AI for Affiliate Managers: From First Chat to First Prospecting Agent
AI for affiliate managers made simple: pick an LLM, build skills, then build a prospecting agent that finds partners in 30 minutes. Watch both episodes. |
You know AI could save you hours. You’ve also opened ChatGPT, asked it something, got a confident-sounding answer and closed the tab thinking “okay, now what?”
This guide covers AI for affiliate managers from the ground up. We start with which tool to pick and how to make it sound like you, then build a working agent that finds, scores and drafts outreach to affiliate partners in about 30 minutes.
It comes from a two-part Affiliate Nerd Out series I recorded with Keegan Lamar, President of ViaOps. Keegan set up Claude for our whole exec team at XPFlow, and he gave me my first real lessons in AI. Both episodes are embedded below.
The short version: AI for affiliate managers means handing the repeatable parts of partner work (research, scoring, first-draft outreach, follow-ups) to AI tools, while you stay the human who reviews the output and builds the relationships.
In this post
Where should an affiliate manager start with AI?
Which AI tool is best for affiliate work?
How do you make AI sound like you?
What is an AI skill, and why does it matter?
How do you build an AI affiliate prospecting agent?
How much of your work should AI handle?
Wrap it up, Dustin
FAQ
Where should an affiliate manager start with AI?
Start by picking one of the big three and talking to it like a curious person, not a programmer. Keegan’s advice was simple: know the big three, pick one (or all of them), and ask it questions.
“I would argue that’s probably the most important step, is just being willing to try.” - Keegan Lamar
Treat it like a glorified Google search at first. Keegan snapped a photo of a tree in his neighbor’s yard and asked what it was. AI told him the species, that it was already full grown, and that it wasn’t going to take over his yard. Low stakes, but that’s how you build AI literacy.
You’re not alone in this either. McKinsey’s State of AI survey found that most organizations now use AI in at least one business function, but most are still stuck in pilot mode. The gap is people who actually use it every day.
Which AI tool is best for affiliate work?
Claude is my pick for business work because it pushes back instead of flattering you. Ask all three the same question and you’ll feel the difference fast.
Tool | Made by | What it feels like | Where it shines |
|---|---|---|---|
ChatGPT | OpenAI | Tells you every idea is a good idea | Image creation (that’s mostly what I use it for now) |
Claude (Chat, Cowork, Code) | Anthropic | Realistic, will challenge you and warn you | Business work, writing, planning, building agents |
Gemini | Consumer-friendly, has Google built in | Quick lookups and image/video play like Nano Banana |
Keegan explained that this is by design. ChatGPT was originally built to earn positive feedback, which led to some famous hallucinations (one guy thought he’d discovered new math, and it was basic algebra). Claude is built to ground things. It once told Keegan’s frustrated brother, “Hold on there, cowboy.” Gemini pulls straight from Google search, so its answers are only as good as whatever ranks.
Dustin’s Choice: Claude. It’s the most personable, it learns how I work, and it doesn’t blow smoke at me. For planning, organizing and building agents, nothing’s beating it.
What this looks like for an affiliate manager: paste a prospect’s URL into Claude before a call and ask for a breakdown of the company and its online footprint. That used to be 15 minutes of prep for me.
How do you make AI sound like you?
Give it context and a master prompt: who you are, what role it plays, and what it should never do. A fresh chat knows nothing about you. A project (or folder, depending on the tool) keeps context across every conversation inside it.
A master prompt tells the AI its persona and its job. Keegan built one for me that basically said, “You are Dustin’s assistant. This is what Dustin cares about. This is the lens you look through.” Then you connect your docs, email and calendar so it has real information to work from.
The part most people miss: tell it what role not to play. Without rules, AI wanders off and guesses. With clear don’ts, it feels like it’s actually listening.
What this looks like for an affiliate manager: create one project per brand you manage, drop in the brand brief, commission structure and ideal partner profile, and every chat starts already knowing the program.
What is an AI skill, and why does it matter?
A skill is a repeatable process you teach AI once, so it does the job your way every time. Finish a task with AI (Keegan’s example was building an invoice), then tell it: “Turn what we just did into a skill.”
Skills are the building blocks. Group a few together and you have an agent: “You’re my marketing agent, here are your skills, here’s what you handle.” Now you’re delegating whole processes, not single prompts.
You don’t have to write them all from scratch. GitHub has free skill libraries. Search “awesome skills,” or look at Corey Haines’ marketingskills repo, which covers copywriting, SEO, cold email and more. You can even ask Claude to comb through a library and pick the skills that fit how you work.
What this looks like for an affiliate manager: a recruitment email skill that knows your voice, your offer and your follow-up cadence, so you never re-explain it.
Stuck on the first step of recruiting? Grab the Affiliate Partner Recruitment Playbook, the same framework we use to find, vet and reach niche partners. [Get the Recruitment Playbook] [PLACEHOLDER LINK]
How do you build an AI affiliate prospecting agent?
Make a project, pull four marketing skills into a plugin, feed it a brand brief and ideal partner profile, then tell it to scout. In AI 201, Keegan built one live in Claude Cowork. Start to finish took under 30 minutes.
