| |||
| |||
| |||
|
| Sponsors |
| |||||||||||||||||||||||
| Quick takes |
| |||||||||||
| Deep dive |
Agent jobs in marketing and sales

TL;DR |
Agents are coming to GTM, and you can’t transform everything at once. Brendan Short (The Signal) and I built the Agent Jobs Matrix to help you decide where to deploy agents and in what order. It scores every agent job on two things: how much the agent does on its own, and how hard it is to deploy. Most jobs move through the same four boxes in the same order. Today I walk through the approach, apply it to paid media agents, and map 24 other agent jobs.
Agents are coming to GTM.
Most marketing and sales teams are in the early phases of this transition.
For any GTM leader, the questions are now:
Where do I deploy agents?
How do I sequence agent rollouts?
Which vendors and resources will help?
Whether you are building or buying, you need a way to prioritize. You can’t transform everything all at once.
To help you in this journey, I partnered with Brendan Short (The Signal) to develop a new framework.
We call it the Agent Jobs Matrix. It’s available as a free interactive resource here.
Four jobs to start: Paid Media, Deal Execution, Outbound SDR, Inbound SDR. Each one shows a “ladder” to go from low-risk “assistants” to higher-complexity, autonomous agents. We want to keep expanding with your feedback.
We’ve also started mapping the vendors and tools.
In today’s newsletter I’ll walk through the general approach, apply it to Paid Media agents, and map 24 other interesting agent jobs. Brendan is going deep on Deal Execution agents in his newsletter. I highly recommend reading his post.
The Framework
Every agent job gets two scores.
Agent autonomy is how much of the job the agent does on its own. At the low end, you ask and it answers. At the high end, it works toward a goal within a budget.
Deployment complexity is how hard it is to roll out. There are a few concepts inside “complexity”: single-player to multiplayer, low stakes to high stakes, and reading data to writing it.

The net combination is four boxes:
Assists the person (bottom left). It answers, reports, drafts and recommends when you ask. You find what works and do the work yourself.
Scales the person (top left). Proactive recommendations and ongoing tasks. What works runs on its own, after every call or every night, and suggests your next move.
Scales the team (bottom right). What works for one person becomes skills and workflows the whole team uses.
Owns the goal (top right). It plans, acts and adjusts on its own, up to owning the goal and the budget.
While I was texting with Brendan, he had a nice riff on this progression: “What does your best rep send in their email after a sales call that works? Great. Now let’s have that generated after every call. That works? Great, now let’s have it for your whole team.”
In marketing, a demand gen marketer comes up with ad ideas and iterates on the best ones with AI. Once they have patterns that work, they let the agent generate variations every night. When that works, they can expand it to other marketers on the team with shared agents. As the whole team gets leverage, they find the parts that can be handed off to the agent.
Each box builds on the last. You learn, the agents improve, and over time the company gets leverage. This is the promise and the dream.
Despite all the performative AI-pilled talk on X, I know most teams are still starting at the bottom left.
The matrix is missing one key thing. It doesn’t prioritize jobs for IMPACT! We’re looking at this from the outside and don’t know your situation. So you’ll need to bring that third dimension.
How to apply it: Paid Media agents
Last week I wrote about paid media agents. Hot right now. Two paths to get one: buy the agent, or build your own and buy the Legos.

This one had a great discussion on LinkedIn, which tells me a lot of people are thinking about it.
So let’s look at Paid Media using the Agent Jobs Matrix. Each dot is a job the agent does in service of the bigger paid media goals. I thought about seven steps:

1.) Draft ads. The agent works through ad ideas with a marketer and writes headline, copy and image variants. Some build a whole campaign as a draft. The marketer edits and launches.
2.) Analyze performance across platforms. You ask why cost per lead jumped on LinkedIn, and it pulls the numbers from every platform and finds the cause. This is the most common way I see paid teams using AI today.
3.) Recommend daily campaign optimizations. Every day it reads results and recommends pauses, budget shifts, bids and new creative, each with its reason. A marketer approves each one. Nobody asked it to look. Now the agent scales the person.
4.) Refresh & remix creatives. Every night it remixes winning ads into fresh variants, swaps them into live ads and retires tired ones, inside one marketer’s campaigns. Nobody approves each swap.
5.) Refresh audiences & target lists across platforms. It rebuilds audiences from CRM and intent data, applies suppressions and syncs them to every ad platform. Marketing and sales both rely on it, and a missed suppression targets your customers. Now the agent scales the team.
I’ve built a version of this with Claude Code. In a few days, I got audiences syncing and refreshes to run autonomously with my night shift agents. Every night they rebuild my audiences and update them via Primer’s API. Those go into LinkedIn matched audiences and other ad platforms as I expand (Primer handles the matching and syncing account lists to all major channels).

6.) Tune bids and budgets. It changes bids, budgets and pacing on its own, inside the budget and limits the owner set. Vector’s Bid Agent adjusts LinkedIn bids every 15 minutes. Synter’s MCP runs pauses and budget changes by default. This is the top right box: the agent does the work itself.
7.) Own campaign goals & budget. This is the “let it cook” version that the most ambitious builders and vendors will go after. It plans, launches and shifts spend to hit the goal.
What this means for the build or buy choice:
If you’re buying, check which step the vendor really sells. “Agent” can mean step 1 or step 7. Ask what the agent does without a person, and what it does with one.
If you’re building, everything below the top step is within reach now. My own setup is a mix. As a team of one, my audiences scale the person. In a B2B team they’d scale the team, because sales shares the list.
Start where a person still approves. Then take away one approval at a time.
The page lists the vendors for each job, split into buy the agent and build your own.
More GTM agent jobs I’m excited about
Paid media is one job family. Here are 24 more, grouped by the box.
Assists the person.
Research a new segment.
Draft blogs, emails and landing pages.
Draft custom decks for AEs.
Turn one recording into a month of content.
Answer “what drove pipeline?”
Rewrite pages so AI search cites us.
Generate a forecast for sales managers.
Scales the person.
Catch competitor moves overnight.
Write the weekly marketing readout.
Find and fix broken forms, links and tracking.
Brief reps before every first meeting.
Keep competitive battlecards fresh.
Clip & remix content for each channel after publishing.
Scales the team.
Check every asset for brand, positioning and legal.
Reopen closed-lost deals when something changes.
Turn one brief into a full launch across teams.
Follow up every event lead in a day.
Review and route inbound leads.
Build account lists from buying signals.
Run live roleplays for sales training.
Owns the goal.
Qualify & book inbound leads (AI SDR).
Personalize onboarding for every new customer.
Run outbound as an AI SDR.
Run ABM plays end to end.
Next steps
This framework is a work in progress. Brendan and I want to keep evolving and expanding this concept.
Everything is live here for you to explore, add vendors, and poke holes.
I’d love to hear what you think. I read all replies.





