You can't automate work you haven't done yet

LinkedIn is an echo chamber. It's also where I find clients, software, and the people I build with.

This platform is one of the greatest hack for your career, and while not perfect at all, it’s the closest we have from a community.

So when a prospect asked me to run their influencer program on LinkedIn, targeting PMMs, I said yes.

I spent 6 months running an influencer campaign by hand, with three different waves.

Then I designed the agents that should have done most of it.

Here's the exercise that told me which parts.

The only experience I had with influencer marketing was a few requests from random folks on LinkedIn and a campaign we did on our podcast for a MarTech software.

So when Guideflow reached out, I consumed as much as possible to understand what great influencer marketing looked like:

When I pitched my initial idea, they were pretty happy and loved the direction I wanted to take to target product marketers.

And the results spoke for themselves, because with 3 launches we got:

→ 3x revenue on the line I supported
→ $200K pipeline on $30K spend (6.6x ROAS)
→ Product of the Month on ProductHunt, November 2025

I even got some words of affirmation from my pal Collin:

But, it was pretty hard on my schedule. I had to drop some clients to give it my 110%, and a LOT of manual work was done.

And I feel like a lot of that “boring” work could have been automated.

You can't automate work you haven't done.

I could have come up with a series of automation for the influencer campaign, but since this was the first time I was running this, it would mostly have been guesswork.

If I built agents on day one, I would have automated the wrong things.

The manual work that I did in 6 months gave me the resolution on which tasks repeat, which judgment calls have no rules, and where handoffs break. It doesn’t need to last 6 months, but think about a task or workflow you can do with your eyes closed, but that has very little value to your skillset.

For this mission, it was communicating back & forth with interested influencers AND setting up payments for accounting.

The back & forth happened since most messages came from my profile, but it added unnecessary bandwidth being taken from my schedule. For the payments, we were getting post drafts + payments information via a Slack automation, but we were still missing some information, which created a bottleneck on me always having answer or request information.

The mapping of these “boring” tasks was necessary because I knew there was other stuff that needed to be automated.

After the campaigns, I sat down and listed every task I'd touched across those 6 months. Once this was done, I listed all of them, from beginning to end.

Two columns: how much time this is taking, and what is the value of my time for this task.

Time ⏰: how long the task ate when I did it.

Cash Value💸: the dollar value of my time on that specific work.

High ⏰ and low 💸 became agents. High 💸 stayed mine.

The point was to keep me in my zone of genius and let agents take everything else.

This is coming from Alina Vandeberghe during her agentic workflow workshop at Above the Fold last February.

Keep the work that makes you a damn good operator. Hand the rest to agents.

You can automate some parts or turn them into skills, but you’re the one with taste, nuance, and feeling.

The robots can’t take that from you.

After mapping the 12 tasks necessary to run a campaign from A to Z, the ratio of ⏰time vs. 💸cash value will tell you which ones you should prioritize building agents for.

8 agents fell out of the table, each one as a verdict above.

Slack is the UI. ~30 min a day reads the entire campaign state.

There’s a layer on the tech stack, and a lot of calibration to do, but the workflow is this:

SCOUT:

Find influencers and creators based on specific fit with our ICP and performance criteria.

OUTREACH:

Pick the brand tone of voice, and draft initial copy to catch personas attention within the first message.

SCORE-FIT:

Filter the creators found based on success criteria (e.g. engagement rate, audience size, ICP fit) and green light the ones over 65% score-fit.

PAYMENTS:

Sync a unique Stripe link for each participants, with an automated receipt once they have paid.

BRIEFS:

Create custom briefs for creators, with their custom tracking links added and campaign messaging based on their target personas.

TRIAGE:

Every replies from interested creators is being synced to pre-approved follow-up messages, including a video explainer. Questions outside the scope are being flag for manual responses.

WEEKLY:

Scrape the database of creators and their status, build a weekly report on the campaign performance, with updates pushed daily. I’m the one approving it before it send.

ASSET-QC:

Checking the first drafts from creators, making sure it respect the messaging, the tone, the product specs, and including the right links and hashtags. Anything that doesn’t get flagged.

Once this has been automated, I still keep the tasks that I handle with my zone of genius. The calls with creators, the strategy of the campaign, and jumping in when there’s a problem.

These 8 agents could save me HOURS of work.

Last week I shipped the same machinery to another prospect. Same exercise, different brand, different ICP.

They'll get the agents in 3 weeks because I already did the 6 months.

You can't automate work you haven't done. But once you have, you should never have to do it again. That’s the neat part.

If you're a PMM staring at your own version of this spreadsheet and you’re looking to automate the boring tasks, there’s two ways I can help you:

Train you on my AI cheat codes

Hire me to automate the boring tasks

I’m already fully booked for June, the best way to keep me top of mind is to fill in the waitlist (takes 4 minutes)

Enjoy the start of the summer ☀️