Automation and profitability¶

You email your internet provider because the router has stopped working. Two minutes later an answer arrives: it has understood the problem, it has opened a ticket with the right technician, and it suggests restarting the router while you wait. Did a person read your email? Probably not.

Somebody decided to pay for that system. Not because it is modern, but because it saves hours of work, and those hours cost money. In September 2026, Telefónica announced that Meliá's technical support, which includes 37 hotels in the Balearic Islands, handles around 240,000 requests a year and already automates more than 70% of them. Today you learn to ask the question every company asks before it buys AI: does it pay? And you will answer it with numbers, in a spreadsheet.

Three steps of automation¶

Automation is doing a task with less human work. Optimisation is doing it better: the best price, the best route, the best moment.

RPA (robotic process automation) is software that copies the clicks and keystrokes a person would make, following fixed rules.

AI does not replace rules; it covers the cases rules cannot reach. In the city-builder game, a timer on each traffic light is a rule; the advisor that watches the traffic overlay and changes the timings is AI.

Where companies use it¶

From data to money¶

flowchart LR
    D["<b>1 · Data</b><br>bookings, cancellations,<br>flights, weather<br>(traffic overlay)"] --> A["<b>2 · Analysis</b><br>a model predicts<br>occupancy per day<br>(the advisor)"]
    A --> DEC["<b>3 · Decision</b><br>prices, staff,<br>purchases<br>(tune the lights)"]
    DEC --> R["<b>4 · Result</b><br>more revenue,<br>lower costs<br>(fewer jams)"]

Without data there is no AI, and with bad data the AI decides badly: garbage in, garbage out. When the data are huge, varied and arrive fast, we call it big data: too much for a spreadsheet, so it needs the cloud from Unit 3 and specific tools. You will study it in Unit 5.

AI can make a company more profitable in three ways: more revenue (sell more, or better), lower costs (time, errors, energy) and lower risk (fraud, breakdowns, fines). But it also costs: collecting and cleaning the data, the cloud service, setting it up and connecting it to existing systems, maintaining the model, and the people who supervise it.

Return on investment (ROI) = (what you gain − what it cost) / what it cost. The payback period is how many months it takes to recover the initial investment.

In class: a real Balearic case¶

In pairs. The question: which company or organisation operating in the Balearic Islands uses AI, or automatic forecasting, in a real process, what exactly does it do, and how do we know? Pick one starting point, or find a case from your own company's sector.

For your case write: company and sector; the process; what the AI or the model does; the number the source gives; the full citation (title, website, date, link); and one line that separates what the source says from what is your hypothesis.

In class: does it pay?¶

In pairs, in Google Sheets (or LibreOffice Calc). The figures are invented for the exercise.

A small software company in Mallorca receives 400 support tickets a month. Each ticket takes a technician 15 minutes on average, and an hour of a technician costs the company €20. They want an AI cloud service that classifies tickets and drafts a reply; they estimate the time would drop to 9 minutes per ticket. The service costs €300 a month, and setting it up (integration and preparing the data) costs €2,000 once.

Step 1 — Build the sheet with formulas, not with results typed by hand. Use only operators (*, /, -), because function names change with the language of the spreadsheet.

Step 2 — Sensitivity. Suppose one reply in ten has to be rewritten, and the real time with AI is 11 minutes. Change only B4. What happens to the payback period?

Step 3 — Your process. Duplicate the tab and model a process of your project company: the one from your Teachable Machine idea, or the Balearic case you found. Every number that does not come from a source goes in a cell with the word hypothesis in column C.

Hand in on Classroom (in pairs, "S17 · A real case and a sheet that pays", by 23:59 the same day): the table of your real case with its citation, and the link to your spreadsheet (shared with the teacher, view only) with both tabs. No generative AI in this activity.

For your project¶

This lesson feeds Deliverable 4 · AI plan:

Key takeaways¶

• Three steps: manual, by rules (scripts, RPA) and with AI, which handles variable, unstructured input.
• Automating means less human work; optimising means better decisions. Each improvement needs a number that measures it.
• The chain: data → analysis → decision → result. Without good data there is no useful AI; big data needs the cloud.
• ROI = (gain − cost) / cost; the payback period says when the investment is recovered. Supervision and errors are costs too.
• A source that says "forecast" is not a source that says "AI". What you cannot cite is a hypothesis.

Quick check¶

You can now say where AI helps a company and whether it pays. Last question of the unit: which sectors use it most, what is it built with, and what does it change for you? Next: Sectors, languages and software.