Everything you need to navigate AI-driven change at work

Before you can assess your own exposure to AI automation, you need a clear model of what AI systems are actually good at and what they are not. The current generation of AI — large language models, image generators, code assistants, and the tools built on top of them — is genuinely impressive at a specific set of tasks: processing and summarising large volumes of text, generating plausible-sounding content in a given style, writing and debugging code, recognising patterns in structured data, and translating between formats. These capabilities are real, they are improving quickly, and they are already changing how work gets done in a wide range of fields.

What AI systems are not good at is equally important to understand. They do not have genuine understanding of the content they produce — they generate statistically plausible outputs based on patterns in training data. They cannot exercise judgment in situations that require weighing values, taking responsibility for outcomes, or navigating genuinely novel circumstances. They do not have relationships, cannot build trust, and cannot be held accountable. They are not creative in the way that humans are creative — they recombine existing patterns rather than generating genuinely new ideas from cross-domain experience. And they are not reliable in high-stakes situations where errors have serious consequences, because they have no way to know when they are wrong.

This matters for career planning because the tasks most exposed to automation are those that map onto AI's strengths: routine information processing, content generation in established formats, pattern recognition in structured data, and translation between systems. The tasks least exposed are those that map onto AI's weaknesses: high-stakes judgment, genuine relationship-building, creative synthesis across domains, ethical reasoning, and accountability for outcomes.

The most common mistake people make when thinking about AI and their career is to reason at the job-title level. 'Will accountants be replaced by AI?' is not a useful question. 'Which of the specific tasks I perform as an accountant are most exposed to automation, and which are most durable?' is a much better one.

To assess your own exposure, start by mapping your working week at the task level. Not your job description — your actual week. What did you spend your time on? For each task, ask three questions: First, is this task primarily about processing information that already exists, or does it require generating new insight from incomplete information? Second, does this task require genuine relationships and trust, or could it be performed by someone — or something — with no prior relationship with the people involved? Third, does this task require taking responsibility for an outcome, or is it primarily about producing an output that someone else will evaluate and act on?

Tasks that score high on information processing, low on relationship requirements, and low on accountability are the most exposed. Tasks that score low on information processing, high on relationship requirements, and high on accountability are the most durable. Most roles contain a mix of both — and the mix is what matters.

The AI Exposure Audit on this site is designed to help you do this assessment systematically. But you can also do a rough version yourself in about twenty minutes with a piece of paper and an honest willingness to look at how you actually spend your time.

Once you have a clear picture of your exposure, the next question is what to build. The answer is not 'learn to code' or 'get an AI certification' — though both of those may be useful in specific contexts. The answer is to invest in the capabilities that are becoming more valuable as AI capabilities expand, and to do so in a way that builds on your existing strengths rather than starting from scratch.

The capabilities that consistently hold their value across industries and roles are: complex judgment in high-stakes situations, genuine relationship-building and trust, creative synthesis across domains, ethical reasoning and accountability, adaptive leadership through change and uncertainty, and deep domain expertise that allows you to evaluate AI outputs critically. These are not soft skills in the dismissive sense — they are the hardest skills to develop and the hardest to replicate, which is precisely why they are valuable.

Building these capabilities is not primarily about taking courses. It is about deliberately seeking out the kinds of experiences that develop them: taking on projects with genuine stakes and accountability, building relationships with people who are different from you, working in domains adjacent to your own, and putting yourself in situations where you have to exercise judgment without a clear right answer. The professionals who develop these capabilities most effectively are those who treat their career as a deliberate practice rather than a series of jobs.

One of the most underrated aspects of career resilience in the AI era is the ability to have clear, honest conversations about your role and your value — with your manager, with your team, and with yourself. Many professionals avoid these conversations because they are uncomfortable, but the discomfort of having them is much smaller than the cost of not having them.

The conversation worth having with your manager is not 'Am I going to be replaced by AI?' It is: 'Where does the value in my role sit, from your perspective? Which parts of what I do are most important to the team and the organisation? And are there parts of my role that are changing in ways I should be aware of?' These questions are professional, forward-looking, and demonstrate exactly the kind of self-awareness and initiative that makes someone more valuable, not less.

The conversation worth having with yourself is harder. It requires honest assessment of where your current skills and experience sit on the exposure spectrum, what you have been avoiding building because it is difficult, and whether the career path you are on is taking you toward more durable work or less. Most people know the answers to these questions at some level — they just have not made the time to sit with them properly.

Career planning in the AI era requires a different time horizon than most people are used to. The decisions that matter most — which skills to build, which relationships to invest in, which direction to move in — take years to compound. The news cycle about AI moves much faster than that, which creates a constant temptation to react to the latest development rather than executing on a longer-term plan.

A useful five-year career plan in the AI era has three components. First, a clear picture of your current exposure and the trajectory of your role — not just where it is today, but where it is heading. Second, a specific set of capabilities to build, chosen because they are durable, because they build on your existing strengths, and because they are relevant to the direction you want to move in. Third, a set of relationships to invest in — people who are doing the work you want to be doing, people who can give you honest feedback, and people who can open doors that are currently closed to you.

The most important thing about a five-year plan is not that it is correct — it will not be, because the landscape will change in ways you cannot predict. The most important thing is that it gives you a basis for making decisions consistently over time, rather than reacting to each new development as it arrives. Professionals who have a clear direction and execute on it consistently almost always outperform those who are more reactive, even when the direction turns out to need adjustment.