Using AI to Drive Down LinkedIn Ads Costs

Today we’re going to talk about two of the most popular tools in the modern B2B marketing stack: LinkedIn Ads and Claude.

If you or your team have tried plugging AI into LinkedIn Ads and gotten something mediocre back, it’s likely because nobody fed it the right data or gave it a system to work inside of. Do it right and the math gets interesting: double your click-through rate on LinkedIn and cut your cost per click in half. That’s worth building a system around, and it’s exactly what Anthony Blatner’s been doing.

Anthony is the founder of Speedwork, a LinkedIn ads agency (I can vouch that he knows his stuff; he’s been running all of our LinkedIn ads for the last two years and we keep spending more and growing…), and he sat down with our head of community, Matt Carnevale, to walk through exactly how he’s using Claude Code with the LinkedIn Ads API and we thought it would be the perfect thing to share in this week’s newsletter.

Here are some of Anthony’s go-to tactics that he’s using for his own clients (or just watch the episode on YouTube):

1. Build a LinkedIn ads agent that flags wasted budget, so you don’t need to babysit every campaign.

Anthony feeds Claude Code the LinkedIn Ads API documentation and has it build a working interface – basically an API agent he can just talk to. It pulls more granular data than the LinkedIn UI ever shows you, especially on demographics.

Once Anthony’s agent is built, it runs two checks on repeat: a bid scan twice a week to catch underspending or overpaying as prices climb through the month, and a demographics pull that flags which companies and job titles got reached that week. Anthony approves all bid changes himself, since Claude doesn’t know what he knows – like a new ad launching tomorrow that changes the math.

2. Personalize campaigns at scale that you could never build by hand.

Building separate, personalized campaigns for every segment used to be a huge pain. Anthony says most teams cap out around 25 before it becomes basically impossible to produce and maintain manually.

Here’s Anthony’s approach with Claude: build a matrix mapping each segment (say, plumbers, electricians, dentists, and chiropractors if you sell software to small business owners) to its own targeting, ad copy, image, landing page, and offer – whether that’s a demo request or a personalized report. Complete the first row of your matrix on your own, then have Claude replicate it across the full list.

3. No budget for dozens of campaigns? Thought leader ads are the best entry point on LinkedIn.

Anthony’s answer for anyone who read the previous point and thought “great, but I can’t afford 50 campaigns”: pick three employee or exec posts and promote each one to a different target segment. Thought leader ads get the best engagement rates, which means they’re also the cheapest way in if you’re starting small.

NOTE: Big budget or small, thought leader ads continue to be the #1 performing unit on LinkedIn. The move here is to post organically first, see what gets traction, then put dollars behind the winners, instead of guessing which posts to boost. (Or do what we did yesterday: I wrote a post to promote our webinar; LinkedIn hates links so it didn’t perform well, but the whole point of the post was to make something that we could boost via paid to drive more registrations).

4. Build the offer your leads want, before they have to ask for it.

The highest-performing move Anthony’s seen isn’t a better headline. He pre-generates a report for each specific target company on his list, before the campaign launches. The ad’s CTA is “get the report for your company.”

Someone clicks, and instead of a form that triggers “a rep will reach out in a few days,” they get their company’s report instantly. The moment you make someone wait for a human, the lead cools. If you can’t make unique industry reports, customize the landing page so it’s built just for them. With AI, Anthony’s team can create ten personalized pages in an hour once the first one – the model – is done.

5. Feed Claude real customer conversations so prospects hear their own words in your ads.

This is the line between a good ad and obvious AI slop. Anthony pulls actual CRM data (emails, call notes, deal history, and call transcripts) then organizes them into folders inside Claude Code by competitor or industry. He starts with a brand guidelines file so every output stays on-voice.

When Anthony kicks off a campaign build, Claude pulls from the folder for that segment. The output sounds like it came from an actual conversation with a customer in that industry, because it did.

6. Start running Competitor Conquest campaigns.

What’s in a name, right? Anthony’s been running these for years and it’s exactly what it sounds like. Some of your prospects are already unhappy with a competitor or coming up on a contract renewal.

Anthony builds a list of the company’s ten most common competitors, then targets the companies known to be using each one, and runs a separate campaign per competitor. Since you know your pitch against each one, the ad copy writes itself. Anthony starts with a test group of three companies before rolling out to the full list.

One of the most impactful things Anthony shared on the pod was that if you’re able to take these tips and double your CTR, you can usually cut your CPC in half. A 50% discount on LinkedIn ads, he calls it. Every one of these tactics is a different way to get there – by making your ads more relevant to the person looking at it.

Alright, that’s a wrap on my word count, but if you want more (there’s much more), give this a full watch/listen. Check out the full episode for more.

PS Anthony is speaking at Drive this September, and we’re running low on tickets. Grab yours here.

Frequently Asked Questions