Roberto Cortez / Blog
Choose what you want to work on: applying AI, improving sales or running your business. Explore processes, decisions and examples from my classes.
Prefer a guide by subject? Explore AI, automation and business.
To hand a project between agents, provide the objective, current decisions, verifiable status and next authorised task. Include references to files and completed checks…
Read article ↗Installing many skills can worsen work when it introduces contradictory rules, unsuitable procedures or choices that are difficult to distinguish. Quantity alone neither…
Read article ↗Create a skill when a task recurs, you already understand its quality criteria and you want to preserve a useful procedure. Save the objective, required materials, steps…
Read article ↗A CLAUDE.md or AGENTS.md file should gather project instructions that remain useful for future work: objectives, boundaries, confirmed decisions and ways to verify…
Read article ↗You preserve project context between sessions by saving decisions, work status and necessary references in documents you can consult again. Conversation helps execution…
Read article ↗An automation can create more work than it saves when the team must correct results, reconstruct context or handle exceptions without reducing total effort. Compare the…
Read article ↗Separate assistants when tasks require sufficiently different instructions, sources or limits to justify their own contexts. Specialisation should reduce confusion and…
Read article ↗Introduce AI through a bounded task, with someone responsible for comparing its result with the expected work. Start with analysis support or draft preparation and make…
Read article ↗You reduce dependence on one supplier when you can recover your data, explain your instructions and continue the process outside the interface where you started. That…
Read article ↗Asking AI to review everything repeatedly only makes sense when each review seeks a relevant error and has a completion criterion. Without that criterion, you add…
Read article ↗To turn a team procedure into an automation, first describe the expected result, required information, decisions and exceptions. Observe an actual execution and make…
Read article ↗Choose between n8n and a custom application by identifying what is growing: integrations, human interaction, rules or data management. A larger workflow may need better…
Read article ↗An automation can work in a demonstration and fail in a business because the data, people and operating conditions differ. A demonstration confirms one possible path…
Read article ↗Choose speed according to the task and the quality needed to complete it. A fast model may work well for simple classification, while an analysis with several…
Read article ↗Autonomy should depend on permitted actions, the consequences of an error and whether you can verify the result. An assistant can prepare useful work without executing…
Read article ↗Before automating a task, confirm that someone needs its output and that its frequency makes sense. If the routine no longer supports a decision, a delivery or a…
Read article ↗To prevent isolated automations, make the objective, data, owner and actions of each process visible. The team can keep experimenting, but work that starts serving other…
Read article ↗You should leave a tool when the cost and limitations of continuing exceed the expected benefit of keeping it, after allowing for the transition. Money already spent…
Read article ↗To measure the return on an automation, compare the actual work before and after, including review, corrections, maintenance and team adoption. Measure recovered time…
Read article ↗In an automated process, people should retain decisions involving insufficient context, consequences that are difficult to reverse, or a need for negotiation and…
Read article ↗An automation owner should monitor results, handle exceptions, keep rules current and decide when execution needs to stop. That person must understand the process and…
Read article ↗A small team's first AI pilot should address a limited, frequent and verifiable task, with available data and a named owner. Choose an experiment that allows mistakes to…
Read article ↗To automate a frequently changing process, start with the stable stages and keep decisions that are still being explored under human review. Separate collection…
Read article ↗An automation can be technically possible without justifying the effort required to build, supervise and correct it. Compare the task's actual frequency with the total…
Read article ↗Buying software makes sense when an existing solution handles a common process well and the team can use it without constant workarounds. Building an internal tool with…
Read article ↗I explain how I collect Stories metrics by sequence and use AI to spot retention, tap-back and exit patterns, without treating correlation as proof.
Read article ↗How to give AI context and feedback so it suggests content you'd actually publish, using the real example of my own Instagram stories.
Read article ↗How I use small projects, explicit rules and ongoing review to build real judgement when working with AI in paid traffic and business.
Read article ↗I apply the distinction between feature and benefit, the one I use to write copy, to selling automation and AI services: turning integrations into changes the client feels, without inflating promises.
Read article ↗A simple WhatsApp automation example after a confirmed purchase: what the first message should say, which tools I used, and when a team member should step in.
Read article ↗Why I keep printed codes separate from final links, use my own domains and deep linking so printed materials don't break when the destination changes.
Read article ↗I revisit tools I've used and still use in the business to explain when paying more makes sense, when testing before buying pays off, and when an old tool stops being worth keeping.
Read article ↗How I prepare AI email reply drafts without letting the AI send them: triage, context and human review before sending.
Read article ↗How to select, clean and maintain support replies as a reliable knowledge base for an AI agent, without mixing exceptions with current policy.
Read article ↗An AI assistant doesn't know your offer, your customer, or your business history. I explain what context is worth giving before asking it to analyse any ad account.
Read article ↗What actually matters when feeding an AI assistant so it stops giving generic answers: behavioral instructions, brand content, metrics and file organization.
