September 05, 2025 | Listen Online | Read Online

The Human Side of AI

Upskill fast, redesign roles, and deploy AI agents with human judgment, empathy, and trust as your edge.

Executive Summary

What I have observed over the last couple of years is that companies are investing in technology ahead of their people. It's logical why they do this. Technology is widely available and often quick to implement, but upskilling and finding AI talent are hard.

That's why I think AI success starts with a people-first strategy: talent is your real moat, and early adopters are embedding agents to augment—not replace—human judgment. Leading firms are formalizing AI leadership, deploying human-agent teams, and building trust through clear cultural boundaries, practical integrations, and focused upskilling.

People‑first is the lowest‑risk, highest‑reward bet. The tech is early; your company's AI capabilities compound, so the best time to start upskilling was three years ago—or today. Here's an example of how Wells Fargo expanded its Google Cloud usage to equip employees with AI agents—an employee‑first program, not just a customer chatbot.
Design work as human + agent systems. Treat agents as teammates for repeatable tasks; keep decisions, judgment, taste, and trust with humans. HR examples are already shipping: Deel's AI Workforce (7+ specialized agents with human oversight) and HR Acuity's AI Hotline (triage + Workday Help integration).
Formalize leadership for responsible adoption. Lululemon just created the Chief AI & Technology Officer role. I actually don't like the idea of a Chief AI Officer; I think that they are a short-lived role, and traditional CIOs should take that as part of their purview to accelerate product innovation and govern risk.
Set the cultural boundary: work with AI, not for AI. New Workday research: 75% of employees are ready to collaborate with AI agents—but only ~30% are comfortable being managed by one. Keep human oversight explicit.
Reality check on disruption. A Stanford Digital Economy Lab working paper finds a 13% employment decline for 22 to 25-year-olds in the most AI‑exposed occupations since late 2022—another reason to upskill now.

Start where value is obvious, embed AI in existing tools, build guardrails early, and upskill by role—principles that remain the highest‑leverage moves.

FROM THE ARTIFICIALLY INTELLIGENT ENTERPRISE NETWORK

🎙️ AI Confidential Podcast - Agents Are the New API Client with Marco Palladino

☕️ AI Tangle - Google's $2.5T Dodge, OpenAI's Buyout of Statsig & Anthropic's Record Fundraiser

🔮 AI Lesson - Boost Writing Results with Meta Prompting

🎯 The AI Marketing Advantage - The Mindset Behind Vibe Marketing

💡 AI CIO - Compression of Software & Analysts

📚 AIOS - This is an evolving project. I started with a 14-day free Al email course to get smart on Al. But the next evolution will be a ChatGPT Super-user Course and a course on How to Build Al Agents.

ARE YOU AI READY?

Join senior executives & enterprise leaders for Charlotte's largest AI conference — two days of strategic frameworks, operational playbooks, and peer-to-peer exchange on scaling AI from pilots to productivity.

Seats are strictly limited and selling fast. Don't miss your chance!

We are currently calling for speakers — share your AI expertise and submit your topic here. The deadline is on September 15.

AI DEEP DIVE

The Human Side of AI

Upskill fast, redesign roles, and deploy AI agents with human judgment, empathy, and trust as your edge.

The Big Shift Happening Right Now. Forget the AI replacement narrative. The real story? Wells Fargo, Deel, and Lululemon are investing in augmentation, not attrition: AI agents as teammates, not replacements. And they're seeing immediate ROI.

Current Trends

75% of employees are ready to work WITH AI agents (Workday research)
13% employment decline for 22-25 year olds in AI-exposed roles since 2022 (Stanford)
Translation: Train your people now or lose them later to layoffs or to another employer who offers opportunities to learn.

Three Examples of AI Pilots That Are Actually Working

Last week's edition of The AIE highlighted how many pilots failed. This week, I wanted to give you a counterpoint.

1. The Banking Model (Wells Fargo)

What they did: Embedded AI agents directly into existing tools across branches, investment banking, and operations
The result: Employees use agents for policy lookup, memo drafting, and compliance checks—but humans own every decision
Your takeaway: Start with document-heavy, low-risk processes. Integrate into Slack/Teams, not a new platform.

