Enhance your abilities with Progress Memo’s weekly knowledgeable insights. Subscribe without spending a dime!
Am I nonetheless an advisor? Or a builder? I’m having an existential second.
My work has ceaselessly modified in a approach I’m nonetheless making an attempt to know. Six months in the past, agentic vibe coding crossed a threshold. Since then, I’ve used AI to boost my influence by a magnitude.
- I designed touchdown pages end-to-end for a serious journey model that made it into manufacturing.
- I automated matter prioritization, web optimization testing, and web optimization reporting for my purchasers with full-blown apps.
- I constructed an array of helpful functions for myself, from automating the SSI (web optimization Website Index discovered within the bimonthly Progress Intelligence Briefs) to Openclaw brokers that assist me with analysis and charts.
The work I shipped improved, whereas it additionally grew to become tougher to outline. However when the price of constructing collapses resulting from AI, judgment is the one factor that doesn’t compress. In the meantime, most operators are nonetheless hiring, budgeting, and measuring as if execution is the constraint.
I’m not alone: AI corporations are reaching $100 million ARR quicker than ever, largely as a result of they’re AI native. Their complete product growth philosophy is essentially totally different. Heck, Anthropic went from $9 to $30 billion USD in six months and is now value about as a lot as Starbucks, Mastercard, or McDonald’s.
And I’ve my emotions about Matt Schumer’s essay “One thing Huge Is Occurring,” however with reportedly 80 million views, it clearly hit a nerve.
So, I need to take a beat from publishing analysis this week and take measure of how agentic coding adjustments software program, distribution, and folks.
The Impact On Software program
In 2024, I made a daring prediction that AI brokers would hit 100 million customers in 2025. I used to be off by a few 12 months. Brokers didn’t hit 100 million customers in 2025, however they did hit manufacturing in 2026, and the good points are measurable:
- METR discovered 1.5 to 13x (!) time financial savings when technical workers used Claude Code.
- A 40% discount in price and 60% discount in time from agentic AI will not be unrealistic.
- Bain & Co estimates a 30-50% acquire in productiveness from deploying AI brokers and automation.
What occurs to software program when non-engineers can ship code?
After the iShares software program ETF (IGV) cratered 24% in Q1 2026 (steepest quarterly drop since This fall 2008), you might sense a panic within the air that AI would make software program corporations redundant. However software program is greater than code.
Enterprise software program has sturdy guards in opposition to AI redundancy. Anybody who has ever bought a CRM or migrated to a different vendor is aware of how arduous that is and the way a lot is concerned.
Enterprise software program is greater than code. It’s code plus integration, safety, uptime, gross sales, and assist … all wrapped up in procurement cycles, IT assessment, and authorized sign-off.
AI can chip away at any a kind of items. For instance, an agent can deal with an integration, run a safety audit, even ebook a demo. However no agent exhibits as much as get sued when a mission-critical system goes down at 3 A.M. Accountability is the half that doesn’t unbundle. Enterprise corporations don’t change this stack; they construct their very own brokers and AI workflows on prime of it.
Self-serve software program is a distinct beast. Anybody can now spin up a easy job tracker in a Kanban format. I personally would slightly pay just a few {dollars} a month and spare myself the effort of bug fixing, however it’s doable and fast. Self-serve merchandise want to maneuver upmarket. The playbook exhibits up in Notion’s, Figma’s, and Canva’s strikes into enterprise.
On this shift, two archetypes stand out:
- Knowledge suppliers.
- System of data.
1. Knowledge suppliers present worth by making information that the market couldn’t in any other case entry. These corporations lose leverage from their person interfaces however acquire it from their information. For instance, let’s say a knowledge supplier offers you app retailer rankings. The person interface for that firm is slowly turning into friction as extra folks can code their very own dashboards. However their information turns into way more fascinating. The sturdy levers for APIs/MCPs on this world are information completeness, uniqueness, stability, and price. The logical transfer is to shift to a headless expertise for early adopters and preserve the person interface for legacy customers.
2. Methods of report (SOR) are the canonical place the place an organization’s personal information lives. Salesforce, Workday, or Coupa are the bane of existence for many individuals, however they’re billion-dollar corporations as a result of they’re extraordinarily arduous to interchange. The moat is the tangle of permissions, audit trails, integrations, compliance posture, and a long time of workflow conventions constructed round that information. An agent can generate a CRM in a day; changing Salesforce at a Fortune 500 is a multi-year change-management challenge. These corporations have already began and can proceed to make use of AI extra to supply higher person experiences. However their levers are depth of integrations, compliance and audit posture, switching price, and the standard of their brokers. The winners within the SOR house are those whose brokers make the present system of report extra helpful, not those making an attempt to interchange it.
