the follow-up triggers, the pipeline updates, are built with AI assistance. AI generates the initial workflow structure. I refine, test, and deploy. Every workflow gets tested manually before it touches live data.
Competitor research. When we run a competitor scan for a client through the Client Blueprint process, AI handles the initial data gathering: scraping public information, identifying competitor positioning, pulling pricing where available, cataloguing their content output. The analysis, the "so what does this mean for your business," is mine. AI can tell you what a competitor is doing. It can't tell you why it matters to your specific situation.
Social content scheduling. Monthly social content calendars get drafted with AI assistance. The themes, the hooks, the platform-specific formatting, AI produces the first version. I review every post, adjust the voice, check the facts, and approve the schedule. Nothing goes live without that review step. This applies across every service in the Content Engine.
Honesty about AI means being honest about its failures too. Here are things I've tried and stopped doing.
AI-generated strategy documents. Early on, I experimented with using AI to draft client strategy sections. The output was grammatically perfect and substantively empty. It said things that sounded strategic but contained no actual strategic thinking. Generic observations about "growing markets" and "evolving buyer expectations" that could apply to any business in any sector. I binned the approach within a month.
AI-written cold email subject lines. The data here was clear. AI-generated subject lines consistently underperformed human-written ones by 15% to 20% on open rates. AI subject lines tend toward the generic and the safe. The subject lines that actually work are often slightly unexpected, slightly specific, and occasionally a bit odd. AI doesn't do odd well.
AI for client tone of voice. Every client has a distinct voice. Some are formal, some are casual, some are technical. AI can approximate tone, but it struggles with the subtle differences between "professional and warm" and "professional and direct." Those are different voices, and the distinction matters when you're writing to a client's audience. I still write all tone guidelines manually and use them to guide AI output, not the other way around.
These failures taught me something important: AI is best at tasks with clear inputs and measurable outputs. Data processing, verification, scheduling, formatting. It's worst at tasks that require taste, judgement, or the kind of pattern recognition that comes from years of experience in a specific field.
This list matters more than the one above.
Strategy. Every client strategy, every campaign plan, every market analysis is developed by me. AI can summarise data. It can't understand a client's market position, their competitive dynamics, their founder's ambitions, or the political reality of their sales team. Strategy requires judgement, experience, and the kind of contextual understanding that only comes from sitting in a room (or on a call) with another person and actually listening.
I wrote about the Client Blueprint process earlier in this column. That blueprint is a human document, built from human conversation, informed by human experience. AI doesn't attend the discovery call. AI doesn't ask the follow-up question when a client says "well, it's complicated." AI doesn't notice when a client's body language changes when you mention their biggest competitor.
Telemarketing conversations. Every phone call made on behalf of AA2 clients is a real person talking to another real person. As I wrote in a recent article about outsourced calling, the value of the First Conversation is in the caller's experience, instinct, and ability to respond in real time. AI can tell you who to call. It can't have the conversation.
Client relationship management. I take the strategy calls. I handle the difficult conversations. I deliver the bad news when a campaign underperforms. I celebrate the wins. If you're a client, the person you speak to is always me. Not a chatbot, not an AI assistant, not a junior account manager reading from notes.
Final sign-off. Nothing leaves AA2 without my approval. Every email, every data build, every report, every piece of content gets reviewed by a human before it reaches a client or their prospects. AI proposes. I decide.
Pricing and commercial decisions. What to charge, how to structure a proposal, when to offer a discount, when to walk away from a bad-fit prospect. These are judgement calls that require understanding the full context of a relationship, not just the data. AI has no business anywhere near these decisions.
Difficult conversations. When a campaign doesn't deliver what was expected, when timelines slip, when a client's expectations need recalibrating, those conversations happen person to person. They require empathy, honesty, and often a willingness to sit with uncomfortable silence while the other person processes what you've said. AI can't do any of that. Nor should it.
The reason I'm writing this is because the alternative is worse. Agencies that hide their AI usage are setting themselves up for a trust problem. When clients eventually discover (and they will) that the "bespoke strategy" was generated by ChatGPT with their company name pasted in, the relationship is over.
Equally, agencies that pretend they don't use AI are either lying or inefficient. If you're still manually compiling campaign reports from spreadsheets in 2026, you're not protecting quality. You're wasting time that could be spent on actual thinking.
The honest position is simple: AI is a tool. Like any tool, it's useful for some things and useless for others. Using it well means knowing the difference.
There's a third camp that bothers me more than the other two: agencies that use AI badly and don't even realise it. They paste a brief into a chatbot, copy the output into a document, send it to the client, and genuinely believe they've done the work. The client receives something that reads like it was written by committee, because it was written by a committee of one person and an algorithm with no industry knowledge, no understanding of the audience, and no accountability for the outcome.
Using AI responsibly takes more effort than not using it at all. You need to build the right inputs, review the outputs critically, and know when to throw the output away and start from scratch. That's not a time-saving shortcut. It's a different way of working that happens to be faster at the right tasks and useless at the wrong ones.
The line between what AI should and shouldn't do isn't technical. It's human. AI should handle the work that doesn't require human judgement: data processing, draft generation, report compilation, workflow automation. Humans should handle everything that requires understanding, empathy, experience, and accountability.
When a client asks me "do you use AI?", the answer is yes, extensively, and here's exactly how. That conversation has never cost me a client. It's won me several. Because the alternative, an agency that either hides its methods or can't explain them, is a much bigger red flag than one that uses technology openly and effectively.
The question isn't whether your marketing agency uses AI. It's whether they'll tell you about it. And whether, when you ask what a human being actually does, they have a clear answer.
At AA2, the answer is: the thinking.
I suspect that within two years, every serious marketing agency will need to have this conversation with their clients. The technology is moving too fast for anyone to pretend it doesn't exist. The agencies that get ahead of it, that establish trust through transparency now, will be the ones that survive the inevitable moment when clients start asking harder questions.
The agencies that hid behind "proprietary processes" and "bespoke methodologies" will have a much harder conversation ahead of them. Because the question won't be "do you use AI?" It'll be "why didn't you tell us?"
Better to have the honest conversation now.
Martin Dugan, AA2
