You know what the marketing funnel is, right? You run ads on Google or Facebook so people learn about your brand or product, or you buy a list of names and send them cold emails. That’s “top of funnel.” Some fish nibble: They fill out a form, you schedule a call, and they don’t show up to the call. You send them a follow-up email saying you hope to talk soon anyway, because you are a worm with no pride.

And so on, down the funnel, until eventually very few people are left, like at the end of a bad party, and you get one of them to sign on the line that is dotted before the lights come back on. Ring the gong! Now you can actually get to work delivering their solution, or building their software, or delivering their catering, or optimizing their yield, or whatever it is you do.

That was then, but now we’ve got AI, lucky us, and it’s infested the funnel. Bad salespeople create even worse virtual salespeople and fill my mailbox with nonsense.

“Hey Paul,” they write. “I saw on LinkedIn that you might need AI-enabled cat-sitting platforms? But you might have missed my email. Not a good time? Is there someone better than you? Someone more interesting? Do you want a gift certificate to Arby’s? I can BURY you under roast beef if that changes things. It’s been hours! How are your wife/wives $WIFENAMES doing at their professional jobs? I wrote you a poem. Here’s a picture of an otter AHAHAHA RIGHT????!?!?! As a $JOB_TITLE, you know how important this is. You smell like oranges in summer. Can I touch your face at night?”

This is how modern, AI-enabled automated email marketing tends to work, spread out over weeks and days. When I imagine these people doing their jobs, I think of a large sewage treatment plant, except it’s churning through LinkedIn profiles.

The problem with this approach isn’t just that it’s annoying—it’s that it’s narcissistic and greedy, and tries to keep all the value for the sender. The pitch is not, “I think my service will be useful to you, as a human with goals of your own.” It’s, “You’re too stupid to know the difference so I’ll fake a persona and nag you with robot-personalized pitches.”

And look, that’s a bad product. If you are unwilling to show any value, and you think a relationship with me is a generic commodity, your product is de-facto going to suck. Why would I keep talking? I used to shrug off cold pitches—everyone’s gotta eat! But I now spam-and-block anything in my inbox that gives off weird AI sweats.

So is this the future for AI marketing? Their bots send garbage and my bots throw it in the trash? That seems like it won’t...work, at least not for long. So what else can we do?

When Rich and I ran an agency, we’d sometimes do free work—design, consulting, strategy—as part of the pitch. There was an implicit understanding that the client wouldn’t take that work and hand it to a cheaper firm, but that sometimes happened. But just as often, the free work landed us the deal. That’s just how a lot of consulting and high-level agency stuff operates: You do a lot of work for no money, sometimes hundreds of thousands of dollars’ worth, and then they give you a contract, because it’s riskier to let you walk away. People like to see that you’ll commit and take a risk, too. Is this how a rational industry would work? Let’s not solve that now.

If you see a funnel, you must flip it.

At the same time, software is getting a lot easier to build, at least prototypes and demos. Reports have gotten easier to generate. Charts are faster to make. All sorts of things can be done with prompts. So let me describe a future scenario, in eight steps:

First , I identify 20 people on LinkedIn—the kind of people who might buy the software I sell. I try to find people who are interesting, relevant, and have said some smart stuff.

Second , I ask AI to help me compile some interesting things about their careers—nothing too intrusive. Did they go from big companies to small, or small to big? What topics do they post on? What are some things they’ve said that I might agree or disagree with? What professional affiliations do they share publicly? Ideally, I’d follow some links and actually read what they said, maybe even add some notes.

Third , I’d ask AI to take that dossier and my notes, and list five software applications that person might need in their job—the more specific and obscure the better. I’d eyeball that list.

Fourth , I’d ask AI to take that list and pick the application that would save the most time and drive the most revenue for that person.

Fifth , after that’s done, I’d say, “Go to my AI-powered software-building tool”—for the purposes of this exercise, I have one of those—”and ask it to build that exact application. It doesn’t have to be perfect. Fill it with sample data.”

Sixth , (this might take ten minutes per app and cost a buck or two) I’d tell it to deploy the app to a secret, impossible-to-guess URL.

Seventh , after all this is done, I’d drop those 20 people emails—maybe I’m cutting and pasting, or maybe AI is helping me. Each email is something like:

Dear Amelia,

You don’t know me but I like your stuff on LinkedIn, and I saw in your post two weeks ago that you are pretty frustrated with your alpaca farm-management software. We actually love alpacas at our company, and so we were inspired to use our custom software building framework to create some new, custom alpaca farm-management software for you. If you’d like to take a look, here’s a working demo (you’re the only person with this link but feel free to share it). The features we focused on were [insert features here]. Utterly glad to hear any feedback. Mostly just curious to see if we did a good job or not. And I won’t bug you further—just reply to this email if you have any interest at all.

Sincerely,

Paul

And finally , I’d send out those emails, with links to real, working software, that they could use and explore. From my actual email address.

See what’s happening? My guess is the future SaaS or services funnel starts with custom product delivery—and then you sell the relationship, the trust, and the long-term support and stability. This future funnel would be totally inverted: It starts with custom, bespoke product delivery and ends with brand awareness.

Now take those 20 people, and turn them into 20,000—or 2 million. Automate even more of the process. How far could you go? How do you keep it from being creepy? There will be a lot to explore here. Is the world ready for opt-out consulting?

And sure, this is speculative—but also possible. I hope to find ways to do it tastefully and ethically. I don’t know if this approach can spread widely, because ultimately we’re talking about delivering products—and marketing people tend to have almost no access to product people. The idea that product culture and marketing culture might converge would lead to mass panic in the streets.

That said, I bet if you opened the conversation by shipping custom software (or creating other high-value demo products), more people would show up to the Zoom calls.

Sliding down the funnel,

Paul