We’re told the future belongs to engineers. I disagree. After watching Marketing Against the Grain sit down with Tina Wong, I’m convinced the edge goes to the operators with deep domain expertise who can wield AI like a scalpel. My view is simple: the winners won’t be the most technical—they’ll be the best at knowing the problem, then using AI to scale their judgment.
Useful Beats Flashy—Every Time
Kipp Bodnar and Kieran Flanagan opened with a bold claim I support: the next wave of builders will be great marketers, sellers, and operators—not pure technologists. One line stuck with me:
“The next generation of builders… are going to be the best marketers, the best salespeople, the best business people. They’re not going to be the best technical people.”
Tina Wong, a data scientist turned creator, backed it up with proof from the trenches. The most useful workflows she builds aren’t flashy bots—they’re boring on purpose. Think reporting your stakeholders actually read. Think customer service replies that arrive fast, never get tired, and feel personal.
“The most useful workflows are usually not the coolest.”
That practicality translates into real retention. Tina showed a simple agent that reads cancellation emails, identifies the real reason, then replies with a targeted solution or offer. It personalizes at scale, and hands off to a human when unsure. It’s not magic. It’s the right tool for a very specific job.
Start With Your Work, Not The Tool
Here’s where many teams blow it: they start with a platform, then go hunting for a use case. Tina flips that. Map your own daily workflow. Find the low-risk, repetitive tasks that drain time. AI shines where availability, consistency, and personalization matter.
“AI is very good at being available 24/7, is very good at consistency… and personalization is a really big one.”
I’ll add a hack I use in my own businesses: record an hour of your screen while you work. Feed it to a capable model that can watch video. Ask what to automate and how much time you could save. You’ll get practical suggestions fast.
And remember the harsh truth Kipp dropped:
“If your prompts suck, your agents suck.”
Great agents start with clear instructions, tight constraints, and examples. Weak prompts equal weak results. That’s fixable.
What Every Agent Needs To Work
Tina used a burger analogy I love. You can swap toppings, but some parts are non-negotiable. Agents are the same. You need the core pieces or the whole thing falls apart.
- A capable language model
- Tools the agent can use to act
- Knowledge and memory (short and long term)
- Optional voice or audio if your workflow needs it
- Guardrails to prevent bad behavior
- Evaluation, testing, and deployment plans
Those last two are where most teams fail. Build the agent, then prove it works. Start with a handful of simple evals. Track where it fails. Improve. Repeat.
“Even if you just start off with like five evals, that’s actually quite good.”
The Real Advantage: Domain Expertise
This is where I’m planting my flag. As someone who’s built companies in crypto, marketing, and social, I’ve seen the pattern: the best agents come from people who live the problem. Not the coder guessing at what pest control techs do. The pest control pro who knows the job cold and can judge quality.
“The people who end up building the most valuable agentic systems… have very deep domain expertise in a field.”
Tools are friendlier now. You can learn enough in a month to ship real workflows. What you can’t fake is judgment. That’s why this shift matters. The power moves from “I can code anything” to “I know exactly what to build and how to measure it.”
My Playbook For Teams Right Now
Don’t overthink it. Pick one messy process and fix it with AI. Here’s a simple plan:
- List your repetitive, low-risk tasks that happen weekly.
- Record one session; ask an AI model what’s automatable.
- Define success. Write five evaluations you can score.
- Draft strong prompts with examples and firm rules.
- Add guardrails and a human escalation path.
- Ship a narrow workflow. Measure. Iterate.
This isn’t theory. It’s how teams cut churn, cut response time, and keep customers from bailing.
Final Thought
AI won’t replace you. But someone with your domain expertise, plus a working agent, will. Stop chasing shiny demos. Build the unsexy systems that save time and keep revenue in the door. Start small, prove it with evals, and expand from there.
If you know your customer and your process, you already have the hard part. Now add the tools. The best builders of this era are the craftspeople who ship workflows that work. Let’s be those builders.
