AI has changed my work. And yours too

Boost your skills with Growth Memo’s weekly expert insights. Subscribe for free!

Am I still a consultant? Or a building contractor? I’m having an existential moment.

My work has changed forever in ways I’m still trying to understand. Six months ago, Agentic Vibe Coding crossed a threshold. Since then, I have been using AI to increase my impact many times over.

• I designed end-to-end landing pages for a major travel brand that made it into production.
• I’ve automated topic prioritization, SEO testing, and SEO reporting for my clients with full apps.
• I’ve built a number of useful applications for myself, from automating the SSI (SEO Site Index in the bi-monthly Growth Intelligence Briefs) to Openclaw agents to help me with research and charting.

The work I was sending improved, but at the same time it also became more difficult to define. But when construction costs collapse because of AI, judgment is the only thing that can’t compress. Meanwhile, most operators are still hiring, budgeting, and measuring as if execution were the imperative.

I’m not alone: ​​AI companies are reaching $100 million ARR faster than ever, thanks in large part to being AI native. Their entire product development philosophy is fundamentally different. Heck, Anthropic went from $9 billion to $30 billion in six months and is now worth about the same as Starbucks, Mastercard, or McDonald’s.

And I have my feelings about Matt Schumer’s essay “Something Big Is Happening,” but with a reported 80 million views, it clearly struck a chord.

So this week I want to take a break from publishing research and measure how agent coding is changing software, sales, etc People.

The impact on software

In 2024, I made a bold prediction that AI agents would reach 100 million users in 2025. I was off for about a year. Agents didn’t reach 100 million users in 2025, but they went into production in 2026, and the gains are measurable:

• METR saw a 1.5x to 13x (!) time savings when technical staff used Claude Code.
• Reducing costs by 40% and reducing time by 60% through agent AI is not unrealistic.
• Bain & Co estimates productivity increases of 30-50% through the use of AI agents and automation.

What happens to software when non-engineers can deliver code?

After the iShares Software ETF (IGV) plunged 24% in the first quarter of 2026 (the biggest quarterly decline since the fourth quarter of 2008), one could sense a panic in the air that AI would make software companies obsolete. But software is more than code.

Enterprise software has strong protections against AI redundancy. Anyone who has ever purchased a CRM or migrated to another provider knows how difficult it is and how much effort is involved.

Enterprise software is more than just code. It’s code plus integration, security, uptime, sales and support… all wrapped up in procurement cycles, IT review and legal approval.

The AI ​​can destroy any of these parts. For example, an agent can perform an integration, perform a security audit, and even book a demo. But no agent shows up to be sued when a mission-critical system goes down at 3 a.m. Accountability is the part that is not unbundled. Large companies are not replacing this stack; They build their own agents and AI workflows on top of it.

Things are different with self-service software. Anyone can now create a simple task tracker in Kanban format. Personally, I’d rather pay a few dollars a month and save myself the hassle of troubleshooting, but it’s possible and quick. Self-service products need to move upscale. The playbook appears in Notions, Figmas and Canvas Steps to Entrepreneurship.

Two archetypes stand out in this change:

1. Data provider.
2. System of records.

1. Data providers add value by providing data that the market would not otherwise have access to. These companies lose influence through their user interfaces but gain through their data. Suppose a data provider provides you with app store rankings. This company’s interface is slowly becoming a source of friction as more and more people are able to program their own dashboards. But their data becomes much more interesting. The enduring levers for APIs/MCPs in this world are data completeness, uniqueness, stability and cost. The logical step is to move to a headless experience for early adopters and keep the UI for legacy users.

2. Systems of Record (SOR) are the canonical location where a company’s own data is stored. Salesforce, Workday or Coupa are the bane of existence for many people, but they are billion-dollar companies because they are extremely difficult to replace. The moat is the tangle of permissions, audit trails, integrations, compliance status, and decades of workflow conventions built on this data. An agent can create a CRM in an afternoon; Replacing Salesforce at a Fortune 500 company is a multi-year change management project. These companies have already started and will continue to make greater use of AI to deliver better user experiences. However, their levers are the depth of integrations, compliance and audit status, switching costs and the quality of their agents. The winners in the SOR space are those whose agents make the existing system of record more useful, not those who seek to replace it.

