What happens to career guidance when people ask AI first?

A person who is thinking about changing career no longer has to wait for an appointment to begin exploring their options. They can open a chatbot and ask what jobs might suit them, what training they need, or whether their existing skills could be useful in another sector.
The answer may be helpful. It may also be incomplete, out of date or confidently wrong. Either way, by the time that person reaches a careers adviser – if they do – their conversation about the future may already have started somewhere else.
That is the question raised by Jaana Kettunen, Professor of Lifelong Guidance at the University of Jyväskylä, in a recent post following the Nordic-Baltic Summit during Scottish Careers Week. “What happens to career services when people ask AI first?” she asks. 1
Kettunen reports that, ahead of the summit, participants identified opportunities for wider access and less routine work for practitioners. Some also observed that people were already consulting AI before contacting a careers service. Her concern is that fewer responses considered what this means for the careers system as a whole. How should services evolve so that AI extends access while human support remains an integral part of guidance?
From tools to systems
Kettunen’s question did not come out of nowhere. In 2025 she wrote with Sally-Anne Barnes and Hanna Pullinen about the use of digital technology in vocational education and training across Europe. Drawing on a comparative study presented at the Skills2Capabilities conference, they described the development of approaches that combine digital and personal support. They also called for investment in digital inclusion, practitioners’ skills and cooperation between organisations. 2
In related work, Kettunen and Barnes looked at how guidance can be connected to other services. Their examples included Finland’s Ohjaamo centres, which bring together different forms of support for young people under 30, and regional cooperation between careers, vocational education and employment services in Norway. These examples illustrate that people do not experience career decisions as neatly separated from education, work and the rest of their lives. 3
General-purpose AI tools have become a visible part of everyday life. Careers services now need to think not only about tools they select or build, but also about the tools people encounter outside the service, over which the service may have little control.
This doesn’t mean that everyone asks AI first, or that an AI-first journey is better. The Ada Lovelace Institute’s 2026 review of AI in career guidance found that some young people in the UK use general-purpose tools for career exploration. It also found uncertainties about accuracy, equity and the role of human guidance. Its evidence does not establish how often people consult AI before an adviser, or whether doing so improves their eventual decisions. 5
A plausible answer is not the same as guidance
Imagine an adult working in a declining industry who wants to retrain. They ask a chatbot what they could do with their experience. The response identifies some potentially useful transferable skills and suggests three occupations. It might even give them the confidence to begin looking at courses.
But it could also recommend a qualification that is not offered locally, overlook the cost of training, or assume the person can study full-time. It may know little about their caring responsibilities, disability, income, transport or previous experience of education. A fluent answer can make those omissions hard to see.
This is not an argument against using AI. A careers adviser can make mistakes or overlook possibilities too. The difference is that good guidance provides space for a person to question the advice, explain their circumstances and explore options over time. It can involve an understanding of local opportunities and the practical support needed to pursue them.
Kettunen and Päivi Häkkinen make a related point in a 2026 article for guidance professionals. They identify the usefulness of chatbots for routine questions and signposting, while warning that unreliable information, unrepresentative data and recommendation systems can distort or narrow the options presented. Their emphasis is on practitioners understanding and on critically evaluating the tools, rather than accepting their outputs at face value. 6
Career guidance is not simply a matter of finding the job that best fits a profile. It can help people discover possibilities they had not considered, question assumptions about what is available to them, and decide what matters in their working lives.
Making the connection to a person
If someone begins with a chatbot, how do they get from there to a conversation with a qualified adviser?
In a reply to Kettunen’s post, Trent Loader makes an important observation: people should not have to diagnose the complexity of their own needs before they can reach human guidance. An apparently straightforward question may be the beginning of a much more difficult decision. 1
That has practical implications for careers service design. A careers service’s own digital tool could make it easy to find an adviser throughout the interaction and offer a choice of contact methods and explain what happens next. A person who wants human support from the beginning should be able to choose it without first having to complete an AI conversation.
The connection is harder when someone uses a general-purpose chatbot that is not part of a careers service. There may be no handover at all. A practical response does not have to start with an expensive integration: advisers can invite people to bring along AI-generated suggestions if they choose, examine them together and check important claims against current course and labour-market information.
What the person shares should remain their choice. An adviser does not need a full transcript of a private conversation to offer good guidance. Any service that does collect or transfer personal information needs to explain what it collects, why it is needed and who can see it.
Access is more than availability
AI may make an initial conversation possible outside office hours, for people in remote areas, or for those who find it easier to start by typing. These are worth exploring. But a service is not inclusive simply because it is available online.
People need suitable devices and connectivity. They may need accessible interfaces, support in a preferred language, or help interpreting an answer. Free versions may differ from paid ones. Some people may prefer not to use AI at all. And a recommendation that looks personalised may still reflect assumptions about gender, disability or the kinds of work thought suitable for a particular group.
