The Post-AI Manifesto

In December 2024, AI wrote better code than most developers. In January 2025, it started writing better copy than most marketers. In February, it began handling customer support better than most agents.

This isn't gradual improvement. It's displacement.

I've spent eighteen months watching businesses I invested in scramble to understand what AI means for them. Some adapted. Many didn't. The difference wasn't intelligence or resources - it was understanding which parts of their business AI would eat first.

Most founders are still asking the wrong question. They ask: 'How do I use AI in my business?'

The right question is: 'What parts of my business exist only because AI didn't?'

Lie #1: 'AI can't do what we do'

This is the most dangerous lie. It's what Kodak said about digital photography. What taxis said about ride-sharing. What bookstores said about e-commerce.

The pattern is always the same: incumbents point to current limitations as proof of permanent protection. But AI doesn't need to be perfect to displace you - it just needs to be good enough at 1/10th the cost.

Lie #2: 'We'll just add AI to what we already do'

This is the 'strategy' of most companies: sprinkle AI on top and call it innovation. But bolting AI onto a business model built for pre-AI scarcity is like adding a GPS to a horse-drawn carriage.

The companies that win don't add AI - they rebuild around what AI makes possible while protecting what AI cannot replicate.

Lie #3: 'Our relationships will protect us'

Relationships do matter - Part 5 of this manifesto is dedicated to them. But 'relationships' based on convenience, habit, or lack of alternatives will evaporate the moment AI offers a better option.

Real relationship protection comes from irreplaceable human bonds - trust built through shared struggle, reputation earned over years, community that can't be manufactured. Not from 'we've worked together for three years.'

AI doesn't eat jobs uniformly. It eats specific types of value creation. Understanding this pattern is essential.

AI Eats: Information Arbitrage

Any business that profits primarily from knowing something others don't is facing disruption. Research services, market analysis, competitive intelligence - if the value is 'we know things you don't,' AI is coming for it.

AI Eats: Process Execution

Anything that can be documented, systematized, and repeated is now AI territory. This includes most professional services, administrative functions, and 'best practices' implementations.

AI Eats: Quality at Scale

The economics of 'good enough at scale' have collapsed. AI can produce B+ work infinitely. Any business model based on producing reliable-but-unremarkable output at scale is vulnerable.

AI Eats: Gatekeeping

Middlemen who add friction without adding irreplaceable value are first against the wall. Recruiters who just match keywords. Agencies that coordinate but don't create. Consultants who repackage frameworks.

This is where the opportunity lives. AI has fundamental limitations that create permanent defensibility - if you know where to look.

Cannot Eat: Physical Reality

AI can design a house but can't build it. Can diagnose plumbing but can't fix it. Can optimize logistics but can't drive the truck (yet). Anything requiring atoms moved in the real world has inherent AI resistance.

Cannot Eat: Trust Relationships

Deep trust is built through shared experience, demonstrated reliability over time, and human judgment in high-stakes situations. Your doctor, your lawyer, your financial advisor - these relationships survive because you're trusting a person with your wellbeing, not just seeking information.

Cannot Eat: Network Effects (Real Ones)

True network effects - where value increases with each participant and switching costs compound - remain defensive. Not 'we have a database of users' but 'our users have built their professional identities here.'

Cannot Eat: Regulatory Capture

Licensed professions, compliance requirements, and regulatory moats don't disappear because AI is capable. A doctor still needs to sign off. A lawyer still needs to appear in court. An accountant still needs to certify.

Cannot Eat: Taste and Curation

AI can generate infinite options. It cannot tell you which one to choose. Human taste - knowing what's actually good, what will resonate, what matters - becomes more valuable as generation costs collapse.

Cannot Eat: Consequences

AI cannot be held liable. Cannot go to jail. Cannot be sued. Cannot be publicly shamed. When decisions require accountability that only humans can bear, humans remain essential.

After analyzing hundreds of businesses for AI vulnerability, I've identified five 'layers' that provide genuine protection. The more layers your business has, the more defensible it is.

Layer 1: Identity Investment

When customers make your product part of who they are - not just what they use - AI can't displace you. Apple users don't just use phones; they're 'Apple people.' CrossFit isn't just a gym; it's a tribe.

Questions to ask: Do customers publicly identify with your brand? Would switching feel like abandoning part of themselves? Is being your customer a statement about who they are?

Layer 2: Relationship Depth

Not all relationships are equally protective. The relationship that protects is one where: the customer would feel personal betrayal if you let them down; you know things about them no AI could learn from data; your history together creates irreplaceable context.

Questions to ask: Would customers take your call at 10 PM? Do they ask your advice on things outside your service scope? Have you celebrated or grieved with them?

Layer 3: Irreversible Stakes

When mistakes can't be undone, humans stay in the loop. Surgery, legal defense, major financial decisions, hiring - anywhere the downside is permanent, humans remain essential.

Questions to ask: Are mistakes in your domain easily reversible? Would customers trust AI alone with this decision? Is 'oops, let's try again' an acceptable response to failure?

Layer 4: Selection Judgment

AI can generate everything. Humans must choose. The skill of knowing what's good - real editorial judgment, taste, curation - becomes more valuable as generation costs hit zero.

Questions to ask: Is your value in creating options or choosing between them? Do customers pay for your judgment about what's good? Could an algorithm have selected what you selected?

