Date · Friday, 8 May 2026, 13:30 to 14:30 CEST
Hosted by · UNICC
Session page · UNOG learning
Speakers
- Matt Valente · Digital Talent Acquisition and Talent Management Lead, UNICC · Bio
- Olga Lehtinen · HR Development Lead, UNICC · Bio
The UN system is evolving, and so is how it defines talent. As organizations across the Common System pilot skills-based approaches, roster reforms, and AI-assisted screening, the traditional path from Intern to D-1 is becoming less linear and more portfolio-driven. Meanwhile, AI agents are reshaping which functions remain uniquely human and which become augmented or automated. This session explores what these shifts mean for UN staff and candidates navigating interagency mobility, internal reassignments, and long-term career positioning.
Key takeaways
- Capabilities travel; whole careers do not. Build a portfolio of skills you can apply across roles rather than optimising for the next job title.
- Skills-based screening looks for three things together: capability, outputs, and evidence. If a CV bullet does not surface all three, rewrite it.
- AI in screening amplifies what the system is configured to value. As panels move toward evidence-based criteria, years and tenure alone will surface less, not more.
- The screening question is no longer “did you use AI” but “how did you use AI”: intentional use, judgment, transparency, and appropriateness. Build the habit of narrating your process, not just delivering the output.
- AI agents are software that draft, coordinate, analyse, schedule, summarise, and report. The Anthropic financial-services agent release (three days before this session) deployed agents in days, not quarters. Plan for them in your daily software within the next nine to twelve months.
- Apply the “unbundling of your job” filter: which of your tasks are automatable, which are AI-augmented, which are uniquely human?. Move your capacity rightward.
- Job seekers who use AI tools well are hired 18% more often than those who do not (NBER field experiment). This is a competitive edge, not a marginal gain.
Olga Lehtinen
Olga’s contribution was to reset what “career relevance” means inside the UN system today. She named four forces that are reshaping how careers unfold: sustained funding pressure (the now-permanent “doing more with less” reality), AI compressing how work gets produced, escalating geopolitical and operational complexity that demands faster pivots, and the UN 2.0 capability shift toward digital and data fluency. Her central argument is that skills, not roles, are now the currency, and that capabilities are easier to move than whole careers. Concretely, she gave a usable formula for what skills-based hiring actually screens for: capability (what you can do), outputs (what you produce or change), and evidence (what proves it), with signals of digital fluency and adaptability now expected across all three regardless of function. She translated that into specific rewriting moves for applications: replace duty lists with capability language (“synthesised input that informed decision making” rather than “supported preparing meetings”), replace tenure with outcomes (“enabled faster alignment across teams” rather than “five years in programme support”), and replace certificates with digital work habits (“use data and AI tools weekly” rather than “completed analytics training”). She was direct about what this implies for AI screening: AI does not decide what to value, it amplifies what the system is configured to value, so the more the panels and ATS configurations move toward evidence, the less tenure alone will carry an application. Her closing prescriptions were three: start using AI tools every week (treating AI as a colleague that needs to learn what you do); stop assuming that years in grade equal readiness for the next step; and continuously update your CV with evidence-based stories so you keep visibility on what you have actually done. She also flagged a sustainability angle: type -AI in Google to suppress the AI summary, and recognise that some tasks do not need an AI tool.
Matt Valente
Matt’s contribution was to demystify AI agents and to show, with concrete numbers, how the assessment side of recruitment is changing. He started with a practical example: at UNICC and elsewhere, technical tests are now being run inside AI sandboxes where the panel can see the prompts, the iterations, and the judgment applied. The differentiator is no longer whether you used AI but how, with explicit signals being looked for around intentional use, judgment, transparency, and appropriateness (not blindly trusting outputs). He then collapsed the futurism around “AI agents”: an agent is just software that drafts, coordinates, analyses, schedules, summarises, and reports, increasingly autonomously but typically with a human in the loop. To make the timeline real he cited Anthropic’s release of ten agents across financial services three days prior to the session, including agents that build pitch books, prepare client decks, read earnings transcripts, and flag model updates, deploying in days rather than months. The implication, framed through the Financial Times’ “unbundling of your job” lens, is that any job description can now be split into automatable tasks (drafting, retrieval, formatting), AI-augmented tasks (synthesis across workstreams, stakeholder management, analysis), and uniquely human tasks (judgment, ambiguity, diplomacy, ethics). His practical move is to apply this filter to your own current tasks and JD and ask where they sit, with the goal of shifting capacity rightward. On hiring data he cited a National Bureau of Economic Research field experiment showing job seekers who used AI tools well were hired 18% more often than those who did not, framing AI fluency as a meaningful competitive edge rather than a marginal one. On how to build that fluency, his counterintuitive but consistent advice was to use the AI tools themselves to learn (ask Claude or ChatGPT what you do not know, where to go deeper, and how to practise), supplemented by the official training paths from OpenAI, Anthropic, Google, and Microsoft, and by builder tools like Lovable for hands-on practice (he built the session’s AI Skills Shift assessment in Lovable). His ethical and risk guidance was equally concrete: redact personal data when applying for roles, avoid putting confidential information into non-enterprise tools, and assume that anything not behind an enterprise licence is a private company with bad-actor exposure. On portability of AI evidence, he advised treating AI work as a portfolio you can talk to in motivation statements and duty descriptions: name the tool, name the workflow, name the impact, but generally do not paste a URL because recruiters will not click it.
