AI agents are the hot new product in AI these days – Satya Nadella says agents will automate much of today’s knowledge work, and Sam Altman is planning to charge $20,000 / month for specialized agents.

But for all the buzz about them, most organizations we talk to aren’t actually using agents yet.

So we sat down with David Freas, Managing Partner at Supermoon, a small, independent brand agency that’s started building out AI agent roles internally, to learn how agents are supercharging their output.

Supermoon’s use case: AI agents for new business outreach

Supermoon is still a team of two, and David and his partner need to focus their time on delivering a great product to clients, which means they don’t have a lot of time to focus on business development.

AI agents can significantly increase their reach and manpower (if they work) by creating target lists, monitoring social media, and identifying companies in need of agency support.

To design Supermoon’s agents, David uses Lutra.ai – an AI automation platform that connects to other applications like Gmail, Google Sheets, Airtable, and Slack to help complete actions. It allows users to define logic and prompt sequences, which are then saved as "playbooks" and function as AI agents capable of executing full tasks.

So far, Supermoon has focused on three types of agents:

• Data enrichment agent: This agent aggregates information on key decision makers in companies Supermoon is targeting outreach to. It pulls from multiple sources, including LinkedIn, company websites, and public databases, and compiles it into a Google Sheet.
• Social media monitoring agent: This agent scans social media platforms for posts containing specific keywords relevant to brand strategy, marketing trends, or company transitions – or posts from a specific set of people.
• Transformational moment agent: This agent is still in the design stages – but its job is to identify companies that are likely to need a brand refresh or repositioning, such as orgs that have recently hired a new CMO or completed a funding round.

The point of these agents is to take over the manual work required to find more sales-ready leads. David estimates these agents save him at least 4 hours of work a week – and that the third will save him an additional 4, once it’s working.

They’re not perfect – AI still hallucinates, so sometimes they’ll make up a contact at a company.

And they don’t replace the need for human connection in a first contact email (we’ve all seen BDR emails that are clearly generated by AI). But David has far more time to focus on the quality of those touches now.

Preparing to work with client-side AI agents

Supermoon fully expects that their clients will be using agents soon, so they’re preparing deliverables (like brand guidelines, style guides, and tone and voice documents) to be used by AI agents, not just humans.

Here’s how they’re doing it:

1. Feeding a traditional deliverable to an AI model and testing its output. Supermoon gives a deliverable like a brand guide to an AI model, and asks the AI to generate posters, emails, social posts, etc. in the brand’s style, voice and tone. The purpose is to test how effective the resources are at guiding the AI – and right now, he says they get there about 80% of the time.
   
   
2. Experimenting with deeper training techniques to capture company character. To reduce time spent giving generic feedback, Supermoon is embedding specific brand traits into the AI’s training. So instead of giving feedback like “make this friendlier,” Supermoon can say, “This should echo our brand voice of ‘approachable nerd’”, a concept the AI has been trained on.
   
   
3. Creating an AI-to-human-to-AI feedback loop. Consistency is a big sticking point in quality of AI output. Supermoon’s human partners review and refine AI’s output, then feed those edits back to the AI along with the reasons for the corrections..

The future of brand agencies in an AI-powered world

Looking ahead, David predicts 3 significant shifts in the brand agency landscape as AI agents become more advanced and more common:

1. Production design work and creation will be the first thing to be replaced. Agencies that rely heavily on basic graphic design, layout work, and repetitive brand asset creation may struggle – and younger designers and marketers will lose traditional learning opportunities.
   
   
2. The irreplaceable parts of branding will become more crucial. Humans need to focus on where they excel over AI: Cultural insight, emotional storytelling, and trendspotting. It’s these more unpredictable elements that humans can provide, and that make brands really stand out.
   
   David says: “If you think about design as a logical path towards ‘this is what people want’, anybody – including AI – can go through the process to figure out ‘this is what a certain set of people want’.
   
   It’s different and interesting to say, ‘How do we do this in a unique and human way that will stand out?’ And in some cases, that means including elements that are illogical or have friction. It’s why Spock couldn’t quite grasp humanity – we need Captain Kirks too.”
   
   
3. Design teams need to be trained in AI-driven workflows. Agencies and brand teams will need humans who work at the intersection of art and AI languages. So teams should start learning how to work with AI now, so they’ll be ready when really buttoned-up agents hit the market.