At this year’s impact.com’s Partnerships Experience (iPX26), CEO Dave Yovanno sat down for a fireside chat with Allie K. Miller, CEO of Open Machine and AI advisor to Fortune 500 companies. He opened by asking a room of partnership managers to raise their hands if they used ChatGPT daily. Nearly every hand went up.
Then Yovanno asked who was running an AI agent harness, tools like Claude Code or Codex. The hands that stayed up dropped fast, pointing to the essential AI skills marketing teams still haven’t built.
Most partnership managers in that room already consider themselves proficient AI users. They’re in AI chats daily: drafting partner emails, summarizing calls, running quick research.
But that kind of daily use caps out at a modest productivity gain. The practitioners pulling ahead have built a system that gives AI persistent access to their program, partners, and goals.
While you’re using a tool, the marketers pulling ahead are running a system
Daily AI use and proficiency aren’t the same thing—and most partnership managers can’t tell which side of that line they’re on. You’re still thinking of AI as a tool if you’ve retyped your top partners’ names into ChatGPT more than once this week.
“If you believe that AI is a tool, you’re still going to gain productivity, but you’re missing this next level of AI as an operating system,” said Miller. “It should act as the substrate of your work. That’s where people are getting 10x gains.”
Better prompting gets more out of the conversation you’re in, but it can’t create the memory that conversation is missing. Every fresh AI chat starts at zero: no memory of your top partners, your program’s current goals, or what came up on last week’s call.
“If you’re doing all your work in ChatGPT, you’re living 2024,” Miller said.
The difference is context: the record of partners, goals, and program history you’re not handing your AI in the first place.
“If you believe that AI is a tool, you’re still going to gain productivity, but you’re missing this next level of AI as an operating system. It should act as the substrate of your work. That’s where people are getting 10x gains.”
– Allie K. Miller, CEO of Open Machine and AI advisor
Building a context system is the essential AI skill most partnership marketers skip
A context system is the piece most partnership managers miss because nobody’s told them it’s a skill worth learning. Even a rough draft outperforms no system at all. Here’s how to build one.
Create a program context doc
A program context doc is the foundational file that tells AI what it needs to know to be useful for your specific program:
- Top partners
- Current goals
- Commission structures
- Key internal stakeholders
- Open problems
Miller keeps an extensive one for her own business—covering what she does, who she serves, and what’s worked and what hasn’t. Yours can be a fraction of that length and still change what AI can do for you. Three bullet points under “top partners” beats a blank doc.
Format also matters less than people assume. “AI loves a good plain text file,” Miller said.
Start with whatever you’ve got: a first-draft context doc that’s mostly bullet points is enough, and it gets updated as your program changes rather than rewritten from scratch.
Capture timely information through voice memos
The program context doc needs to stay current, and typing detailed notes after every partner call isn’t realistic for busy marketers.
“Dictation is one of my most powerful AI hacks. I often go on a walk just dictating thoughts that have been boiling inside my head,” she said. A shorter version works fine: right after the call, cover what you learned, what the partner needs, and what happens next. AI transcribes it and files it.
Dictate a quick note after every call, and you build the kind of real account intelligence that a simple status update can’t capture.
Run your AI system based on goals
Many agent tools let you type ‘/goal,’ hand over an objective, and let the system loop on its own work until it’s done.
“You’re typing out what the actual goal is,” said Miller. “You’re not prompting for 17 steps. It’s going to loop on itself up to 20 times, spinning up secondary agents to review every single round going, ‘did I do it?’ It may even work for hours while you’re sleeping.”
| Dimension | Prompt | Goal |
|---|---|---|
| What you give AI | One task to complete right now | An outcome to work toward, with AI figuring out the steps |
| Time horizon | Immediate (this conversation) | Ongoing or multi-session |
| What AI does | Executes what you ask | Plans steps, loops, self-corrects |
| Best for | Quick answers, drafting | Recurring workflows, complex tasks |
Miller has her own version of this morning briefing: an AI-generated summary of her email, calendar, and meeting recordings. It flags urgent items, catches her up on people she’s meeting with that day, and checks whether her schedule actually lines up with her stated goals. “It shows me if I’m leading a goal-aligned day,” she said.