Frameworks and models
| Name | What it stands for | How to use it |
|---|---|---|
| Capability + Outputs + Evidence | The three components AI screening looks for in a candidate: what you can do, what you can produce or change, and what proves it. Digital fluency and adaptability signals are now expected across all three. | Use as a rewrite test on every CV bullet and motivation paragraph. If a sentence does not name a capability, an output, and a piece of evidence, it is not yet ready. |
| AI Use as a Skill (Intentional Use, Judgment, Transparency, Appropriateness) | The four signals panels increasingly look for when assessing how a candidate uses AI in tests and tasks. | Use as a self-check before delivering AI-assisted work: did you intend the use, exercise judgment on the output, narrate it transparently, and choose AI where it was appropriate? |
| Unbundling of Your Job (Automatable / AI-Augmented / Uniquely Human) (note: this article is protected by paywall; not extracted as a standalone framework page on the explicit instruction not to reverse-engineer paywalled content. The lens is referenced inside Skills-First Approach and Capability Frontier.) | The Financial Times framing in which any job description can be split into tasks that AI can automate (drafting, retrieval, formatting), tasks AI augments (synthesis, stakeholder management, analysis), and tasks that remain uniquely human (judgment, ambiguity, diplomacy, ethics). | Apply to your current JD and weekly tasks. For tasks on the left, decide what to delegate to AI. For the middle, decide how to amplify with AI. For the right, decide how to deepen as your differentiator. |
| Builder) | The maturity scale used in UNICC’s AI Skills Shift assessment, mapping how often and how deeply a person uses AI across capability areas. | Take the assessment to locate yourself, then target one or two adjacent moves (e.g., from Adopter to Practitioner in process automation) rather than trying to advance everywhere. |
| Skills Inflating vs Deflating in Value | The pattern that judgment, sense-making, stakeholder navigation, and political/organisational acuity are inflating in value, while routine drafting, formatting, and information retrieval are deflating. Documented as the “Inflating vs deflating skills” section of the Skills-First Approach page. | Use as a filter when choosing which skills to deepen and which to delegate to AI. Invest learning time on the inflating side. |
| Capability Language (replacing duty lists with capability + outcome) | Olga’s rewrite move: replace “I prepared meetings” with “I synthesised input that informed decision making and aligned stakeholders”. Documented as the “capability-language rewrite move” section of the Skills-in-Use CV Pattern page. | Apply line by line on CV and cover-letter drafts before submission. |
| 20-Minutes-a-Week Rhythm | Matt’s prescription for continuous AI learning: a fixed weekly slot on one tool or one new feature, instead of one-off training courses. Documented as the “20-minutes-a-week rhythm” section of the Reframe, Adapt, Lead page. | Schedule it as a recurring calendar block. Pick a tool or feature on Monday; write up what you learned and what changed in your workflow on Friday. |
Resources
| Resource | What it is / What it’s for | Link |
|---|---|---|
| AI Skills Shift Test (UNICC) | Self-assessment that maps your AI usage across the capability frontier (Explorer / Adopter / Practitioner / Builder) and produces a personalised action plan. | https://ai-skills-shift.lovable.app/?session=R9PMFP |
| UNICC AI Hub | Official UNICC page on its AI work, covering governance, services, and the persona-based learning approach (consumer vs practitioner) that underpins how UNICC frames AI fluency across the UN system. | https://www.unicc.org/artificial-intelligence/ |
| Anthropic learning resources | Official training and webinars on prompting, agents, and skills, from the makers of Claude. | https://www.anthropic.com/learn |
| OpenAI learning resources | Official training resources from the makers of ChatGPT. | https://platform.openai.com/docs |
| Google Generative AI learning path | Structured Google Cloud training on generative AI. | https://www.cloudskillsboost.google/paths |
| Microsoft AI Skills | Microsoft’s AI skills learning programme. | https://learn.microsoft.com/en-us/training/ai/ |
| UN 2.0 | The UN Secretary-General’s initiative to upgrade the system’s capabilities for the 21st century, with a strong digital and data dimension. Olga recommended familiarising yourself with it. | https://un.org/two-zero |
| NBER field experiment on AI-assisted job seekers | The study Matt cited showing job seekers who used AI well were hired 18% more often than those who did not. | https://www.nber.org/ |
| Anthropic financial-services agents announcement | The release Matt referenced (three days before the session) of ten production AI agents for the financial services industry, deploying in days. | https://www.anthropic.com/news |
Last updated 2026-05-10.