Choose a realistic route

What you will be able to do

Choose a target role that you can actually reach, with a reason that you can explain. A realistic choice uses what you already have, closes a gap that fits your time, and has a second option in case the first takes longer.

Before you startWhat to know and to have readyAbout 35 minutes

What you need to know

• The job families and the routes around them, from The main job families: Six job families and Technical and adjacent routes: How much code does each route need?.
• Why evidence is better than a self-rating, from Evidence over self-ratings: Two kinds of evidence.

What you need

• Your inventory with evidence and your two main gaps, under ## Inventory with evidence in notes.md. You wrote them in Evidence over self-ratings: Guided practice: update your inventory.
• Know how many hours each week you can study, honestly. Include weeks with holidays, family and work peaks.
• An internet connection, to find five real job advertisements in the guided practice.

Four common starting points

People come to AI from many places. Here are four common starting points. Find the one closest to you. If you are between two, read both.

Fresh graduate

• Usually has: recent study habits, time, and some maths or programming from university. Field of study as a small domain.
• Usually lacks: experience of delivering work with a team, and evidence that employers trust.
• Realistic first roles: junior data analyst, junior data engineer, junior AI engineer. A research role is realistic mainly after a master’s degree or a PhD.
• Add first: one complete project that works for a real user, and team skills such as Git and code review.

Software developer

• Usually has: programming at level 3, version control, testing, deployment, and experience in one industry.
• Usually lacks: statistics, and the habit of measuring quality instead of checking only that code runs.
• Realistic first roles: AI engineer, ML engineer, data engineer. Often inside the current company.
• Add first: how language models behave, how to evaluate their outputs, and enough statistics to compare results.

Analyst

This includes data, business and financial analysts.

• Usually has: SQL and spreadsheets, statistics in practice, communication with people who decide, and domain knowledge.
• Usually lacks: programming beyond analysis, and building systems that run every day.
• Realistic first roles: data scientist, evaluation specialist, analytics engineer (an analyst who builds reliable data tables).
• Add first: Python, machine learning basics, and one project that goes beyond a report.

Professional from another field

This includes accountants, nurses, lawyers, teachers, store managers, support leads and engineers from other industries.

• Usually has: demonstrated domain knowledge, communication and project delivery. These are three of the five areas.
• Usually lacks: AI literacy and programming.
• Realistic first roles: subject-matter expert on an AI project, evaluation specialist, AI adoption lead, or AI product manager in their own field. Technical roles in their field, with more time.
• Add first: AI literacy. Then decide between the three ways from Your experience is a head start: contribute, lead or build.

The phrase “in their own field” is important. A nurse who moves to a health AI company has rare knowledge there. The same nurse in a general software company competes with everybody else on programming alone.

Primary role and adjacent option

Choose two things:

• A primary role: the role that you aim for first.
• An adjacent option: a related role that uses most of the same skills. It is useful if the primary role takes longer than you expect, or if you find that you prefer the adjacent work.

Good pairs share most of their skills. AI engineer and data engineer share programming. Evaluation specialist and AI product manager share AI literacy and domain knowledge. A pair such as “researcher and AI product manager” shares very little, so it divides your time instead of giving you a second chance.

How to choose: four questions

For each role that you consider, answer four questions:

1. Do I want to produce what this role produces? Look again at The main job families. You produce it every day, so this matters most.
2. Does it use my strongest evidence? A role that uses your “demonstrated” areas starts with an advantage.
3. How big is the gap, and does it fit my time? Count the course hours, and divide by your hours each week.
4. Can I find this role near me, or remotely? Look at 10 real job advertisements for it. Read the tasks, not only the title.

Worked example: Daniela chooses

Daniela, the accountant, can study 6 hours each week. She considers two roles.

Question	AI product manager for finance tools	Data analyst in finance
1. Wants to produce it?	Yes: decisions and products that help finance teams.	Partly: she likes analysis, but misses working with people.
2. Uses her strongest evidence?	Yes: domain, communication, project delivery.	Yes: domain and spreadsheets.
3. Gap and time	AI literacy, evaluation and some Python.	Python, SQL and statistics.
4. Can she find it?	Fewer advertisements, but her company plans an invoice AI project.	Many advertisements.

