AI-native services: sell the outcome

The thesis everyone now cites: the next giant won’t sell software, it will sell the outcome, because services dwarf software roughly six to one. The distinction that earns its keep is intelligence versus judgment, AI is racing up the first, the second still rests on taste, experience and accountability. That is exactly where senior operators stay valuable rather than getting commoditised. What it underplays for my world: in regulated industry, trust and liability slow the handover from intelligence to autonomy.

The sharpest case yet for where AI value really accrues: sell the completed work, not the tool, because the services budget dwarfs the software budget. The team is the product; AI is internal leverage. It reads like a description of my own practice, senior judgment and relationships up front, AI doing the heavy lifting underneath. His critical test is the right one: does your service get stronger as the models improve, or do they commoditise you? My addition from regulated industry: the real bottleneck is trust and accountability, exactly the part no model closes for you.

The new GTM playbook: prospecting and the funnel as an engineering discipline, where the moat is how you find and engage buyers, not which tools you own. Right for self-serve and mid-market. My caveat from the industrial side: enterprise deep-tech deals are still won on trust and navigating the buyer’s org, procurement, security, the eighth stakeholder, exactly the part that does not automate. GTM engineering is necessary, not sufficient.

The most useful filter in this debate: which service markets are actually ready for autonomous disruption, scored on Hands (physical work), Hearts (trust and empathy) and Handcuffs (regulation and liability). Handcuffs is exactly the friction I keep flagging in industrial and regulated sectors, the reason the playbook lands faster in some markets than others. Read it as a screen: the fewer the H’s, the faster autonomy wins; the more, the longer a human-plus-AI service stays the right answer. The 44x gap between Europe’s legal-services and software markets shows how much value sits in the service layer.

The necessary counterweight to the services-eat-software narrative: a16z argues AI expands the application layer rather than gutting it, and that classic moats, switching costs, process, distribution, still hold. The line worth keeping: better models make the application layer more capable, not thinner, because the hard part was never raw intelligence, it was knowing what to do with it. Read against Sequoia and Lazarov, the truth is not either/or: the winners wrap judgment, workflow and accountability around the intelligence, whether you call the wrapper software or a service.