Read article ↗When you're starting a business, AI helps prepare materials, organize information and cut down on repetitive work. But that only works if you can explain the service…
Read article ↗Changing the background of an ad doesn't mean you're testing a different message. Before generating images, decide what you want to say and to whom, because otherwise…
Read article ↗A stop button that errors out from a network drop shows why you should never trust a panel without server-side confirmation.
Read article ↗How I built an auditable rule matrix to adjust campaign budgets without relying on opaque AI decisions.
Read article ↗I explain the difference between the command center I built for Ad Summit and the multi-client portfolio dashboard vision I described in class, and what is worth taking from that logic.
Read article ↗How to define who approves what when onboarding marketing clients: main contact, metric alignment, asset audits and setup trade-offs.
Read article ↗Before accepting a digital marketing client, I need concrete questions about the business, its margins, its digital assets, and who decides what. A practical discovery guide.
Read article ↗Before letting automation touch an ad budget, I define first the numbers it can and cannot move: minimum spend, daily cap and an alert on every change.
Read article ↗A WhatsApp report is only useful if it helps you decide something within seconds. That means the message needs to state the period, the results that matter, and the…
Read article ↗What you need ready before opening Claude Code for the first time: subscription, isolated folder, and care around exposed API keys.
Read article ↗From a live class where I built a WordPress → Google Sheets → n8n → Gmail workflow on air, I explain why it pays to test each step in isolation before letting an automation run unsupervised.
Read article ↗I choose an AI model by looking first at the concrete task, never at whatever name is trending that week. I define real test examples, compare the cost per run against…
Read article ↗How to organize data and context before asking AI to work on it, with real project examples and security precautions.
Read article ↗Before asking AI for an opinion on your ads, confirm exactly what you're handing over. Check the period the file covers, which campaigns it includes and what each result…
Read article ↗A practical inventory of tasks, decisions, time, and unmet needs to choose your company's first automation with judgment, including the real cost-per-decision example from class.
Read article ↗If the decision you want to automate rests on numbers and fixed comparisons, a conditional rule solves everything without complications. If the work requires…
Read article ↗Organise B2B prospecting with Apify, n8n and AI: qualification criteria, data review and handing opportunities over to your sales team.
Read article ↗Before coding with Claude Code, define rules, data and permissions. My planning process, using a football predictions application as an example.
Read article ↗Use AI with business context, organised data and clear processes. Examples to move from conversations to useful tools for your team.
Read article ↗The landing page should continue the conversation opened by the advert: confirm the offer, develop the benefit and explain the next step. When people encounter a…
Read article ↗The choice starts with the customer's position: are they already looking for a solution, or must they first recognise the problem and discover alternatives? Search can…
Read article ↗Prepare a creative refresh when consistent signs of weakening response appear, after checking measurement, the offer and operations. Use the history to choose new…
Read article ↗A winning advert can reveal which situation, benefit or objection deserves more attention in your offer. To learn from it, confirm the commercial outcome and investigate…
Read article ↗You are testing an idea when you change the reason someone should pay attention or take the next step. You are testing presentation when you preserve that reason and…
Read article ↗When you have little history, define a test budget around the risk you can afford and the question you need to clarify. Reserve money for execution, measurement and…
Read article ↗If a campaign generated sales but the money has not arrived, base the next spending decision on the schedule of receipts and payments. Confirm the sale, estimate the…
Read article ↗To separate new customers from existing customers, match each sale against purchase history and apply a stable definition of first purchase. Show both groups in the…
Read article ↗Platform CPA measures advertising cost per defined action; CAC aims to measure the cost of winning a new customer. To compare them, confirm the event counted, the actual…
Read article ↗Break-even ROAS depends on the margin available for advertising after the variable costs of a sale. Divide one by the pre-advertising contribution margin, expressed as a…
Read article ↗Before touching an ad I need data that supports the decision. I explain how I use the CPA multiple, time windows, and the rules that protect an account's history.
Read article ↗Before touching budgets, audiences or creatives, there are business questions a strategic media buyer should ask first, about offer strength, real revenue and historical data.
Read article ↗Why optimizing the conversion of visitors already on the store usually matters more than adding traffic, and where the common mistakes that block sales tend to hide.
Read article ↗Why a low-margin entry product can still make sense in your business, and what to look at before deciding to cut it.
Read article ↗How to create real urgency in an offer, without fake countdowns, using examples from the copy class.
Read article ↗Before asking AI to write your sales page, gather what only you know about the customer. I explain how to organize that brief, based on a class I gave on copy.
Read article ↗How really listening to customers reveals hidden beliefs no internal brainstorming finds, and how to organise those objections without turning one isolated case into a rule.
Read article ↗A simple test to know if your offer is clear to strangers, with examples from services, objections and beliefs.
Read article ↗What I ask a new client before touching ads, social media, or their site, and why it matters.
Read article ↗I explain why doubling your ad budget rarely doubles the result, using the real example of my book campaign and the sensitivity test method I use.
Read article ↗A campaign can look weak on its own and still sustain sales of other products. I explain how to think about this before cutting budget.