Then she calculates the time, with the course hours from the course catalogue:

• Product manager route: Mathematics and statistics (12–16 hours), LLMs (8–12 hours), Python (16–22 hours), LLM applications (12–16 hours), Evaluating AI systems (12–16 hours). Total: 60–82 hours. At 6 hours a week, that is about 10–14 weeks of courses.
• Analyst route: Python, Mathematics and statistics, SQL. Total: 40–54 hours, about 7–9 weeks.

Her decision:

Primary role: AI product manager for finance tools. It uses my domain knowledge and project experience, and my company’s invoice project gives me a place to start, as the subject-matter expert. Adjacent option: data analyst in finance. The two routes share Python and statistics. If the product role takes longer, the analyst role uses the skills that I have learned by then.

She starts with the courses that both routes share: Python and Mathematics and statistics. That keeps both options open for the first two months.

Three shorter examples

Person (fictional)	Starting point	Primary role	Adjacent option	Why
Arjun	Fresh graduate in computer science	AI engineer	Data engineer	Strong programming. Both roles build software. He needs one complete project and team skills.
Lucía	Backend developer for six years, in logistics	AI engineer in logistics	ML engineer	Level-3 programming plus domain knowledge. She adds LLM behaviour, evaluation and statistics.
Kofi	Business analyst in a telecom company	Data scientist	Evaluation specialist	SQL, statistics and stakeholders already demonstrated. He adds Python and machine learning.

Guided practice: your choice

1. In notes.md, add a heading: ## My route.
2. Write your starting point, and your study hours each week.
3. Choose two roles that you consider. Answer the four questions for each, in a table like Daniela’s.
4. For question 4, find and read at least five real job advertisements for your top role. Write how many of them match the work that you expected.
5. Write your primary role and your adjacent option, each with one sentence of reason.
6. Write which courses both options share. You start with those.

Check: your reason uses your evidence from the last lesson, not only your interest. “I like AI” is not a reason. “It uses my demonstrated domain knowledge and communication, and my gap is 60 hours” is.

If you are not sure which courses a role needs, use the learning path recommendation. It shows a route for your goal, and the courses that you may skip.

Common mistake

Choosing the most famous role

Symptom: you choose “ML researcher” or “AI scientist” because it sounds like the centre of AI. Your plan says two years of maths and programming before your first application. After three months, you stop.

Cause: the choice came from fame, not from the four questions. The gap was too large for the time, and the role did not use any of your strengths.

Fix: answer the four questions honestly. Choose a primary role that you can reach in months, not years, and that uses your evidence. You can move toward a more distant role later, from inside the field. That is easier than from outside it.

Check your understanding

A teacher with 15 years of experience chooses 'machine learning engineer at any company' as the primary role. What would you suggest, and why?

Show the answer

Consider a role that uses the teaching experience, such as AI product manager or evaluation specialist for education technology, or learning design for AI tutors. There, 15 years of teaching is rare evidence. As an ML engineer at “any company”, the teacher competes only on programming, the weakest area, and loses the head start. If they really want to build, “ML engineer in education technology” keeps the domain advantage.

Check your understanding

Why does Daniela start with the courses that both of her options share?

Show the answer

Because they are useful for both roles. If she changes her mind after two months, no study time is lost. It also gives her time to learn more about each role, for example in her company’s invoice project. Then she chooses the courses that only one role needs.

Recap

• Fresh graduates, developers, analysts and professionals from other fields each start with different strengths and gaps.
• Professionals from other fields are strongest in AI work in their own field.
• Choose a primary role and an adjacent option that share most of their skills.
• Use four questions: the output that you want, your strongest evidence, the gap against your time, and real advertisements.
• Start with the courses that both options share.

Next, in Module 3, you see a complete AI project from the inside, starting with how to frame the problem.

Next: Problem framing