You could build the same thing scoped to your program—a briefing that surfaces which partners need follow-up, what’s at risk, and what’s on today’s calendar. All of this information is pulled from the context doc and relationship notes already in place.
The same shift works outside a daily briefing, too: instead of asking AI to write an email to a partner, hand it the relationship history, the partnership goal, and the current situation. With that context already in place, drafting the outreach, flagging risks, and suggesting next steps becomes trivial.
“You’re typing out what the actual goal is. You’re not prompting for 17 steps. It’s going to loop on itself up to 20 times, spinning up secondary agents to review every single round going, ‘did I do it?’ It may even work for hours while you’re sleeping.”
– Allie K. Miller, CEO of Open Machine and AI advisor
Where AI agents take a marketer’s context system next
Once you’re working from goals instead of individual prompts, an agent harness is what turns that goal into finished work without you sitting in the chat window the whole time.
Miller’s own setup shows what that structure can look like at the high end. “I have a 34-agent AI workforce,” she said. “I have a chief of staff. The chief of staff has an assistant whose entire job is just keeping memory and context updated. Following the chief of staff, I have six directs… From that there’s a bunch of tiny little sub-agents.”
You don’t need an army of agents to start. One agent running a morning briefing is already a 10x upgrade over prompt-by-prompt ChatGPT use.
Ultimately, the mindset shift matters more than the tool. “We’re moving from single-threaded one-chat-at-a-time to multithreaded multitasking, and we’re moving from reactive to proactive. We’re no longer prompting the AI—AI is prompting us based on those loops and goals we’ve already given it,” Miller said.
“We’re moving from single-threaded one-chat-at-a-time to multithreaded multitasking, and we’re moving from reactive to proactive. We’re no longer prompting the AI—AI is prompting us based on those loops and goals we’ve already given it.”
– Allie K. Miller, CEO of Open Machine and AI advisor
Put your partnership program context to work
The hands that dropped in that ballroom measured something more specific than tool preference: how much context each person’s AI actually had to work with.
Closing that gap starts with the same first move Miller described—a context doc, built this week, covering the partners and goals already living in your head. Build the doc, add a voice-memo habit after your next few partner calls, and shift one recurring task from a prompt to a goal.
The partnership managers doing that now are the ones who’ll stop asking their AI to start from zero every morning. Everyone else is still waiting to get caught by the next hand-raise question.
Your own AI proficiency is one half of the equation.
The other half is how AI is changing whether your brand gets found at all. See how AI is changing how brands get found—and why partnerships are the strategy.
FAQs
The most essential AI skill for partnership managers and affiliate program managers is building a context system: a persistent, structured record of your program, partners, and goals that AI can reference in every conversation.
A program context doc, a habit of dictating quick notes after every partner call, and a shift from single prompts to goal-based tasks are the three concrete pieces that make up that skill.
An AI context system improves affiliate program performance by giving AI persistent access to what actually matters in the program—top partners, current goals, commission structures, and relationship history—instead of starting from zero in every conversation.
Allie K. Miller, CEO of Open Machine and AI advisor, ties the difference in results directly to this: people without a context system report small productivity gains, compared with gains closer to 10 times that for those who build one.
In practice, that means AI can draft partner outreach, prep account briefings, and flag at-risk relationships using real program details instead of generic templates.
Partnership managers can build an AI context system by starting with a single program context doc—a plain-text or markdown file listing top partners, current goals, commission structures, key stakeholders, and open problems.
From there, dictating a short voice memo after partner calls keeps that document current with details that never make it into a CRM.
The final step is shifting from single prompts to goal-based tasks, such as a daily briefing that surfaces what needs attention across the program—also the point at which an AI agent harness becomes useful.
