5 Levels of Agency

High-agency employee? Worth 5x more.
High-agency agent? Costs 100x more to build.
That’s why 87% of companies have stalled at 'chatbots with suggestions.'

Think about your team.
An entry-level analyst who spots problems? Costs maybe $60K.
A VP who spots problems, diagnoses them, picks the best solution, and executes? Closer to $300K.

You pay 5x more because you get 10x more value.
Linear investment, exponential return.

AI agents work backwards.
The value scales up.
But the cost? It doesn't climb—it explodes.

Take a look at what happens as agents climb the agency pyramid.

—Level 1—
The Lookout
It's like having a smoke detector. Cheap, reliable, saves you from disaster. But it won't put out the fire.
Employee:
"Boss, something's wrong." Your intern refreshing the dashboard, Slacking when numbers look funky.
Agent: @Datadog screaming at 3am. Security tools flagging weird login patterns.
Cost: $100s/month. Basic pattern matching.

—Level 2—
The Detective 
Now you need real AI, not just if/then rules. It's Sherlock Holmes—brilliant at finding clues, but won't tell you what to do about the murder.
Employee: "The revenue dropped because the tracking pixel broke in yesterday's deploy." Your mid-level dev who can actually read the logs.
Agent: @GitHub Copilot explaining why your code faceplanted. @AWS anomaly detection showing which service is bleeding money.
Cost: Low $1000s/month.

—Level 3—
The Consultant 
Think Tony Stark's JARVIS in Iron Man—helpful suggestions, but Tony's still making the calls.
Employee:
"Here's what broke, and here are three ways to fix it—fast, cheap, or good." Your senior IC who's seen this movie before.
Agent: @HubSpot's AI building three email campaigns based on segment behavior. @Gong suggesting different talk tracks for your struggling rep.
Cost: $10Ks to deploy. Multi-model systems, retrieval, scenario planning.

—Level 4—
The Consigliere 
Think Tom Hagen in The Godfather—he'll tell you exactly what he'd do, but Don Corleone still makes the call.
Employee: "Here's what happened, why, three options, and here's the one I'd pick if it were my call." Your director with taste, judgment, and scar tissue.
Agent: @Harvey AI reading 40-page contracts and saying "change clause 7, keep clause 12, this precedent is risky." @Jasper building a full campaign strategy with rationale.
Cost: six fugures to deploy. Guardrails, validation, domain expertise, human oversight.

—Level 5—
The Autopilot 
Fully autonomous. Like Skynet before it got moody.
Employee: "I found the problem, diagnosed it, picked the solution, executed it, and here's the report." Your borderline-unemployable rockstar who makes $500K because they're worth $5M.
Agent: @Citadel's trading algos moving billions. @Tesla FSD navigating rush hour. @Intercom's Fin resolving 50% of support tickets with zero human touch. Cost: millions to build and run.

An employee 5x more expensive delivers maybe 3-10x more output. 
An agent 100x more expensive? Can deliver 10,000x more throughput.

That's why everyone wants Level 5 agents.
And why 87% of deployments stay stuck at Level 2.
The gap between Levels 3 and 4 isn't technical. 
It's organizational. 

Can your company handle an AI making recommendations that contradict your VP's gut? Can you let an agent refund $5K without a human signoff? Most can't. So they build expensive agents and then wrap them in so much red tape they might as well be Level 2.

The real question:
Are you paying $200K humans to do $100/month agent work? (Many companies are.)
Or are you trying to build $2M agents when a $25K one would actually ship? (Also common.)

Where are your agents on the pyramid? 
And what's stopping them from climbing?

Previous
Previous

AI and the end of the no-asshole rule

Next
Next

Picking an AI brand ambassador