AI That Works On Demand: Episode 4

Most organisations have already found a few useful AI prompts. But what if you could take those prompts, add the right knowledge, and turn them into an AI assistant that helps your team complete tasks consistently, accurately and at scale?

In Episode 4 of our AI That Works webinar series, Fraser Dear, Head of AI and Data Innovation at BCN, demonstrated how to build your first AI agent using Microsoft 365 Copilot.

Without writing a single line of code, Fraser showed how anyone with a Copilot licence can create agents that support everyday tasks, from sales preparation and contract reviews to marketing content creation.

Watch the recording and you’ll learn:

• What AI agents are and how they differ from standard Copilot conversations
• The four building blocks of a successful agent
• How to create and share your first Microsoft 365 Copilot agent
• Real-world examples that can save hours of repetitive work each week

Key Takeaways

From prompts to assistants

Up until now, much of the AI conversation has focused on prompting.

You start with a blank chat window, provide context and instructions, and Copilot generates a response.

The challenge is that if you regularly perform the same task, you often find yourself repeating the same prompt over and over again.

AI agents solve that problem.

Instead of rewriting the same instructions every time, you create an agent with a specific role, clear instructions and approved knowledge sources. Once configured, the agent can perform that task repeatedly using the information you’ve provided.

The four ingredients of a successful AI agent

Every agent demonstrated during the session followed the same structure:

1) Purpose

Start with a clearly defined job.

What is the agent responsible for doing?

The more focused the purpose, the better the outcome.

2) Instructions

Provide detailed guidance on how the agent should behave.

This includes role definition, guardrails, response style and specific tasks it should complete.

3) Knowledge

Give the agent access to relevant information.

This could be documents, SharePoint locations, websites, pricing sheets, policies or guidance documents that provide the grounding for its responses.

4) Starter Prompts

Suggested prompts help users understand what the agent can do and encourage consistent use across the organisation.

Demo One: Sales Deal Preparation Agent

The first live demonstration focused on a sales preparation assistant.

Many sales professionals spend significant time before customer meetings searching for product information, pricing guides, discount policies and competitive positioning.

The agent centralised that knowledge and delivered tailored customer briefing packs in seconds.

The demo showed how the agent could:

• Prepare customer meeting briefings
• Recommend products based on requirements
• Apply pricing and discount guidelines
• Identify approval requirements
• Provide competitor positioning guidance

Because the agent only used approved source material, it could confidently provide information while avoiding guesswork or hallucinations.

Demo Two: Contract and NDA Review Assistant

The second example focused on a common challenge for legal, procurement and commercial teams.

Reviewing contracts often involves checking for the same issues repeatedly against internal policies and playbooks.

Using approved contract guidance and review criteria, Fraser demonstrated an agent that could:

• Review agreement clauses
• Identify potential risks
• Highlight deviations from company policy
• Surface high-priority issues for human review
• Draft summary communications for stakeholders

Importantly, the session highlighted that these agents support decision-making rather than replace it.

The agent provided a first-pass review, but legal responsibility remained with the appropriate people.

Demo Three: Brand Voice Assistant

One of the most relatable examples for marketing and communications teams was the brand voice assistant.

Maintaining consistency across emails, social media, campaign messaging and sales communications becomes increasingly difficult as organisations grow.

By grounding an agent in brand guidelines, approved messaging and campaign briefs, the team demonstrated how AI can help ensure communications stay on-brand.

The agent was able to:

• Create LinkedIn posts
• Draft customer communications
• Rewrite content into an approved tone of voice
• Check messaging against brand guidelines
• Prevent the use of unapproved claims or language

For marketing teams, this creates a practical way to scale content creation while maintaining consistency across the business.

Governance matters

A key theme throughout the webinar was responsible AI use.

While agents can significantly improve productivity, organisations still need to think about ownership, AI governance and accountability.

The session highlighted several best practices:

• Keep humans involved in important decisions
• Clearly define agent limitations
• Use trusted knowledge sources
• Maintain and update agent content regularly
• Respect existing user permissions and security controls

In short, agents should help people work faster, not remove human oversight from business-critical decisions.

Quick wins you can try this week

If you’re looking to get started with AI agents, Fraser’s advice was simple:

• Identify a repetitive task you perform regularly
• Gather the documents and information you rely on
• Create a focused purpose for the agent
• Use strong instructions and guardrails
• Test it thoroughly before sharing it with others

Many organisations find their first successful agent starts with a process that already exists, rather than trying to reinvent how work gets done.

What’s Next

Episode 5 – Building AI for Your Business: Creating Assistants for Different Business Functions

In the next webinar, we’ll take the next step beyond personal productivity agents and explore how organisations can create more advanced AI assistants using Microsoft Copilot Studio.

This includes connecting to wider business systems, integrating data sources and creating agents that support entire departments and business functions.

Register here.

Next Steps

If this session sparked ideas about how AI agents could help reduce repetitive work, improve consistency or unlock more value from Microsoft Copilot, we’d love to help.

Explore our wider AI services, from secure Microsoft Copilot adoption and user training through to automation, custom AI agents and data-readiness programmes.