Success is in the agency of...agents?

We’ve shipped three new AI-powered products at RevBoss in the past few weeks — all built with AI agents, deeply integrated into our process, and already making life easier for us and our clients.

Are they actually products? Or are they AI agents? Maybe?

I’m honestly not 100% sure what “agents” actually means.

But I’m 100% sure that I’m 100% a true believer that this is how work happens going forward and that the implications are deeper and wider than you think.

Thanks for reading.

Eric Boggs
CEO @ RevBoss

PS: As a reminder, you’ve received this email either most likely because you completed a form on revboss.com or (less likely) because we’ve interacted on LinkedIn. The unsub URL is in the footer if you don’t want to hear from me / us anymore.

I think that we all understand the pipe dream:

“Buy my tickets for the show, book my dinner reservation, and schedule my Uber back home. Thanks, Magic Robot!”

But I don’t think many of us understand what it takes to get there and what “agency” actually means for an AI-enabled workflow…which I think is basically the state of the art for AI agents today: AI-enabled workflows.

Think prompts that are linked together with conditional logic or branching logic and some basic “memory” that follows along the prompt chain…and maybe with some out-of-the-box “thinking” from the AI model of your choosing or some data integration from 3rd parties or internal tools.

Friend of RevBoss Robbie Allen wrote a great piece about agents a couple weeks ago — give it a read if you want to go a little deeper.

The recent bolt of lightening realization for me was how easy it has become to use an AI agent to build products that solve internal workflows.

Zapier is the OG here, obviously. Clay is the most popular tool du jour. Both are amazing products that are easy to use for simple stuff…but get unwieldy awfully fast.

The Replit agent and all of the other vibe-coding (I hate that word) agents are a different beast altogether.

Instead of point solutions, one-off tasks, and rickety “integrations”, these tools enable products. Like actual products with users, integrations, reporting, etc.

AI coding tools aren’t for the faint of heart — you need to deeply understand products and at least know enough about software engineering to be dangerous. And you can definitely create a huge mess if you don’t know what you’re doing.

But if you know what you’re doing — which increasingly I think I might and we do at RevBoss — these tools can change how you work.

Jason Lemkin agrees with me.

There is an old start-up adage that “every spreadsheet is a business opportunity”.

Not so much anymore.

Now a new SaaS idea isn’t competing with every other similar product, it is competing against someone with a little time, inkling, and know-how that’ll just build it on their own.

Said another way — the previous iteration of revboss.com cost us $50k I think.

The latest version took ~$200 in Lovable credits and maybe ~20 hours of my time, plus some time from the team.

There is obviously a decreasing returns issue here — I’m not going to spend a week building a solution that I can buy and implement for $50/month.

But for something that we’re paying $1000s per month to use…but maybe only using a small piece?

Or for a problem that doesn’t get adequately addressed by an off-the-shelf product?

I think that savvy operators will increasingly build, deploy, and maintain software internally.

And I think that there will be a big opportunity for people and organizations that can marshal these tools to solve real problems good, fast, and (relatively) cheap.

The improvement in these tools over the past ~6 months has been staggering and I think they’ll only get better.

Maintenance that is my only direct concern.

I know enough about software to know that the bigger it gets, the more problems you’re gonna have — many of which are hard to predict or understand.

But my (maybe overly optimistic) take is that if a homemade app gets big enough such that it actually needs proper infrastructure, redundancy, monitoring, and security, then great! That’s a pretty quick and easy hire to make.

My primary indirect concern is the all-in cost of these tools as denominated by my time.

At this point, I’ve spent 100s of hours building stuff on replit — including some fun stuff that has absolutely nothing to do with RevBoss.

(Ask me about https://ratemysoccerclub.com hahaha.)

Was that time worth it for me to learn a new set of skills? Yes.

Is it worth it for me to spend 10 to 20 hours per week hacking out tools for RevBoss? Also yes. It’s an expensive investment on an hourly cost basis, but it’s already paying off.

It’s also incredibly fun and rewarding.

You’ll see all of this soon enough if you’re a RevBoss client. Happy to share and discuss if you’re curious — just reply to this email.

Campaigns: a mechanism for us to build segmented, qualified prospect lists and for our clients to approve prospects and personalize messaging for campaigns.

Leads: a daily “sales to-do list” app that uses triggers, scoring, and notifications to tee up a daily queue of high impact sales activities for our clients.

Content: a workflow app to capture content inputs and ideas from our clients that integrates into our AI / copywriter production line.

All of these started as “I wonder if we could…?” experiments. They’re all now in production (or close to it) with our team and clients.