Forget 'stay curious.' Here are 9 things you can actually do this month.
AI upskilling advice usually stops at ‘be curious’ and ‘practice more,’ which is true and useless. Below are 9 specific actions: pick one repetitive task and build a real prompt workflow around it, block a recurring 30-minute practice slot, enroll in one structured course instead of grazing free videos forever, start a one-page prompt library today, verify every output against a primary source before it goes anywhere that matters, find one real use case and pitch it to your manager, trade prompts with two or three peers, cut your AI news sources down to three, and measure one before/after number so you know it actually worked. PwC’s 2026 Global AI Jobs Barometer puts the average wage premium for AI skills at 62%, up from 57% the year before, and Deloitte’s 2026 State of AI in the Enterprise report found insufficient worker skills is the single biggest barrier companies report to getting real value from AI. The gap between those two numbers is basically the point of this article.
Here’s my honest problem with most ‘how to upskill in AI’ content: it tells you how to feel about AI, not what to open on your laptop. Stay curious. Embrace experimentation. Don’t be afraid to fail. I’ve trained more than 2,000 professionals on AI across corporate workshops and bootcamps, and not one of them ever asked me how to feel more curious. They asked what to actually do on a Tuesday afternoon between meetings.
So here are 9 steps, not feelings. Each one is something you can start this week, most of them in under an hour. None of them require a technical background, a budget approval, or permission from IT. A few of them will feel almost too simple to write down, and that’s exactly why most people skip them and stay stuck at the same skill level for years.
The stakes are real, even if the advice around them usually isn’t. PwC’s 2026 Global AI Jobs Barometer, which analyzed more than a billion job postings across 27 countries, found the average wage premium for workers with AI skills has climbed to 62%, up from 57% the year before[1]. Jobs specifically requiring AI skills are growing roughly 69% faster than the job market overall, which is sitting at about 9% growth[1]. That’s not a someday problem. That gap is compounding right now, every quarter you don’t close it.
Here’s the honest part most people skip: the professionals pulling ahead aren’t smarter or more naturally curious. They just did something specific and kept doing it. That’s the entire difference between step 9 below and a New Year’s resolution.
Don’t start by ‘experimenting with AI’ in general. That’s how you end up with forty scattered ChatGPT conversations and nothing to show for any of them. Start by picking one task you do at least weekly, ideally something a little tedious: summarizing meeting notes, drafting the same three types of email, turning messy data into a clean report, or writing the first draft of something you always rewrite from scratch anyway.
Then build an actual workflow around it, not a single clever prompt. Write down the exact instructions that get you a usable result, save that prompt somewhere you’ll find it again, and refine it every time it falls short. After two or three weeks of doing this on one task, you’ll understand more about how these tools actually work than you will from a month of random poking around.
The fastest way to plateau is trying five different tools on ten different tasks and mastering none of them. Depth on one real task beats breadth across a dozen fake ones, every time.
This sounds almost too small to matter, and that’s exactly why it works. Block a recurring 30-minute slot, once a week, labeled something boring like ‘AI practice.’ Treat it like a meeting you can’t skip, not like a task that slides to next week every time something urgent comes up (and something urgent always comes up).
Use that time for the thing you never get to during a normal workday: trying a feature you haven’t touched, revisiting a prompt that half-worked last time, or reading one thing about how a tool actually functions instead of guessing. Thirty minutes a week is 26 hours a year, which is more structured AI practice than most professionals get in three years of ‘staying curious’ whenever they happen to remember to.
Free tutorials are genuinely useful, and also genuinely easy to graze on forever without ever building real depth. A structured course or cohort forces two things a scattered YouTube habit doesn’t: a sequence that builds on itself, and a deadline that makes you actually finish instead of bookmarking it for later.
You’ve got real options here. LinkedIn Learning and Coursera both have solid self-paced AI tracks if a fully independent format works for you. If you want live instruction, a cohort, and a group of people to compare notes with instead of learning alone, Future Factors runs practical, non-technical AI courses and corporate workshops built specifically for working professionals, not developers. I’ll be straightforward about it: it’s one legitimate option among several here, not the only one, and it’s not free. What actually matters is picking a structured format that has an end date and someone besides you checking whether you finished it.
Whichever you pick, the honest test is simple: did you complete it, and can you point to one specific thing you do differently at work because of it? If the answer to either is no, the course wasn’t the problem. The lack of structure around it usually is.
Every professional I’ve trained who’s genuinely good at this has some version of the same document: a running list of prompts that actually worked, organized by task. It doesn’t need to be fancy. A single page in Notion, Google Docs, or even a plain text file is enough to start.
Every time you land on a prompt that gets you a usable result, paste it in with a one-line note on what it’s for. Every time one fails in an interesting way, note that too. In three months you’ll have a personal playbook that’s more useful to you than any generic prompt list you’d find online, because it’s built entirely around how you actually work (for a fuller system for doing this as a team, our guide to building an AI prompt library walks through the team version of the same habit).
Let’s be honest about the one habit that separates people who use AI well from people who eventually get burned by it in front of an audience: checking the output before it goes anywhere that counts. AI tools are confident by default, even when they’re wrong, and a wrong answer delivered smoothly is more dangerous than one that sounds shaky.