Read article ↗Your maximum cost per lead is the customer acquisition budget multiplied by the rate at which leads become customers. Set that budget using customer value, delivery…
Read article ↗Before saying an ad doesn't work, check whether it actually got enough budget and delivery to be tested properly.
Read article ↗What I record when a campaign or partnership ends to avoid repeating mistakes: hypotheses, limits set beforehand, and what the data doesn't support.
Read article ↗If your offer looks the same as several competitors', the client has little basis to decide with any real criteria. Price ends up carrying most of the weight because…
Read article ↗Before requesting a bigger ad budget, identify where opportunities actually get lost. Sometimes demand is genuinely low; other times enough requests arrive but the team…
Read article ↗When I hire someone to work on campaigns, I want to understand how they think about the business. What information do they ask for before spending money? How do they…
Read article ↗Build a funnel for your service business: offer, landing page, form and sales follow-up, with examples from Sextas Ímpares.
Read article ↗Improve sales follow-up: qualify enquiries, assign responsibilities and prepare the next step. A class with Beatriz Teixeira.
Read article ↗Digital marketing agencies in Porto and nearby: services, current websites and criteria for choosing a team. A list reviewed by Roberto Cortez.
Read article ↗To calculate how many clients you need to win, add expected departures to the growth you want over the same period. Then compare retention scenarios while keeping the…
Read article ↗Keeping a client starts to harm the team and margin when it requires recurring effort that the relationship cannot correct or sustain. Assess additional work, effects on…
Read article ↗To distinguish a delivery failure from misaligned expectations, compare what was agreed, what happened and what the client expected to receive. Seek evidence for each…
Read article ↗Before a cancellation request arrives, ask the client what has changed in their priorities, where they encounter difficulty and which results matter to them. Request…
Read article ↗To understand whether a client recognises the value you deliver, compare the work completed, the observed change and what they find useful. Show examples connected to…
Read article ↗Tell a client about a mistake with a factual account of what happened, the known impact and the action you are taking. Accept your share of responsibility and…
Read article ↗When an offer does not sell, first check who saw it, what they understood and why they chose not to proceed. Limited exposure, unclear messaging, low trust and an…
Read article ↗To test a new offer, present a concrete solution to people with the need you want to address and observe the commitment they make. First define what you want to learn…
Read article ↗Before starting a project, ask the client for the information, availability and decisions that make delivery possible. Link every requirement to a specific need in the…
Read article ↗An entry product needs human support when customers face decisions or difficulties that the materials alone cannot resolve within the promise they bought. Assess the…
Read article ↗Adding a product makes a business harder to manage when it creates work, decisions and expectations that the current operation cannot absorb. Before launching it, assess…
Read article ↗A real agency case shows what happens when the team measures revenue and the client measures ROAS. How to agree on a North Star metric before a project starts.
Read article ↗How I use questions written live by an audience, tied to cost per decision, to decide which problem to solve next.
Read article ↗Reflections on training without execution, being the sales bottleneck, closing profitable projects, and the real value of masterminds, from an interview I asked an AI to run on seven years in business.
Read article ↗How to move beyond purely operational work and become a strategic partner clients won't swap for a cheaper price.
Read article ↗How I separated my talent from operational work and removed repetitive decisions from my schedule, based on the conversation with Andreia Rocha in class 98 of Sextas Ímpares.
Read article ↗What to do with time AI frees up on your team: how to check it's real, avoid scattered minutes, and shift from execution to curation.
Read article ↗I explain how I build a value ladder tied to one customer pain, with real examples of events, books and high-ticket services.
Read article ↗I reflect on when it pays to start with a cheap entry product and when it makes sense to sell high-value services directly, with real examples from my own funnel.
Read article ↗When your schedule is nearly full, taking on one more client has a real cost. A reflection on risk, recurrence and portfolio value from the class on freelance pricing.
Read article ↗Before closing a services client, I weigh business risk, required upfront work, founder maturity and warning signs that can raise the price or make me decline the proposal.
Read article ↗Before any quote, trust is built through the questions you ask and how you define the deliverable. How to work on that perception without relying on followers or titles.
Read article ↗Why I separate a project's launch work from the regular monthly fee, and how I negotiate that with clients who resist paying for the initial phase.
Read article ↗How to separate ROAS from real profit, decide if your business runs on CAC or lifetime value, and find the right bottleneck in your funnel.
Read article ↗Working on client retention in a services business comes down to generating concrete results, communicating regularly and owning up to mistakes with maturity when they…
Read article ↗Why trying to fix everything in the first week of a marketing or automation project sets an expectation you can't sustain, and how to structure the first month by priorities.
Read article ↗Why hourly billing punishes efficiency, and how to anchor your service price in the deliverable, responsibility and risk instead of the clock.
Read article ↗If AI shortens a task, revisit the service built around it: what you deliver, what the client values, and what work still depends on you. In Sextas Ímpares #148, I…
Read article ↗How to calculate your business's cost per decision: time, infrastructure, error and latency, using the lead exercise from class #128 as a starting point.
Read article ↗What I learned about growing an online business: margins, team autonomy, client relationships and applying AI to operations.
Read article ↗