2. The HR Automation Model (Deel + HR Acuity)

Deel's approach: 7+ specialized agents handle PTO, offboarding, payroll—with explicit human checkpoints
HR Acuity's innovation: AI Hotline integrated with Workday for instant triage and routing
Your takeaway: Pick repetitive, rules-based workflows first. Build escalation paths from day one

3. The Governance Model (Lululemon)

The move: Created Chief AI & Technology Officer role (not just "AI Officer").
Why it works: Ties AI to operating performance, not innovation theater
Your takeaway: AI needs operational ownership, not a separate silo

Your 90-Day Implementation Roadmap

I think that it's helpful to provide an implementation roadmap, but every company is different. I'd not only do this for my company but myself to give myself a leg up on my competitors and to improve my career in general.

Week 1-4: The Audit

Map every role's top 10 tasks
Label each: Automate / Augment / Avoid AI
Define non-negotiable human checkpoints
Document your "no-go zones" (HR decisions, compliance verdicts, trade secrets)

Week 5-8: The Training Sprint

Managers: Agent oversight, exception handling, risk triage
Analysts: AI-assisted research, verification protocols
Frontline: When to escalate, how to QA agent outputs
Create role-specific playbooks (2-page max)

Week 9-12: The Proof Points

Launch 2-3 internal pilots (pick document-heavy workflows)
Measure: Cycle time reduction, error rates, adoption curves
Publish wins weekly to build momentum
Adjust based on user feedback, not executive assumptions

The Fatal Mistakes to Avoid

❌ The "Adopt or Die" Approach: Recent CEO ultimatums backfired spectacularly. Build capability, showcase wins, let early adopters evangelize.

❌ AI as Judge and Jury: Never let agents make final calls on HR, compliance, or customer escalations. Assist mode only.

❌ The Shiny New Tool Trap: If it's not in Slack/Teams/existing workflow, adoption dies at 10%. Meet people where they work.

❌ The "Chief AI Officer" Theater: AI isn't a department—it's an operating model. Give it to ops leaders who own P&L. Here's one take on this that falls in the middle.

The Bottom Line Action Items

This week: Schedule a task audit with your top 3 departments
This month: Define your human/AI boundary rules (make them explicit)
This quarter: Ship one internal pilot that saves 10+ hours/week

Reality Check

That Stanford study showing 13% employment decline for young workers in AI-exposed roles? It's not a future risk—it's happening now. The companies building human + AI systems will poach talent from those still debating.

The choice isn't whether to adopt AI agents. It's whether you'll design the system or let it design you.

AI TOOLBOX

I've been revisiting my toolbox lately as I am trying to improve my creation of presentations. So far, they aren't killer apps, but they are some improvements over a few months ago. Where I am seeing the most improvement is in the agent-based tools that can create better presentations.

Manus.im - I am using Manus for all sorts of tasks, and while the features look like ChatGPT agent mode, Manus is much more reliable.
Genspark AI - Just like Manus.im, Genspark AI is an agentic AI tool that can do lots of tasks, but one of those tasks is that it helps users draft and design presentation content more efficiently.
Gamma - Generates full, story‑driven presentations from a single prompt; exports to Google Slides; built for collaboration.
Beautiful.ai - Creates sleek, polished decks with minimal input but offers less creative flexibility.

PRODUCTIVITY PROMPT

Prompt of the Week: Presentation Prompt

If you haven't noticed, the prompts I share have gotten pretty long; that's on purpose. I am really sharing the prompts I've used and gotten good results with. I have tweaked them and added lots of context, but that also means they are pretty long, which means they are triggering some spam filtering when I send them via email.

So, going forward, I'll give you the description but then ask you to click through to the website to see the prompt.

For example, I am doing a presentation to a group on a one-off presentation. I've been tweaking it using Manus.IM to edit it. I had an outline for the content, but it was for a special use case.

So here's what I think most people do, including me until recently.

Help me create a presentation on how AI works for a group of mid-level managers.

If you want to really create a good presentation, then you need to give it as much context as possible. I think a lot of AI users, including myself, often don't take the time to create repeatable tasks. That means spending 15 minutes to direct an AI agent to complete the task entirely. It's a huge time saver over going back and forth with a chatbot and if you do a good job with deconstructing your task you should save a lot of time over the course of doing repeatable tasks.

Here's a presentation I created in Manus.IM using the prompt below.

This is an incredibly long and detailed prompt. I did rewrite it to make it more generic and general use, but it's how I created the presentation above. It took me about 10 minutes of my time, then about 10 minutes for the agent to run. It's not perfect, but it's very close. Over time, I think I need to really figure out how to define the templates better to make sure the look and feel is polished, but the content was very good.

I used Manus, but you can try this in Agent Mode in ChatGPT to get the same effect, though in my tests Manus completed the task more easily.

Get the prompt

Remember: Great presentations are not about slides—they're about connecting with your audience and delivering value. The slides are just visual support for your message.