Here’s the build, step by step:
Create a project in Claude Cowork. His was called “Brand Scout.” The project instructions act like a project manager telling everything how to work together. Keegan didn’t write them himself. He had AI write them.
Pick your model. Keegan’s take: Sonnet is the daily workhorse and the safest bet for this.
Grab four skills from the marketingskills repo: product marketing, prospecting, cold email and influencer marketing.
Bundle them into a plugin called Partner Scout, with a /scout command. A plugin is just a group of skills plus instructions, which is basically an agent.
Add two files: a brand brief and an ideal partner profile (what I call partner personas). His demo brand was Carnwell, a made-up outdoor coffee company.
Run it. “Find about 10 ideal partners and tell me why we should work with them.”
About six minutes later, it came back with 13 ranked partners, each with a fit score out of 10 and a pursue-or-pass verdict. It found gear review sites like Treeline Review and GearLab, personal blogs and YouTube backpacking channels. Then it drafted outreach for every “pursue” partner, using the creator’s first name and pointing to their actual camping coffee maker roundup.
From there you can tell it to drop the drafts in Gmail, or have it build a shareable breakdown doc for your team.
On permissions, Keegan’s advice is to start skeptical. Click “allow once” until the project knows you, then loosen up. Skip all approvals too early and you might come back after an hour to a mess you have to redo.
Now, the honest part. This build is the kindergartner. As Keegan put it, Alfie is the middle schooler all the kids look up to. In the time this agent found 13 partners, Alfie likely found 200 better ones, customized to your exact brand, because we’ve been building and training that model for over a year. If you want the full grind handled, Alfie sends 1,000+ personalized outreach emails a month and saves affiliate managers an average of 14 hours a week.
Want Keegan’s exact build? Head to dustinhowes.com/keegan for the project instructions, the skills, the plugin, every prompt he used, the brand brief and partner profile templates, and a prompt to put the whole thing on a schedule.
How much of your work should AI handle?
Aim for about 70 percent now, with you reviewing as the human in the loop, then push toward 80 and 90. That’s Keegan’s benchmark for time spent typing on a computer.
How you get there:
Notice when you say the same thing twice. That’s your next skill.
Schedule tasks. An end-of-day email rundown, or a morning brief that lists the emails to answer, the follow-ups due and today’s meetings.
Chain agents. One clarifies, one executes, one reviews, one delivers. You police the output.
What this looks like for an affiliate manager: a morning brief that flags new partner applications, partners who went quiet, and follow-ups from yesterday’s calls.
Two more habits from the episodes. Be polite to your AI. Clear, kind feedback gets better results, and I don’t want to lose that muscle with actual humans. And remember why you’re doing this. Big deals still get made between people. AI just frees you up for what Keegan calls the “kissing babies and shaking hands” side of business. We wrote more about that idea in Automate the Ordinary. Elevate the Human. on the XPFlow blog, and Chief of Chaos makes the case that AI won’t replace affiliate managers, it will expose or supercharge them.
Wrap it up, Dustin
Pick a tool (I’d pick Claude). Give it a project and a master prompt. Turn your repeat work into skills. Then build your first scout. Even the $20-a-month Claude plan gets you Cowork, which is all you need for this build. Keegan pays $200 a month and says it pays for itself.
If you’d rather skip the kindergarten phase, see how Alfie finds and recruits niche partners for you. And if you want a hand setting up AI across your team, connect with Keegan at ViaOps (he’ll help you tune your prompts if you mention the episode).
Ready to recruit smarter? Download the Affiliate Partner Recruitment Playbook and pair it with your new scout, or with Alfie. [Get the Recruitment Playbook] [PLACEHOLDER LINK]
Built your first agent? Tell me what it found. And if you’re still deciding where to look for partners, start with my guide on how to find affiliates.
Keep on recruiting.
FAQ
What is AI for affiliate managers? It means using tools like Claude, ChatGPT or Gemini to handle repeatable affiliate work: partner research, scoring, outreach drafts, follow-ups and reporting. The manager reviews the output and owns the relationships.
Which AI is best for affiliate managers? Claude is my pick for business work because it pushes back instead of agreeing with everything. ChatGPT is great for images, and Gemini is handy for quick Google-powered lookups.
What is an AI skill? A skill is a saved, repeatable process that tells AI exactly how to do a task your way, like writing a recruitment email. Several skills grouped together make an agent.
Can I build an AI prospecting agent without coding? Yes. In Claude Cowork you create a project, add skills from a free library, bundle them into a plugin and give it a brand brief and ideal partner profile. Keegan built one live in under 30 minutes.
Do I need a paid AI plan? The free Claude plan gives you chat. The roughly $20-a-month plan unlocks Claude Cowork, which you need to build the agent shown in AI 201.
Will AI replace affiliate managers? No. AI takes over the machine work so managers can spend more time building partner relationships, which is where deals actually get done.
How is a DIY agent different from Alfie? A DIY agent is a fast, rough start: about a dozen partners from a quick web search. Alfie is purpose-built for affiliate recruitment. It finds far more partners, tailors them to your brand and runs outreach at scale.
