Let me introduce you to a person I will call the planner. She works in operations at a regional distributor, she is good with spreadsheets, and she has quietly decided she wants a job in artificial intelligence within eighteen months. She has evenings, a modest budget and a lot of tabs open. Every guide she finds tells her to learn everything: linear algebra, five frameworks, three clouds, a portfolio of ten projects. She closes the laptop feeling slightly worse than when she sat down. This post is for her, and if you are somewhere near her position, it is for you too. The idea is simple: an AI career plan built for one real person beats a syllabus built for everybody.
Why the “everything” roadmap fails
Roadmaps aimed at a crowd have to include every route, because the author cannot know which one you need. That makes them accurate and useless. A person with a statistics background and a person who has only ever managed rotas need very different first months, yet they are handed the same list.
When you plan for one person, a few questions collapse the list quickly. What does she already do well? What does she worry about? How many hours a week can she truly protect? Answer those honestly and half the roadmap disappears, which is a relief rather than a loss.
Step one: describe yourself the way a friend would
Take ten minutes and write three or four sentences about yourself as if explaining to a friend over coffee. For our planner it might read: “I understand how stock moves through a warehouse. I clean messy data every week. I do not like maths that looks like Greek letters. I can give about five hours a week, mostly on weeknights.”
That paragraph is more valuable than any aptitude quiz. It points to roles where her existing strengths count, such as data quality work, analytics support, or product operations for a machine-learning team. It also names a boundary, the dislike of heavy notation, so she does not sign up for a theory course in week one and abandon it by week three.
Step two: choose the first skill by its payoff for her
A skill is worth learning first if it connects to something she already touches. For the planner, tidying and checking data is the obvious thread. Learning to profile a dataset, spot odd values and document what she changed is real work that teams in this field value, and she can practise it on the files she already handles.
Contrast this with picking the trendiest topic of the month. That is writing a plan for a stranger. A better test is the “Monday question”: after this week’s study session, what can she do differently on Monday at her actual desk? If the answer is nothing, postpone the topic.
- Pick one skill that links to her current job.
- Pick one small task where she can use it at work.
- Decide what “done” looks like in two sentences.
Step three: learn how decisions get made with these tools
Beginners often assume AI work is mainly about building models. Much of it is about helping people use outputs sensibly. Reading a piece such as our overview of how machine intelligence supports business choices gives a feel for where analysts, product people and operations staff sit around a model, and that is where the planner’s background is a genuine advantage. She knows what a late shipment costs. A model builder fresh from a course often does not.
Likewise, understanding the limits matters as much as understanding the capability. A predictive tool can be useful and still be wrong at the edges. If you want a gentle primer on that tension, this plain explanation of how computers estimate what comes next is a good evening read. Being able to say “here is what this forecast can tell us, and here is where I would not rely on it” is a skill employers notice.
Step four: make your learning visible to one reader
Here is the part many career-switchers skip. Writing about what you learn forces clarity, and it also leaves a trail that a hiring manager can follow. The trick is the same as everywhere else in this post: write each entry for one person. Maybe it is the version of you from last month, who did not know what a validation set was. Maybe it is a colleague who keeps asking how your course is going.
Where you publish matters less than the habit, but it helps to have a calm place of your own. One option is iBlogUp, a publishing site that gives each writer a subdomain of their own and lets readers follow along or subscribe through a feed. Its own pitch is that it costs nothing, which suits a learner on a tight budget. Another option is a plain notes page on a personal site. Choose whichever you will actually update.
A good learning post for one reader has a recognisable shape: the problem she had, what she tried, what broke, what she would do next time. No jargon she cannot explain. No claims she cannot support. That is also exactly the kind of evidence that persuades an interviewer, because it shows thinking rather than a list of certificates.
A realistic weekly rhythm for five hours
Consider how the planner might spend a typical week:
- Two sessions of one hour each on the skill of the month, with a tiny exercise after each.
- One session of ninety minutes applying it to real, non-confidential work data or a public dataset.
- One short session writing up what happened, aimed at that one imagined reader.
- One lighter session reading or listening to something broader, so she keeps the wider field in view.
Notice what is missing: a marathon weekend, a lot of money, a heroic mindset. A plan that survives a tiring Thursday is a plan that works.
What to say when someone asks, “So what do you do in AI?”
You will be asked, probably by a relative. Practise answering in one sentence for one listener. “I help teams trust their data before it feeds a model” lands better than a list of tools. The same exercise helps with interviews: picture the person across the table, think about what keeps her up at night, and speak to that. For an operations manager that is cost and reliability; for a startup founder it is speed and focus.
Common wobbles, and what to do about them
Three wobbles turn up almost every time somebody starts a career change.
Comparison. You will see people who seem to be years ahead. Their path was built for their life. Go back to your own paragraph from step one and read it again.
Tool churn. New products appear weekly. Ask whether a new tool changes what the reader of your work needs, or merely changes the buttons. Usually it is the buttons.
Silence. Without a teacher or a team, progress can feel invisible. That is another reason to keep a record in public or semi-public form. Even a handful of readers who follow along make the effort feel witnessed.
Checking in after three months
At the quarter mark, reread your plan and ask four plain questions. Did I protect most of the hours I promised? Did the first skill show up in my day job? Can I explain one thing I learned to a friend in under a minute? Do I still like this? If the answers are mostly yes, extend the plan by another quarter and add one new skill. If not, adjust the plan, not yourself. A change of direction after twelve weeks is information, not failure.
The planner, by the way, does not need to become a research scientist to belong in this field. There is room for people who clean, check, translate and explain. What gets her there is not a bigger syllabus. It is a smaller, truer picture of herself, a few honest hours a week, and one reader to write for, who may be one step behind her on the same road.