There Are No Low-Level Tasks

Brian Sowards told me that a team of three opened $100 million in channel-source pipeline, and my honest first reaction was that it had to be a typo. It wasn’t. Brian is a go-to-market engineer (his words, and the right ones) who builds agentic GTM systems for operators inside fast-moving organizations, and that number is the kind of thing that either breaks your model of what’s possible or confirms everything you already believed about lean teams and AI. For me it was the second one.

Most people hear a number like that and assume the AI was off in the corner doing the boring work, freeing up the humans for the important stuff. Brian’s actual point is the opposite of that, and it’s the most useful reframe I’ve heard all year.

“I don’t know anyone in marketing whose biggest problem right now is how do I publish fifty blogs I’ve already written. There are no low-level tasks. The power of AI is leveling up the quality of what you do and the speed at which you do it.”

That single shift changes where you point AI in your business, and almost everyone is pointing it in the wrong place.

The work AI should be doing is the expensive work

When teams first got access to real AI, they went looking for the cheapest, most repetitive tasks to knock off the list. Brian’s early work with Aviatrix proved how badly that instinct undersells what’s actually available.

Aviatrix sells multi-cloud orchestration into the cybersecurity space, which is about as complex as B2B sales gets. The buying process runs through multiple heroes (the senior AE who can build the business case with the C-suite, the technical SAs and SEs, the forward-deployed engineers whose time has to be protected because they’re carrying a virtual pager everywhere they go). Marketing sits in the middle of three completely different motions trying to figure out how to help. This is not a “publish fifty blogs” environment. It’s the kind of problem most people assumed AI couldn’t touch.

What AI unlocked there wasn’t low-value at all. It was taking the dense technical documentation the engineers wrote for support issues, the stuff that captured the exact reasons a company would actually buy, and turning it into high-performing content. Then they paired that with an organized LinkedIn motion and a partner and client roadshow built around AWS re:Invent. The result was an 80% reduction in marketing effort and $10 million in net-new pipeline opened in a single month. None of that was grunt work getting automated. It was expensive, high-judgment work that had simply been too hard to do at scale before.

“Get organized. That’s what the challenge is. Everyone who’s been in this business knows that meeting people and having a roundtable and building relationships works, and it’s a nightmare to organize.”

That’s the real gap, and it’s why Brian describes agentic operations as a bit of a Trojan horse. You’re bringing AI into the organization, but what the AI is actually executing is the stuff the people doing the job already know matters. The relationships, the roadshows, the personalized outreach. The work that’s always been valuable and always been a logistical nightmare to run consistently.

The most important work in your business has no software

The Moxie Power story is where this really lands. Moxie was replacing diesel generators with smart battery systems, selling into a channel-driven, relationship-heavy world where you don’t just win the partner, you have to win their salespeople, and then those salespeople have to actually succeed at selling you. Brian called it a good-old-boy network where the person who sold you the equipment is the person you call when something breaks. You don’t out-spreadsheet that. You earn your way in.

So a team of three did the only thing that scales in that environment without losing the human core of it.

“We just agentically enabled them to communicate with the entire distribution network, but to do it in a way that set up relationship building. Hey, we’re going to be at the same conference. Let’s get together.”

That is $100 million in pipeline built on top of relationships, not in place of them. And it points at something I’ve been saying for a while from a different angle, which Brian put more bluntly than I ever have:

“Take the most important thing in the business, there isn’t software for it. There has never been a company that I didn’t drop Claude Code or Codex into that didn’t 5X the amount of software they wanted overnight, and all of it was stuff they couldn’t buy that was endemic to the value chain.”

This is the anti-SaaS thesis stated by an operator who’s lived it. The era of buying fifteen tools and duct-taping them together is ending, not because the tools are bad, but because the work that actually moves your pipeline is specific to you, and nobody is going to sell you software for it. You build it. The $60k stack becomes a $3k stack plus a system shaped exactly like your business.

Where to point this in your own org

If you want to find the work AI should actually be doing for your team, three questions get you there faster than any list of tasks to automate.

Stop making a list of cheap tasks to automate

The “what can we knock off our plate” exercise caps your ceiling at incremental. You’ll get a faster version of the same work and call it transformation. Start instead from the work that drives revenue and ask why you aren’t doing more of it. The answer is almost always that it’s too hard to organize, which is exactly the constraint a system removes.

Look for the work that’s too expensive to do, not too cheap

The highest-value motions in most companies (personalized channel activation, turning real technical expertise into content, running a genuine relationship motion across a whole partner network) don’t get done because they don’t scale with human hours. That’s the work to point AI at. Not because AI replaces the judgment, but because it removes the logistical tax that kept you from doing it at all.

Build the thing nobody will sell you

Audit the most important process in your business and check whether software exists for it. If it’s truly core to how you win, it probably doesn’t, because it’s specific to you. That’s not a gap in the market. That’s your moat, and it’s now buildable in an afternoon with Claude Code instead of a six-month vendor evaluation.

The honest part

I’m not going to tell you this hands everyone their evenings back, because the most honest thing Brian said all episode was that it doesn’t. He works more now that he has AI agents than he did before, and so does nearly every leader he talks to. The leverage is real, but we’re in the middle of a transformation, and transformations don’t fit neatly into nine Zoom meetings and a tidy schedule. The people getting these results are doing it in the early mornings and late nights, and pretending otherwise would be the exact AI-hype dishonesty I started this podcast to avoid.

What I’ll stand behind is the shape of the opportunity. The systems Brian builds, the client walks away owning. A personal, proprietary AI tech stack, not a dependency on his GitHub repo. That’s the same principle underneath everything I believe about this work: your lived experience and your internal data are the only real moat left, and a system built around them compounds in a way no off-the-shelf tool ever will. Small teams get things done. A team of three just opened $100 million in pipeline to prove it. The only question worth asking now is whether the most important work in your business is getting AI’s attention, or whether you’re still using it to publish blogs you already wrote.

---

This was my conversation with Brian Sowards on Barely Shipping. The full episode is on YouTube, and you can find Brian at sowards.ai. The free book this thinking is built on, Pipes Before Chocolate, is here.