Plex Consulting Briefing

Leading in the
Age of AI

Jeff Sutton

March 2026

Who is using AI
every day?

Hands up. Be honest.

The Shift

The Old Way
🔍 Search
10 blue links
You find the answer
The New Way
💬 Chat
One conversation
AI gives you the answer

The internet just changed its interface.

The Arms Race

OpenAI Annual Revenue

2022

$28M

2025

$5B+

178x growth in 3 years

The Reality

42%

of companies approve AI use
but have no platform deployed

23%

just "look the other way"

Source: Section AI Survey 2024

Who Wins?

Incumbents

Slow to move
Rich in data
Heavy on capital
Loaded with tech debt

AI-Native
Disruptors

Fast & scrappy
Creative
Zero legacy debt
Small teams, big output

Fast-Moving
Incumbents ✓

Data advantage
AI-augmented teams
Speed of a startup
Best of both worlds

This is where you want to be.

AI Isn't New

You've been using it for 25 years

The AI You Already Know

2001
Google Search Ranking
2004
Gmail Spam Filter
2011
Siri
2016
Google Neural Translate
2022
ChatGPT
2024
Agents
2025
AI Army

What Changed?

The interface.
You can now TALK to it.

How It Works

Massively simplified

🧑
You type
🧠
LLM
Large Language Model
Response

Trained on trillions of words. Predicts the most useful next token.

Your New Teammate

  • Knows all the internet's information
  • Learns incredibly quickly
  • Analytical and creative
  • Never gets tired
  • Moves across domains seamlessly
  • Available 24/7 for $20/month

The Evidence

Harvard, BCG, Wharton studies (2023-2024)

+12%

More tasks completed

+25%

Faster completion

+40%

Higher quality output

It's an Intern,
Not Software

A Different Mental Model

SOFTWARE

  • Deterministic
  • Same input = same output
  • Install and forget
  • 0% error rate expected

AI

  • Probabilistic
  • Brilliant but inconsistent
  • Needs management
  • 5-10% hallucination rate
"You don't install an intern.
You onboard them."

The Onboarding Framework

SET

Expectations
Guardrails
Context

TRAIN

Good prompts
Examples
Feedback loops

SUPPORT

Review output
Iterate
Improve

Verification Tiers

Match review rigour to the stakes

🔴 Human-Verified
Official comms · Financials · Client work
🟡 AI-Verified
Code · Copywriting · Internal reports
🟢 Limited Verification
Ideation · Brainstorming · Internal drafts

"We bought Copilot licenses"
is not an AI strategy.

It's a talent development program.

The Last 90 Days

What just changed

Agentic AI

AI that doesn't just answer — it DOES things

  • Browses the web autonomously
  • Writes and runs code
  • Sends emails and messages
  • Deploys websites
  • Manages complex workflows

Reasoning Models

AI that thinks before it answers

Anthropic
Claude
OpenAI o3
DS
DeepSeek R1

Competition-level maths. PhD-level science.

Cost Collapse

DeepSeek proved frontier AI can run at 1/10th the cost

AI Coding Tools

Developers are 2-5x faster

  • Cursor — AI-native IDE
  • Claude Code — terminal agent
  • GitHub Copilot — inline autocomplete
  • Code is no longer the bottleneck

The gap between adopters
and laggards is widening
WEEKLY

What Now

Three Scenarios

Scenario 1

Niche Tool

Some industries find use cases. Most don't change.

Scenario 2 — Most Likely

Embedded Copilot

Everyone uses it daily. Like email in the '90s.

Scenario 3

AGI

Autonomous problem solver. Full agent economy.

Today vs Tomorrow

TODAY

Generate
Evaluate
Critique

SOON

Invent
Persuade
Transact

The Toolkit

Start at Tier 1. Level up as you learn.

Tier 1
Start Here — Claude Pro / ChatGPT Pro ($20/mo)
Tier 2
Level Up — Perplexity, Projects, custom instructions
Tier 3
Build — Cursor, n8n, Figma AI, automations
Tier 4
Enterprise — Ollama, RAG, AI agents, custom models

Cloud vs Local

☁️ Cloud

Best quality models
Always up to date
Data leaves your network
Per-usage cost

🏠 Local

Free forever
Data stays home
Runs on your hardware
Smaller models

🔀 Hybrid ✓

Cloud for quality
Local for sensitive data
Smart routing
This is the play

Building Your AI Army

A progression, not a switch

1
Solo User
Chat with AI for answers
0 agents
2
Power User
Custom instructions, projects, workflows
0 agents
3
Builder
Create automations, connect tools
5 agents
4
Operator
Manage AI teams, monitor quality
20 agents
5
Commander
Full AI platform, enterprise scale
250+ agents

The Mental Model

Humans bookend the process

🧑
YOU
Instigator
🤖
AI
Processor
🧑
YOU
Judge

Human judgment never leaves the loop.

5 Questions to Ask
Your CEO on Monday

  1. What's our AI policy?
  2. Who's responsible for AI governance?
  3. Where does our data go when staff use AI?
  4. What's our 90-day AI pilot plan?
  5. Are we the fast-moving incumbent — or the one being disrupted?

You just got a superpower.

But so did everyone else.

The question is:

Where's the
LEADERSHIP?

Now let me
show you...