The name's IC Agents, 00-IC Agents

A CEO experiments with AI by outsourcing their keynote presentation and talking points to a collection of specialist AI agents (individual AI computer programs, like GPTs (Generative Pre-trained Transformers such as ChatGPT), but can act independently in a set environment.

These multi-agents were each trained on specific knowledge and expertise on different things to focus on specific tasks.

Agents understand a task and the context, make decisions, interact with other agents to query and fine-tune, and can take actions to achieve their specific goals, much like a digital assistant or robot helper).

What happened?

Shopify's CEO, Tobi Lutke, recently gave a standard leader presentation during the first half of their 2024 Summit.

In the second half of the Summit, he explained how his entire presentation was created by?multiple AI agents with specialised skills, each with a specific role.

They worked individually and collaboratively in a virtual office, periodically convening?in the conference room to brainstorm and assign tasks.

There was even a virtual water cooler for spontaneous encounters (which would happen because the “employees” were programmed to feel thirsty and get water to simulate our human way of working and creativity).

That AI team autonomously conceived, drafted, researched, wrote, edited, and designed the slides and talking points for Tobi’s talk. The Summit crowd was enthralled—quite rightly so.

People as the conductors

You'll notice that Tobi was still the human at the centre. Not knowing the complete details about his process, I'm sure that the output creation wasn't perfect, which it shouldn't ever be, and would have been further iterated by Tobi and others offline to get to the final version.

We're also seeing this in the generative video space, with people who have access to OpenAI's Sora, that the final creation needs to involve more than just AI.

What this could mean for IC?

This is a cracking IRL (In Real Life) example of where a trained internal communications professional with expertise in AI augmentation and integration into workflows could support leaders in the future.

I've been fascinated by AI Agents and Agentic AI Framework design this year since Chinese company OpenBMB (Open Lab for Big Model Base) released their ChatDev coding, which allows you to simulate specialist multi-agent collaboration. Also, Nvidia's digital twin simulation and training worlds.

A fun visualisation of how the AI IC agentic framework could look like

This is my take on what Tobi could have designed through an internal communications lens and the specific knowledge experts needed:

There's a lot of potential here, not just to help leaders but also many other stakeholders, including early/newbie IC professionals at the start of their internal comms careers, where AI isn't about replacing but fast-tracking knowledge and experience - because without this info you can't thoroughly prompt, iterate and use AI to help augment and empower the person to deliver good IC work.

To get the most out of AI, you need to know what to ask and how to iterate further using the right knowledge and experience rather than relying on just a few simple basic prompts.

For example, if you were to create an IC-simulated office with different expert agents in IC best practice both in theory (based on IC qualification syllabuses) and with expert IRL knowledge, the input could be to create a three-month training action planner to be used to create a GPT that prompts and tests someone new to IC joining x business in y industry, using real recent in-house work examples.

This local GPT robot would then be available 24/7 to help fast-track a new person's knowledge and experience. It can be a personal IC training partner/assistant, and the insights, conversations and testing, would be used with the person's manager to discuss their outputs and learning.

This is a very quick, simple example.

The importance of understanding IC workflows

Creating such AI agentic frameworks is only possible and realised once you map out, diagram, and understand your comms workflows to examine where AI could be augmented and integrated into how you and your team work on the many different user cases you're all involved in.

This is the space I'd love to help internal comms professionals with and what I plan to develop over the next three months, using all of my AI and comms knowledge and experience together.

Based on how I've been using AI over the last year and a half, I'll soon share my take on IC user cases and what integrated AI workflows could look like. I'll also share my insights on the best AI tools and how to use them, compare their strengths and weaknesses, and much more.

Please follow me in the meantime as we explore this new space together.

If you'd like to find out more, please message me as I'd love to hear about what you're interested in related to AI and IC.