Using AI Agents to enable a culture of experimentation

Project Managers are always looking to drive value for clients.  Delivering projects is often a delicate balance, where project teams have a “day-job” and need to maintain their BAU activities while also adding value through more strategic improvement projects.  AI is frequently cited as a potential route to efficiency, but it presents challenges for teams and individuals on where to start, let alone adopt new ways of working. However, the first step doesn’t need to be complex and can move AI from a generic tool with potential, to a powerful asset that project teams can leverage to their benefit.

Using content for context

For many, initial interactions with AI might look impressive but typically produce responses that lack relevance.  But real value can be gained by providing AI with context from existing documentation – and every project has a wealth of presentations, data and policies/processes already in governed company spaces.  This can be easily done through the setup of an AI Agent.  This simple shift, from relying on a vast publicly available information to leveraging a company’s own trusted data within an AI Agent, is the key to unlocking relevant and powerful results. It transforms a generic AI response into a powerful resource that uses the project’s own unique context.

Embracing Experimentation

There’s no doubt AI Agents are going to change the workplace, but there is no single playbook for rolling out such a wide-impacting technology.  When it comes to technical solutions, “tried and tested" is often cited as a risk mitigation approach however, AI Agent adoption requires teams to be involved, and in fact lead the way, on trying and testing.  This can be intimidating, so empowering teams with a safe space to "just try it" will encourage curiosity, engagement and ultimately create a cultural shift towards shared learning.

Letting the Team Take the Lead

Often, business changes are mandated from the top of the organisation – this experimentation approach encourages teams to:

• Collaborate: Work together, learn together and grow together which strengthens bonds and overall team performance
• Be curious: Foster a mindset where teams feel empowered to ask, "What if we used AI to...?"
• Remove the fear of failure: Knowing that not every experiment will end in a perfect result is critical - every attempt will provide valuable learning.
• Move from theory to action: Shift the focus from simply discussing AI to actively engaging with it and learning what impacts are made.

This behaviour aligns perfectly into what Project Managers aspire to develop within project teams.  

Practical Use Cases to Get Started

It doesn’t need to be ringfenced to projects only, to help any team find a place to start, here are some low-risk, high-reward use cases that can be implemented today:

• The New Starter Agent: Use existing team documentation to create an AI Agent that helps new hires quickly learn about the business and team. This enables more technical and focussed discussions with time-pressed team members once the foundations are laid.
• The Meeting Insights Agent: Connect meeting notes and transcripts on one topic (team meetings) to an AI Agent to automatically summarise key decisions, identify action items, and track responsibilities. This transforms multiple meeting outputs into a dynamic resource.
• The Proposal Agent: Create an AI Agent using historic proposals and strategic road maps that can then help draft future proposals, ensuring consistency in messaging and alignment with larger programmes of work.
• The Project Documentation Agent: For large, complex projects, an AI Agent can serve as a central knowledge base, answering questions about project plans, risk logs, and stakeholder information, reducing time spent searching for information.

First Steps for Your Team

Ideas to get started with your journey of experimenting with an AI Agent:

1. Identify your pain point - Find a single, specific task that is time-consuming or repetitive and has existing, well-documented information (e.g., a process document, a presentation, a policy).
2. Locate your source data and connect it to an AI Agent - Use tools like Copilot and Sharepoint to securely connect this documentation to an AI agent within your company’s governed space. This is often far simpler than people imagine and can be done natively in Sharepoint – likely already if you have Copilot!
3. Go straight in with conversational questions – you don’t need to have a fully formed “prompt”, simple questions are enough.
4. Take time to share your results.  Encourage team sharing in collaborative sessions – it great for team members to share their thoughts on what they found, ways it could be improved, what was impressive, what other use cases this is identified.
5. “Own” experimentation – it’s OK not to have all the answers on how the technology works, be curious and use published resources.  There are FAQs on Copilot to find answers to common questions, but share the links with the team so they can take responsibility for their own specific concerns.
6. Be the advocate for experimentation – continue to engage in all the learnings.  Celebrate finding what doesn’t work as much as what does!