This doesn’t mean re-doing all the work yourself. It means treating any specific factual claim, statistic, or quote the same way you’d treat a rumor from a coworker: worth checking against the actual source before you repeat it to your boss, a client, or a room full of people. I learned this one the hard way early in my training career, getting a stat wrong in front of 200 people because I trusted a clean-sounding AI summary instead of opening the original report. It’s a fast habit to build and an expensive one to skip.
This is the step that actually turns personal upskilling into career upskilling, and it’s the one most people never take. Once you’ve built a real workflow in Step 1, don’t keep it to yourself. Bring it to your manager as a specific, scoped idea: ‘I’ve been using this approach for our weekly reports, it’s cutting the draft time roughly in half, can I pilot it more broadly on X.’
Deloitte’s 2026 State of AI in the Enterprise report, based on a survey of over 3,200 senior leaders, found that insufficient worker skills is the single biggest barrier organizations report to actually scaling AI’s value[2]. Managers are actively looking for people who can show, not just tell, how this technology fits their team’s actual work. A specific pitch backed by something you’ve already tested is a different conversation entirely from ‘we should use more AI.’
Learning this in isolation is slower than it needs to be. Find two or three people, coworkers or friends in a similar role, and set up a low-effort habit of trading what’s working: a prompt that saved real time, a tool feature you just discovered, a mistake worth avoiding. A shared Slack channel or a 15-minute call every couple of weeks is plenty.
The value here isn’t just efficiency. Explaining why a prompt worked forces you to actually understand it instead of just having gotten lucky once, and hearing how someone else approaches the same task will surface blind spots you didn’t know you had.
AI news moves fast enough that trying to follow all of it is its own kind of time sink, one that feels productive and mostly isn’t. Pick two or three genuinely credible sources (not a firehose of every AI account on social media) and let everything else go. You do not need to know about every model release the week it happens.
What you do need is enough signal to notice when something actually changes how you should work: a new feature that replaces a workaround you’ve been using, or a real shift in what a tool you rely on can do. Doomscrolling AI headlines feels like staying current. Mostly it’s just noise wearing the costume of productivity.
Before you start any of the above in earnest, write down one concrete number: minutes spent on your chosen task, number of drafts before something’s usable, whatever’s easiest to actually track. Then check it again after 30 days.
This step gets skipped constantly, and it’s the one that turns ‘I feel like I’m getting better at AI’ into ‘I cut my weekly reporting time from 90 minutes to 35.’ The second sentence is the one that gets you taken seriously in Step 6, and it’s the only way you’ll actually know whether any of the other eight steps did anything at all.
If you can’t point to one concrete thing that’s measurably faster, clearer, or better because of how you’re using AI, you’re still in the experimenting phase, and that’s fine. Just be honest with yourself about which phase you’re actually in.
A single weekend of intense AI experimentation followed by three months of nothing is worse than 30 minutes a week, every week. Consistency beats intensity here by a wide margin.
Tool-hopping feels like progress and usually isn’t. Pick the one or two tools your actual job runs on and go deep before you go wide.
Keeping your prompt library and your wins to yourself caps your own growth. The people who improve fastest are usually teaching someone else what they just figured out, which forces real understanding.
None of these 9 steps requires a technical background or a big budget. They require picking one thing and actually finishing it, which is rarer, and more valuable, than it sounds.
Most professionals I’ve trained see a real, noticeable shift after about 30 days of consistent practice, meaning the weekly 30-minute block plus actually using AI on one real task rather than sporadically. It’s not instant, and it’s also not a multi-year project. Consistency over four weeks beats intensity over one weekend every time.
No. Every step above is about how you use the tools, not how they’re built under the hood. PwC’s 2026 data shows the wage premium applies broadly across roles, not just technical ones, and the entry-level jobs seeing the biggest gains are being asked for judgment and leadership skills, not engineering skills.
Free tutorials are genuinely useful for exposure, but most people graze on them indefinitely without ever finishing anything or building real depth. A structured course with a start and end date, whether that’s Future Factors, LinkedIn Learning, Coursera, or another option that fits your schedule and budget, forces the sequencing and accountability that open-ended free content rarely provides on its own.
Treating it as a one-time project instead of a habit. A single intense weekend of AI experimentation followed by months of nothing produces less real skill than 30 focused minutes a week, sustained. The compounding comes from consistency, not from a single burst of effort.
Don’t lead with the request. Lead with a result. Pick one real task, build a working AI approach to it on your own time first, measure the before-and-after (Step 9), and then bring that specific, already-proven case to your manager. Deloitte’s 2026 research found insufficient worker skills is companies’ single biggest barrier to getting real AI value, which means a manager with a concrete, working example in front of them is usually receptive, not skeptical.
This article draws on PwC’s 2026 Global AI Jobs Barometer (an analysis of more than one billion job postings across 27 countries and territories, released 15 June 2026) and Deloitte’s 2026 State of AI in the Enterprise report (a survey of 3,235 senior leaders across 24 countries). All figures were confirmed directly on each organization’s own published page this week. The 9 steps reflect patterns from training 2,000+ professionals across corporate workshops and bootcamps on practical AI skills.
Weekly tips. Real examples. Practical help for busy professionals.
We care about your data, check out our privacy policy.