The Impact On Distribution
Distribution is extra necessary than product, or so the saying goes, however getting it in 2026 is difficult. Platforms are closing (by decreasing clickouts and preserving customers inside), and so they’re taking alternatives away to transform or construct direct relationships with guests exterior of the platform.
- AI Overviews and AI Mode make greater than 50% of clicks redundant and preserve customers on the Google platform.
- AI chatbots ship a tiny fraction of site visitors out.
- Social is flooded, phrase of mouth is uncontrollable, and paid will get dearer.
From The Model Tax:
Value per go to climbed 9.4% in 2025 alone, including to a 30% cumulative improve over 3 years. Conversion charges fell 5.1%.
How do you get distribution on this AI-first world? Two levers compound:
- Velocity.
- Product.
1. Velocity means you execute quicker (and higher) than your opponents. When all distribution channels decline, and no alternate options open up, the one option to develop is to leverage them higher. Play the sport higher than the competitors. Quick transport pace turns into desk stakes, and concepts + compute develop into the differentiators.
PwC discovered AI speeds content material manufacturing up by 3-10x. In plain phrases, we have to automate extra. However not at the price of belief. Whenever you lose belief, you lose the sport.
2. Product is the advertising now, with two distinct results:
- AI sees via advertising gloss. Brokers can learn ingredient lists, parse evaluations, examine specs. “We’re the perfect X on the earth” doesn’t survive an agent that truly checks. However sturdy merchandise get chosen persistently.
- Free product is the brand new prime of the funnel. Standalone instruments that clear up an actual drawback are simpler to construct than ever, and so they purchase higher than advertisements. Ramp Sheets routes customers towards Ramp’s core product with no advertising funds.
When product is the advertising, the emphasis shifts to product progress: onboarding, engagement, retention. The fastest-growing merchandise nowadays all have product-led progress motions. So, advertising and product growth soften collectively.
The Impact On Individuals
AI functionality is racing forward, however human cognition … isn’t. Till we attain AGI (God is aware of when; I hope not any time quickly), human cognition is what limits AI productiveness. We are able to solely ship as a lot as we will assessment.
AI instruments can absorb extra enter than ever earlier than, whereas our personal human consideration span is declining: AI’s context home windows grew 3,906x (!) during the last 10 years, from 512 to 2 million tokens, whereas human consideration has shrunk. We’re outsourcing considering quicker than we’re studying to examine it.
Two price curves are racing one another: the Value to Automate (exponential decay) versus the Value to Confirm (biologically bottlenecked). In “Some Easy Economics of AI,” Catalini et al., argues that duties with a verifiable output will likely be automated the quickest. Work that requires a human to examine it compounds slower, so we’ll automate work that’s simple to measure quicker. I really feel it at any time when I’m working 4 terminal home windows without delay: the main focus drain is as excessive because the throughput. At scale, what holds us again is how a lot we will proofread and direct.
When anybody can construct something, the methods we’re restricted change: Talent and instruments matter much less. However judgment, concepts, and time resolve whether or not you run in the suitable path or in circles. It’s very simple to get distracted with AI as a result of the associated fee to construct is now so low.
Judgment is the half that doesn’t compress. I can ask Claude Cowork for a contract assessment, however I’ve to know what it missed. Claude will fortunately write me a This fall plan, however it’s solely nearly as good as my learn on which market to assault and what my opponents are about to do.
During the last six months, I applied extra agentic and automatic methods than I’ve completed hands-on work. My purchasers now have entry to distinctive software program they’ll’t get wherever else that solves their distinctive issues.
Three issues at the moment are compressing towards zero: the associated fee to construct software program, the associated fee to provide content material, the associated fee to spin up a instrument. However one other price is trending far-off from zero: The fee to know whether or not any of it’s proper.
I’m indirectly “doing” the work anymore in a conventional sense. I’m now constructing the factor that does the work, then checking it. The work that issues now’s the half I can’t hand to an agent … figuring out what to construct, what to kill, and what the agent missed. And I’m right here to determine what which means – with you.
Extra Sources:
Featured Picture: Match Ztudio/Shutterstock; Paulo Bobita/Search Engine Journal