Layer 5: Accountability Requirement

Someone must sign. Someone must testify. Someone must be responsible when things go wrong. This legal and social reality creates permanent human roles.

Questions to ask: Who goes to jail if this goes wrong? Whose name is on the line? Could responsibility for this outcome be delegated to software?

Here's a practical framework to assess your own business. Score each dimension honestly.

Revenue Source Analysis

Look at your revenue line by line. For each source, ask:

• Is this revenue tied to a specific human or relationship?
• Could AI replicate the value we provide here?
• What percentage of this revenue is 'commodity work'?
• If AI were 10x better tomorrow, would customers still pay for this?

The Replacement Test

For each major service or product you offer, complete this sentence: 'Customers pay us because AI cannot...'

If your honest answer is 'AI cannot... yet,' you're in trouble. If your answer is 'AI cannot... because physical presence/deep trust/legal accountability/genuine taste is required,' you're safer.

The Compression Test

Imagine AI reduces the time required for your service by 90%. What happens?

• If customers pay less: You're selling time, not value
• If customers pay the same: You're selling outcomes
• If customers pay more: You're selling status/trust/irreplaceability

If your assessment reveals vulnerability, here's how to rebuild. This isn't about adding AI - it's about repositioning around what AI cannot touch.

Step 1: Separate Commodity from Core

Make a brutal list of everything you do. Divide into: what AI can do (commodity), what AI can assist (hybrid), what AI cannot do (core). Your future revenue should come only from core. Commodify or automate everything else.

Step 2: Price on Outcome, Not Effort

If you charge by the hour, you're competing with AI on efficiency - a losing battle. Restructure pricing around outcomes, not inputs. What result does the customer actually pay for?

Step 3: Build Irreversibility Into Your Offering

Create stakes. Make your work matter in ways that can't be undone. Sign your name. Accept liability. Be the one who goes to court if it fails. This isn't comfortable, but it's defensible.

Step 4: Invest in Real Relationships

Not 'customer success touches' or 'relationship marketing.' Real relationships: know their kids' names, remember their struggles, show up when they're hurting. This doesn't scale - that's the point.

Step 5: Become the Curator

Position yourself not as the creator but as the selector. The editor. The one who knows what's actually good. Your taste is your moat.

Here's a thought experiment that clarifies everything: Imagine AI improves at 3x per year for the next decade. What survives?

Year 1: AI handles routine tasks

If your business is mostly routine, you feel this immediately. Administrative functions, basic research, first-draft writing - all compressed.

Year 3: AI handles complex cognitive work

Analysis, strategy recommendations, creative production at scale. 'Knowledge work' becomes 'knowledge verification.'

Year 5: AI handles judgment calls

Most decisions that once required human judgment can be delegated. What remains: decisions where being wrong is catastrophic and irreversible.

Year 10: AI handles everything that doesn't require a body or a soul

Physical presence and genuine human connection become the only defensible advantages.

Every founder should sit with these questions. They hurt. They should.

On Value:

• If AI could do everything I do, what would I do instead?
• What do customers pay for that they don't actually need anymore?
• Am I solving a real problem or exploiting information asymmetry?

On Relationships:

• How many of my customer relationships are real vs. transactional?
• If a competitor offered 50% lower prices, how many customers would stay?
• Do customers want to work with me, or do they just want the outcome?

On Identity:

• Is my work part of my customers' identity, or just a tool?
• Would my customers defend my business in public? Why?
• What would customers lose beyond the service if I disappeared?

On Stakes:

• What's the worst thing that happens if I make a mistake?
• Who is personally liable for my work product?
• Would customers trust AI alone with this decision?

Some industries are more vulnerable than others. Here's my assessment of the current landscape:

High Vulnerability

Medium Vulnerability

Lower Vulnerability

Theory is worthless without action. Here's what to do this week:

Day 1: Revenue Audit

Print out your revenue sources. For each one, write whether the value comes from: information, process, relationships, judgment, or physical presence. Circle anything that's primarily information or process.

Day 2: Customer Conversations

Call three of your best customers. Ask: 'If AI could do everything we do at a fraction of the cost, would you still work with us? Why or why not?' Listen to what they actually value.

Day 3: Competitive Analysis

Google your service + 'AI.' See what exists. Try it. How close is it to what you offer? What's the gap? Is that gap shrinking?

Day 4: Team Discussion

Share this manifesto with your leadership team. Have an honest conversation about which parts of your business are vulnerable. Make it safe to be brutally honest.

Day 5: Action Plan

Based on the week's learning, identify one thing you're going to stop doing (commodity work), one thing you're going to do more of (core work), and one investment you're going to make in relationships or irreversibility.

This manifesto isn't pessimistic. It's realistic. And realism creates opportunity.

The founders who understand these shifts first will build the companies of the next era. Not by fighting AI, but by building around what makes us irreplaceable.

We are at an inflection point. The businesses built for the pre-AI world will struggle. The businesses built for the post-AI world will thrive.

The choice is simple: adapt or decline.

AI is the most powerful tool humanity has ever created. Use it to eliminate the parts of your business that shouldn't exist. Double down on the parts that only you can do.

The future belongs to the builders who get this right.

The best time to plant a tree was twenty years ago. The second best time is now. The worst time is after the forest fire has already started.

The fire is here.

Start building.