Three skills every team needs to scale agentic AI
Read this guest blog by Silvia Lehnis, Chief AI Officer at UBDS Digital, for techUK’s Tech and Innovation Focus Week 2026.
Skill 1: Frame the right problem
When you can build at machine speed, the expensive mistake is building the wrong thing quickly. So the critical skill becomes user-centred design: the ability to look past what people ask for to what they actually need, how they behave in real conditions, and the root cause underneath. Only then can you judge which problem is worth an agent, and what a good outcome looks like. That also shapes the principles to inform quality evaluation at machine scale, because if a person has to check every output the efficiency gains can be lost. A study of more than 10,000 developers found AI-assisted engineers merge 98% more pull requests, but review time rises 91% and throughput barely moves (Faros AI, 2025). Quality assurance systems must be built in: using machines to run the checks they can do well, people reserved for the edge cases and critical stage-gate reviews – equipped with the right information to be able to meaningfully influence the outcome. This is becoming law as well as good practice. Under the Data (Use and Access) Act 2025 reforms to UK GDPR automated decision-making rules, the ICO has made clear that a human rubber-stamp is not meaningful review.
Skill 2: Manage organisational knowledge
An agent can read documents, intranet pages and knowledge articles, but it cannot tell which are current, authoritative or where the gaps and unspoken knowledge are. Cognizant announced it would deploy 1,000 "context engineers", but the knowledge that matters cannot be fully outsourced to a new job title. It lives with the domain experts who do the work, which makes engineering context part of everyone's role. I have seen this with AIOps: an agent resolving incidents is only as good as the knowledge articles behind it, and where those are stale or missing, it recommends the wrong fix. In government, this is increasingly becoming standard practice. The GDS data asset management policy requires departments to identify their critical data assets, record metadata, name an owner and maintain a data quality plan reviewed annually – a strong policy that is ahead of where most teams are in practice. Data asset management policy in government - GOV.UK
Skill 3: Learn to manage agents
Managing agents is becoming part of the job. Context windows – the information an LLM can process – are finite, so you cannot build one agent to do everything. Without splitting work across agents, each scoped to the context its task needs, machine learning becomes machine forgetting. Knowing how to carve up the work and pull the right decisions back into human hands is a genuinely new skill.
So what can you do today? Build three assets for yourself, and ask an AI to interview you to draw them out – this can be done in as little as an hour. The interview should cover:
Your style: so an agent produces work that reads as your tone and brand.
Your goals: your role, accountabilities, outcomes and priorities, as well as the values that drive your decisions, so it knows when to defer to you.
Your workflows: capabilities and processes you need to reach your goals, the things people come to you for. You can ask AI what it recommends as the top skills given the goals you just provided, build AI workflows for you, identify tools and knowledge required. These are the foundations of your own context layer, and the same three questions scale from a person to a team to a department.
The organisations that lead on agentic AI will be the ones whose people learned to frame the right problem, manage their knowledge in a way that both humans and AI can use effectively, and manage their agents.
These skills are becoming as fundamental to modern organisations as digital literacy became during the cloud era. You can read a more in-depth version of this article here.
