Editorial series · 15 weeks · Week 4

The 5 Pillars: How to Ask Anything of an AI and Get It Right

Roberto Ocampo September 9th 2026 Philosophy / Method (Ocampo IP)

Last week we left off with something important: when you ask an AI for something without a role, without inputs, without a format, and without a limit, it's not the machine that's failing — it's the order. We looked at four parts that are almost always missing. Today I'm putting a full name on the method behind those four parts, because I didn't invent it for this course: I've been using it for years in a real business, with real teams, on tasks that have nothing to do with each other — from avocado pricing to international contracts — and it works just as well for any trade.

They're called the 5 Pillars. Five questions. When all five are answered, the AI doesn't get lost. When one is missing, it guesses, and what it guesses is almost never what you actually wanted.

Where they came from (and why it matters that I didn't invent them yesterday)

I didn't draw this on a whiteboard for a slide deck. It came out of the daily need to hand work to a system and have that work come back right the first time, without back-and-forth. Early on I called them "the five little facts" when I'd ask my team for them before sending me any material: "send me the info with these five points and the system won't get lost." Over time, the same structure served executive instructions, technical work, any kind of assignment. Five questions, three ways of phrasing them, the same backbone underneath.

That backbone is what I'm handing you today, translated into your trade.

The 5 Pillars, one by one

Pillar 1 — Objective. What is this about?

The topic and where it comes from. "Write me a quote" isn't enough: a quote for what, for whom, and why do you need it now? A contractor doesn't tell their crew "fix that up" — they say "this is for Mrs. Miller, who called Tuesday because she wants the yard redone before it gets cold." Origin and purpose are part of the order, not extra detail.

Pillar 2 — Inputs. What am I giving it?

The reference material: measurements, prices from your list, photos, an old contract you want to use as a template, that client's history. This is where everything you already know about your trade comes in — the part the AI doesn't come with out of the box. Without inputs, the AI doesn't lie on purpose — it fills the gaps with the generic, and generic almost never fits your specific business.

Pillar 3 — Focus. What should I pay attention to?

This is the one people skip most, and the most valuable one. It's not just "analyze this," it's "analyze it paying attention to X." A real example from my own work: "this presentation is outdated because the avocado tariff changed" — that tells the AI exactly where to look and why. In your trade it'd be something like: "concrete prices went up this month, adjust the estimate accordingly" or "this client already complained once that the estimate didn't include final cleanup, don't leave it out again." Focus connects directly to Pillar 1: it tells the AI where inside the objective to put the effort.

Pillar 4 — Deliverable. What do I want back, and how?

Format, length, who it's for, and by when. A table? An email ready to paste? A short paragraph for a text message? Do you need it today or for Friday's meeting? If you don't say, the AI decides for you — and it decides with roughly a coin-flip's odds of getting it right.

Pillar 5 — Mode and conditions. What hat should it wear, and under what rules?

This pillar has two halves and both matter. The first is the role: do you want the AI thinking like your accountant, your salesperson, whoever drafts messages to a demanding client? The second is the most important of the five, and the one almost nobody includes: an explicit Yes — what you actually want to achieve, or which sources/rules are valid — and an explicit No — what's completely off the table. "Don't invent prices." "Don't promise dates we haven't confirmed." "Don't use a casual tone with this client." The Yes/No isn't optional, and it isn't for special occasions: it's the difference between an AI that polices itself and one that fills gaps with whatever seems plausible, which is not the same thing as correct.

Why it works the same way in any trade

What makes this method strong isn't that it's complicated — it's the opposite: it's the same structure no matter the topic. Pillar 5 in particular is deliberately designed not to stay stuck to a single domain. In a scientific case, the Yes/No is valid sources or not. In software, it's the coding standards that are followed or not. In design, it's the brand guidelines that are respected or not. In your business, it's going to be your own trade rules: the things a good employee of yours already knows without being told, but that an AI will never guess on its own.

And here's something important: you don't need all five pillars filled out to get something useful. The more complete they are, the faster and more precise the result — whatever's missing, the system has to guess, and you already know what happens when it guesses. The discipline isn't filling out a bureaucratic form every time; it's holding all five in your head as a quick check before you hit send.

The 5 pillars applied to a trade example

Let's go back to last week's quoting example, now with all five pillars instead of four:

  1. Objective: quote for artificial turf installation for Mrs. Miller, who called Tuesday.
  2. Inputs: 400-square-foot backyard, concrete border around the perimeter, my reference material prices from this month.
  3. Focus: concrete prices went up this week, adjust the estimate; this client already asked about warranty, don't leave it out.
  4. Deliverable: proposal in English, list format, ready to paste into an email, needed today.
  5. Mode and conditions: act as my quoting assistant; Yes — include a one-year warranty; No — don't invent prices, leave the field blank where an amount goes.

Five answers, fifteen extra seconds to write them, and the difference between a draft you have to rewrite from scratch and one you paste straight into the client's email.

A brand method, not a course trick

I'll say it plainly, because it's part of what you'll learn with us: this framework is the protocol Ocampo-Infra uses to fire off any task we hand to an AI system, from a small quote to the entire infrastructure that runs this very blog. It's not a list of tricks I put together to sell a course. It's how we already work, every day, with real clients — and the reason our systems don't get lost when the work gets complicated.

This week's practice

Take anything you asked an AI for in the last two weeks that didn't come out the way you expected. Check it against the five pillars, one by one. You'll almost always find one was missing — usually Pillar 3 (the focus) or half of Pillar 5 (the explicit No). Ask again with all five filled in and compare.

Next week we go back to fundamentals, with the engine that makes all of this possible: what an LLM actually is, explained in the language of your trade, without extra jargon.

Want to build it with us?

This is exactly what we teach —and build with you— in Cohorte 2026. You walk out with a finished AI project that Ocampo-Infra can run for you. Grab the free guide and claim your seat.

Roberto Ocampo · ro@ocampo.ai · https://ocampo.ai