154: How We Use Grok Bot in Our Outbound Process

Welcome back to the Practical Prospecting Newsletter!

I’ve been using Grok Bot inside the agency for a month now. Figured it would be interesting to share the 3 areas where it’s actually making an impact, and where it isn’t.

Quick context if you haven’t used it: Grok Bot doesn’t necessarily need API keys. You give it a login, and it signs in and does the work for you like a person would. It just feels much faster at connecting to your tools, understanding what you want, and actually getting them done, without a ton of trial and error.

However, as you’ll read, I still prefer to use Claude for deeper/more complex tasks.

Agenda:

1. Monitoring the Little Things

2. Replying to Leads

3. List Building

Influ2 is one of the first companies changing ABM with contact level intent. 

Instead of telling you which account is "interested," they show you the specific people who are, so you know exactly who to reach out to. 

They're hosting a webinar on this exact topic on November 5th at 12pm ET, so you can see exactly how it works. 

Sign up here!

#1 - Monitoring the Little Things

When you’re running hundreds of campaigns across clients, there are a ton of small checks that make a big difference if you do them every day.

Individually they’re easy. But it’s tedious when done one by one across every campaign.

Stuff like:

• Is one variant outperforming the other?

• Did bounce rate go up anywhere?

• Did reply rate or positive reply rate drop?

• Is any campaign running out of leads?

• Is any campaign sending too little volume, or too much?

• Is a campaign going out with a missing merge field?

• Are any campaigns sending from burned inboxes?

• Are there offers, angles, or subject lines working for other clients that could work here?

Setting this up in Grok Bot was so easy. I just told it the things I wanted it to check, and within minutes it built an agent that runs every day and sends me a Slack report.

Step one was just visibility. So I could see all of the action items in one place every morning.

Step two was having Grok Bot actually do those tasks. For example…

• When a campaign runs low on leads, it automatically submits a list request to our GTM engineers

• When a campaign has burned inboxes, it automatically creates a request for new ones

• When one variant is outperforming another, turn of the bad performing variant

Step three is the one that’s had the biggest impact:

Grok Bot does a weekly analysis across every campaign. Then it sends me a report that says, roughly, “here’s a trend I’m seeing in what’s working, and here’s how you could apply it to your other clients.”

The best ideas from one client end up helping five others.

#2 - Replying to Leads

Out of all three, this one has been the most impactful.

For a while, one of our biggest problems was speed to lead. We’d generate a lot of interested replies, but it took too long for those leads to be replied to. And even when they did, they often didn’t get followed up with.

So there were meetings left on the table that should’ve been booked.

We’re starting to use Grok Bot for this. I’ve tried a lot of reply agents, and most of them tend to go a little off the rails every now and then with their replies. But Grok Bot is really good at following tight guardrails and instructions.

Here’s what I had it do:

For each client, it read through ALL the human replies across our inboxes from the last several months.

Then it built a report of the common replies we get, and how we’ve responded to them in the past.

Now it uses that to reply (or a variation of it) going forward.

The thing is, there are only so many types of replies you can get. After a few months of data, there’s rarely a new one. So now it replies to all of them, in minutes.

It also handles the follow-up. If a meeting hasn’t been booked, it sends the next follow-up, then the next one.

And when someone says “reach back out in two months,” it sets the reminder and follows up when it’s time.

A couple other things it’s doing in the inbox:

• Reply tagging. 

  • We have Smartlead auto-tagging set up, but it gets things wrong sometimes. So every day, Grok Bot reads the individual replies and fixes the tags. That way we know how many true positive replies, true out of offices, and true not interested replies we actually got.

• Reply analysis. 

  • We get thousands of replies that nobody’s going to read one by one. Grok Bot reads all of them and spots trends, like “a lot of people are raising this objection, you might want messaging that addresses it.”

That last one has been huge for improving copy across clients.

#3 - List Building

We were already doing a lot of list building in Claude. And we still use Clay for a lot of our use cases.

But Grok Bot has been really good at list building for one reason: it can just log into tools and doesn’t need an API key or MCP.

So right now I have it doing things like:

• Logging into our Prospeo account

• Using the Clay to run templates and some baseline stuff

• Running a few scrapers we have

I’m not having it do complex list building work. It’s just the connector between tools.

The cool thing is we already have SOPs for how we build these lists. So we literally just tell it, “log into Prospeo, apply these filters, export the CSV, and push it into Smartlead.”

And it does every step, one by one. Creates the campaign, pulls the list, uploads it.

For really complex lists, we still use Claude. But for the simple ones that just need to be built, exported, and uploaded, Grok Bot has been handling it.

Where it falls short: messaging

Grok Bot is really good at being efficient and getting work done. It’s not great at deep work, at least from what I’ve seen so far.

Claude still does a lot of the analysis on our messaging, and it’s way better at coming up with initial messaging angles. Grok Bot’s first drafts tend to be pretty shallow, even when I give it the exact same prompt I give Claude.

So for anything that takes real thinking, like finding the angle, writing the sequence, or working through variants, we still use Claude.

Thanks for reading,

Jed

P.S. If any of you are in the Boston area, I’ll be there this week for an in person (and free) outbound workshop. More details here!