Agent skills: build a leadership command centre from one AI chat

Agent skills let you turn one good AI answer into a standard output your whole team can reuse. At aiblLIVE London 2026, Gianna Gard of Dust showed how, building a leadership command centre live on stage.

She took one request, a competitor battle card, from a single chat to a skill. Then she made it an agent that runs every Monday at 7am. This page sets out the method. The full guide adds the prompts she typed, a starter brief and a checklist.

What is an agent skill?

An agent skill bundles instructions, data and tools into one reusable package, so an agent produces the same standard of output every time. Gard described it as a Markdown file that captures how your company wants a task done. It can run on a schedule or when an event happens.

A skill is where we’re bundling a combination of data and tools together with a set of instructions, in order for us to create standards for our company.

Gianna Gard, AI Deployment Strategist, Dust

How do you turn an AI chat into a skill?

Get one good answer first, then ask for a skill that reproduces it. Gard asked for a battle card on her top three competitors, with a tab each and a comparison grid. A presales colleague reviewed it and added security monitoring. She then asked for a skill with the same level of detail and logos.

What should a leader’s command centre track?

Start with one of three lenses. Flow asks where work waits in your team and where you need to hire. Direction uses AI as a thought partner on benchmarks and where the organisation is heading. Network asks which accounts and conversations to move forward this week.

How do you stop an agent guessing at numbers?

Tell it how to read each data source. For a narrow usage dataset, Gard set her agent to run a SQL query every time, instead of searching semantically for values. She did this to cut the risk of hallucination on data insights.

I have very narrow data. I just want to run a SQL query anytime I’m asking for some type of data insight.

Gianna Gard, AI Deployment Strategist, Dust

Key takeaways

• Start with one weekly question. Pick flow, direction or network.
• Turn your best answer into a skill. Then any team can reuse it.
• Put it on a trigger. Time it to land before the meeting that needs it.
• Track the cost from week one. Check tokens per question and switch off idle triggers.

About the workshop

Gianna Gard is an AI Deployment Strategist at Dust, a platform for building AI agents on company data, where she works on customer deployments. She ran this workshop on the AI Infrastructure track at aiblLIVE London on 20 October 2026.

What’s in the full guide

If you keep asking AI the same question every week, this guide shows you how to stop.

• The four building blocks of an agent, and how to set each one.
• An eight-step method from one chat to a scheduled, shared agent.
• The prompts from the live build, plus a starter command centre brief.
• What the room asked about shared context, where to start and cost.

For CEOs, COOs, sales leaders and the people who build their AI tools. Seven pages, about a nine-minute read.