The Agent Will See You Now: Your Next Business Partner Isn’t Human

What if your next business partner wasn’t human? Not a tool you command, but a proactive collaborator that anticipates your needs, suggests your next move, and works alongside you to get the job done. This isn’t science fiction. It’s the reality of agentic AI, and it’s already being built today. Forget the simple chatbots you’ve argued with online; we’re entering the era of the digital colleague.

At McGill and Partners, a specialist insurance and reinsurance broker, this future is already taking shape. They’ve built an AI agent that does more than just answer questions—it takes action. It’s a digital team member designed to support their client brokers, and its existence raises a fascinating set of questions. How do you build, manage, and, most importantly, trust a non-human colleague? This isn’t just a story about technology; it’s a story about a new kind of collaboration, and it offers a glimpse into the future of work itself. Check out the full podcast episode here.

Building the Digital Colleague

The mission was clear: create a digital agent for McGill’s external portal that could provide first-line support to client brokers, reducing complexity and increasing efficiency. The team, including Shwetha Balaji and Shreya Datta, turned to Salesforce’s Agent Force platform, a low-code solution that allowed them to construct a sophisticated AI without getting bogged down in years of development.

But what they built was far more than a glorified FAQ. The crucial difference lies in the shift from reactive to proactive intelligence. A traditional chatbot is reactive; you ask, it answers.

An AI agent, however, is proactive. It doesn't just wait for a command. As Shwetha explains, the key distinction with agents is "the fact that an agent can think, act, learn, and adapt to what's coming their way." This isn't just answering based on a set of pre-programmed responses; it's applying intelligence to anticipate needs and take initiative.

As Shreya describes their implementation, their agent doesn't just tell a broker how to raise a risk. It notices patterns and takes initiative. It might say, "By the way, you have risks that are due to expire in the next 60 days. Shall I start the renewal process for you?" Or, "I can see you've just placed a policy for a particular line of business. Would you be interested in our related products that could complement it?"

This is the core of agentic AI. It’s the ability to execute actions, to anticipate needs, and to function as a genuine partner in the workflow. The agent isn’t just a repository of information; it’s an active participant in the business process, constantly looking for ways to assist its human counterparts.

The Human Side of Agentic AI

Creating a digital colleague introduces a host of new, uniquely human challenges. It requires a fundamental shift in mindset, skills, and strategy.

First came the art of conversation. Shreya describes the process of setting up the agent as learning to “converse with the agent.” It wasn’t a matter of writing rigid code, but of guiding a probabilistic system with natural language. She found herself in a delicate dance of trial and error, experimenting with phrasing and even removing punctuation to see how it would alter the agent’s understanding. This is the new skill of the AI era: not just programming, but “AI whispering”—the ability to communicate effectively with a non-human intelligence.

Next came the guardrails imperative. Before a single feature was designed, the team focused on what the agent must not do. In the highly regulated world of insurance, this was non-negotiable. The agent could not make decisions on behalf of a user. It could not access data it wasn't authorised to see. As Shwetha explains, "We have taken small steps towards this so that every step we take towards implementing AI, there's a bit of additional trust on the technology." This incremental approach, building from non-controversial questions to more complex actions, created a foundation of confidence. The agent was built from the ground up to be a supportive partner, not an autonomous actor.

The Agent in Action

So, what does this digital colleague do day-to-day? Its work can be broken down into several key functions. It handles the high volume of repetitive queries that can bog down human brokers, such as questions about specific product features or internal processes. This alone frees up significant time for its human colleagues to focus on more complex, high-value work.

But its real power lies in its proactivity. By monitoring broker activity, it can identify risks that are approaching their expiry date and prompt the renewal process, preventing potential gaps in coverage. It acts as a vigilant assistant, keeping an eye on the details so the human broker can focus on the client relationship. Furthermore, by understanding the broker’s portfolio, it can intelligently suggest cross-selling opportunities, helping to drive business growth in a way that feels helpful rather than intrusive.

The impact is tangible. It leads to greater efficiency, a smoother and more responsive experience for clients, and it allows human brokers to operate at the top of their licence. It’s a powerful demonstration of the business value that can be unlocked when we move beyond simple automation and begin to build true agentic AI.

The Future is Agentic

The story of McGill and Partners is a microcosm of a much larger shift. We are moving from an era of using tools to an era of collaborating with agents. This has profound implications for the future of work. The most valuable skills will no longer be about technical execution, but about management, guidance, and collaboration with these new digital team members.

As the podcast discussion highlights, the key skill for the future may be metacognition—the ability to think about thinking, and to learn how to learn and adapt alongside AI. The future belongs to those who can embrace this uncertainty and forge effective partnerships with their new, non-human colleagues.

Hear the Full Story

We’ve only touched on the highlights of this deep and insightful conversation. To truly understand the challenges and triumphs of building a real-world AI agent, you need to hear the full story. In the complete podcast episode, Shwetha and Shreya go into much greater detail on the specifics of their “guardrails-first” strategy and the practical steps they took to build trust in their AI.

Listen to the full episode to get a masterclass in the realities of agentic AI deployment. You’ll learn more about the unique hurdles of testing a probabilistic system and hear the team’s hard-won advice for any organisation looking to move beyond the hype. This conversation is a practical guide to creating true digital partners, and it will change how you think about the future of your own industry.

Find the Project Flux podcast on your favourite platform and listen today.