What does an Agent first Sales person look like? Marc Benioff said something in the latest SaaStr Podcast that really sharpened my thinking. He said "When I now write a V2MOM* I always have a Salesforce executive and Agent with me....I like having an AI Partner...it is another tool in the toolchest.."

I've used Copilot, AonGPT, ChatGPT, Gemini (yet to fully immerse myself in Claude) for sometime to support my client interactions and sales processes for myself and the team. Getting to the crux of the problem for an enterprise sales person; deals are complex, especially at Aon. We don't struggle with a lack of capability or content but we do need to maximise the following:

The information across teams

High quality and consistent deal qualification

Utilisation of internal expertise

Our time to prepare deeply for every client interaction

Internal relationship mapping

If we get this right we always show up informed and can create tangible value for our clients.

This is how I use an agent most effectively. At Aon we use Microsoft Co-pilot so this is where I have built my agent called 'Aon Sales and Client Insights'. The goal is simple; 'Help me (and others) think more strategically before every client interaction'.

My agent

The UX is simple, I have created 7 bespoke prompts that I can choose to start my Agent from preparing a client meeting, analysing as sales opportunity, planning Aon Client Leadership support, leveraging our Human Capital and Risk Capital Structure to the next steps in a deal.

I found using AI to help you build these and the agent description and instruction really valuable. I spent time in ChatGPT explaining what I was trying to achieve with my agent to build the prompt title and system prompt. This is fully aligned with how we work at Aon. For example, ACL will not mean too much to someone in another business but at Aon, it is our superpower.

Building out the prompts

Behind this is a clear Agent Description and Instruction. This part is crucial to get right. The description explains what it is and when it should be used, so colleagues (and AI agents, if enabled) can use it effectively. The Instructions direct the behavior of the agent, including its tasks and how it completes them.

Agent Builder for System Prompt

The Instruction is 5,100 characters, I've shown a screenshot above but again, take time and use AI to help write this out. It is a critical part to the effectiveness of the AI. Especially how you want to think about the prompts i.e. for my area of work, a human capital lense is crucial.

Knowledge. What are the sources I want my agent to use to generate responses? Firstly I have allowed the agent to search all websites. In addition I have uploaded X17 specific documents that are proprietary Aon datasets and further specific sites (including Sharepoint sites) that I want it to focus in on. For example, we just released our Human Capital Trends Study, so this is one of those datasets.

So what is an example of the output? I was recently preparing for a meeting with a top 3 global asset management firm.

Using one of my prompts the Agent provided me with a comprehensive overview of the situation for this client based on their annual report and other clients notes I have. It then ran a comparison to their key objectives against the broader market by analysing 17 different Aon datasets (surveys, benchmarks and case studies that I had uploaded) and then provided me with key insights and most importantly, 4 qualification questions to help qualify the gaps in the opportunity.

What I love about this is whether you are in HR, Customer Success, Broking or Finance, this is completely scalable. I can deploy this to my sales team and we can iterate to ensure this is going to deliver best value today and in the future. It helps focus your meeting prep and reduce time looking for information.

As our CEO, Greg Case , said recently "We are a firm of 60,000 colleagues today....we do not aspire to be a 30,000 colleague firm in the future doing the same work we do today, we are committed to a firm of 60,000 colleagues using AI to deliver to deliver more value for our clients...".

For me, this is such a great demonstration of the value of AI to ultimately deliver better decisions to our clients.

I'd be curious to understand how others are using Agents in this space.

*V2MOM is a management and communication process used by the team at Salesforce. It is an acronym that stands for vision, values, mission, objectives, and measures.