The impact on distribution

Distribution is more important than product, they say, but getting it in 2026 is difficult. Platforms are closing down (by reducing the number of clicks and keeping users inside) and taking away opportunities to convert or build direct relationships with visitors outside the platform.

• AI overviews and AI mode eliminate more than 50% of clicks and keep users on the Google platform.
• AI chatbots send a tiny fraction of traffic outward.
• Social media is flooded, word of mouth is uncontrollable and pay is getting more expensive.

From the trademark tax:

In 2025 alone, the cost per visit increased by 9.4%, a cumulative increase of 30% over three years. Conversion rates fell by 5.1%.

How do you receive a distribution? in this AI-first world? Two-lever connection:

1. Speed.
2. Product.

1. Speed ​​means you work faster (and better) than your competitors. If all sales channels fail and no alternatives open up, The only way to grow is to use them better. Play the game better than the competition. Fast shipping speed becomes the deciding factor, and ideas + calculation become the differentiator.

PwC has found that AI accelerates content production by three to ten times. In plain language: We need to automate more. But not at the expense of trust. If you lose trust, you lose the game.

2. The product is now the marketing, with two different effects:

• AI sees through marketing gloss. Agents can read ingredient lists, analyze reviews and compare specifications. “We are the best X in the world” doesn’t survive an agent who actually checks. But strong products are consistently selected.
• The free product is the new top of the funnel. Standalone tools that solve a real problem are easier than ever to create and are a better buy than advertising. Ramp Sheets guides users without a marketing budget to Ramp’s core product.

When product is marketing, the focus shifts to product growth: onboarding, engagement, retention. The fastest growing products today all have product-led growth movements. This is how marketing and product development merge.

The effect on people

AI capability is advancing rapidly, but human cognition is not. Until we reach AGI (God knows when; I hope not soon), Human cognition limits AI productivity. We can only ship as much as we can verify.

AI tools can accommodate more input than ever before, while our own human attention spans are shrinking: AI’s context windows have grown 3,906x (!) over the last 10 years, from 512 to 2 million tokens, while human attention has shrunk. We outsource thinking faster than we learn to examine it.

Two cost curves compete: the cost of automation (exponential waste) versus the cost of verification (biological bottleneck). In “Some Simple Economics of AI,” Catalini et al. argue that tasks with verifiable output are automated the fastest. Work that requires a human to review slows down, allowing us to automate work that is easy to measure more quickly. I feel it when I use four terminal windows at the same time: the loss of focus is as high as the throughput. What holds us back on a large scale is how much we can proofread and direct.

When everyone can build anything, our limits change: skills and tools count less. But judgment, ideas and time determine whether you go in the right direction or in circles. It’s very easy to get distracted by AI because construction costs are so low now.

The judgment is the part that doesn’t compress. I can ask Claude Cowork to review the contract, but I need to know what it missed. Claude likes to write me a Q4 plan, but it’s only as good as what I’ve read about which market to attack and what my competitors are planning.

In the last six months, I’ve implemented more agent and automation systems than I’ve done hands-on work. My customers now have access to unique software that they can’t get anywhere else and that solves their individual problems.

Three things are now approaching zero: the cost of creating software, the cost of producing content and the cost of putting a tool into operation. But another cost is far from zero: the effort involved in figuring out whether any of this is correct.

I no longer “do” the work directly in the traditional sense. I’ll now build the thing that does the job and then check it. The work that matters now is the part I can’t leave to an agent…knowing what to build, what to kill, and what the agent missed. And I’m here to find out what that means – with you.

Additional resources:

---

Featured Image: Fit Ztudio/Shutterstock; Paulo Bobita/Search Engine Journal