AI Agents & Personalisation

1 x 2.5 Hour Bespoke Session (online)

The step past chat: turning the tasks you repeat into scheduled prompts and agents, personalising the tool to your own role, then building small working things without writing code.

Summary:

A bespoke session for staff who already use AI chat and want the repeated parts of their week to stop being manual. It moves in three steps: first the repeated task, from a scheduled prompt that runs on its own to a proper agent that holds its own instructions and knowledge; then personalisation, so the tool stops having to be briefed from scratch every morning; then a look beyond chat entirely, prototyping small working tools with no code involved. Every agent shown is one the audience would actually use, including one grounded in the regulation their own work runs on. It closes honestly, on what large language models still get wrong and on the rules for using public AI safely.

Course Breakdown:

Tackling repeated tasks: from general chat towards AI agents

• Scheduled prompts
• A scheduled web search that arrives without being asked
• Use case: a weekly generative AI news digest
• How agents work: reusable assistants for repeated work
• Use case: a business communication coach that improves text
• Use case: a regulation assistant, grounded in the actual regulation rather than in general knowledge
• The different ways an agent can support work, from text support through to knowledge, simulations, decisions and tools
• What an AI agent actually is
• What a connector adds to it
• Use case: a personalised calendar administrator, a tool-based agent with a live calendar connection

Personalisation

• The six ways to personalise: custom instructions
• Traits and style
• Personality and tone
• Occupation and background
• Naming and identity
• Memory
• Use case: standardised ticket replies
• Memory: when AI should remember something for later
• Custom instructions versus memory: stable rules against remembered context
• A prompt that writes both your custom instructions and your memory for you

AI beyond chat: build, test, adapt

• What building with AI means when there is no code involved
• Use case: a job detail page, built as a working prototype
• Use case: an interactive text adventure, to show what embedding AI directly into an everyday tool looks like

The honest close

• The limitations of large language models: why AI forgets, loses context, or needs a stronger set-up
• Using public AI safely: public, anonymised, generic or invented content only

Outcome:

Participants leave able to hand a repeated task to a scheduled prompt or an agent instead of doing it again, with a chatbot set up around their own role so it stops asking for the same context every time, and with a working sense of where an agent’s usefulness ends. They also see what can be built without a developer, which usually changes what people think to ask for next.