Rachel Buchanan’s research on AI and career access makes the wider social-justice issue clear: AI is becoming relevant not only to guidance but also to recruitment and selection. Guidance professionals have a role in helping people understand these systems and in challenging practices that reproduce existing inequalities. 8
This matters especially in vocational and adult education. Learners may be moving between jobs, returning to learning after a long break or trying to have skills gained through work recognised. A useful system should help them explore these possibilities, not steer them towards a narrow set of familiar occupations because those are what its data most readily describes.
The ethical principles are already being discussed within the profession. The International Association for Educational and Vocational Guidance’s guidelines call for transparency about the use and limitations of AI, explicit consent and an option to decline, protection of privacy, professional judgement and equitable access. Those are professional guidelines, not proof that any particular tool meets them. 9
What would a better service look like?
A better service would begin by asking what people need from guidance, rather than starting with the technology. For some, an independently used digital tool might be enough to find a course or prepare questions. Others may want a longer conversation, practical assistance or support from several organisations. People should be able to move between those forms of help without being treated as failures of the digital system.
It would also give practitioners time and training to work with the questions people bring. That includes checking AI-generated information, discussing uncertainty and helping learners make their own judgements. If AI reduces some administrative work, the resulting time needs to be protected for guidance—not assumed to appear automatically.
And it would be evaluated on more than the number of chatbot conversations. Who reaches an adviser? Who is unable or unwilling to use the digital option? Do learners understand their choices better? Can they identify suitable training and make progress towards goals they have chosen? Are some groups getting a noticeably poorer service? These are harder questions than counting clicks, but they are closer to the purpose of guidance.
The Ada Lovelace Institute recommends further research into different combinations of AI and human guidance, including their effects on decision-making, understanding and outcomes for different groups. That is a useful reminder that the right balance cannot be settled by a product demonstration or an appealing slogan. 5
Jaana Kettunen’s question is therefore not simply whether AI can answer career questions. It is whether career services can remain visible, accessible and accountable when the first answer may come from somewhere else. The opportunity is to connect new ways of exploring with the human relationships, professional expertise and public services that help people turn possibilities into choices.
References
Attwell, G. (2024, February 23). AI and career counseling, advice, information and guidance and generative AI. AI Pioneers.
Buchanan, R. (2025). Exploring the role of AI in career access through a social justice lens. Nordic Journal of Transitions, Careers and Guidance, 6(1), 1–13. https://doi.org/10.16993/njtcg.77
International Association for Educational and Vocational Guidance. (2025). IAEVG ethical guidelines. https://iaevg.com/Ethical-guidelines
Kettunen, J. (2026, November 11). What happens to career services when people ask AI first? This was one of the questions I explored at #NordicBalticSummit26 [Post]. LinkedIn. https://www.linkedin.com/posts/jaanakettunen_nordicbalticsummit26-scotcareersweek26-activity-7526338344310480896-kSh
Kettunen, J., & Barnes, S.-A. (2025, July 17). National practices supporting VET systems: Multi-agency collaboration. University of Jyväskylä. https://www.jyu.fi/en/blog/national-practices-supporting-vet-systems-multi-agency-collaboration
Kettunen, J., Barnes, S.-A., & Pullinen, H. (2025, December 9). How technology is transforming VET career guidance across Europe. University of Jyväskylä. https://www.jyu.fi/en/blog/how-technology-is-transforming-vet-career-guidance-across-europe
Kettunen, J., & Häkkinen, P. (2026). Demystifying artificial intelligence: Key concepts for guidance professionals. IAEVG Magazine, 99, 11–14. https://urn.fi/URN:NBN:fi:jyu-202608186616
Loader, T. (2026, November 12). An important shift in the question, Jaana. If AI becomes the first point of contact, service design must consider how [Comment on a LinkedIn post]. LinkedIn. https://www.linkedin.com/posts/jaanakettunen_nordicbalticsummit26-scotcareersweek26-activity-7526338344310480896-kShT
Organisation for Economic Co-operation and Development. (2024). Digital technologies in career guidance for youth: Opportunities and challenges. https://www.oecd.org/en/publications/digital-technologies-in-career-guidance-for-youth_c9ab23da-en.html
Tanner, E., & Samson, R. (2026, April 13). Navigating the future: A landscape review of AI in career guidance for young people. Ada Lovelace Institute. https://www.adalovelaceinstitute.org/report/navigating-the-future/
About the Image
Image generation systems resemble a Rubik’s Cube, whose hidden facets represent data points and invisible human labour. To solve it, the algorithm continuously adjusts connections. For people sharing generated images on LinkedIn, mentioning ‘cr’ acknowledges the algorithm that solved the cube. But why play this game? Plugged into overheated data centres, the cube’s frenetic rotation consumes vast amounts of energy. The spectre of electronic waste looms large as the rapid expansion of AI systems renders older devices obsolete. This image was created in Paint, remixing Unsplash images, a personal e-waste photo and the LinkedIn C2PA Content Certification icon.
