AI agents on the org chart: what founders need to know now
The org chart question is no longer theoretical. Two OpenAI employees who were managing teams of agents at age 24 wrote up what AI could actually look like inside a company’s structure. Their framing: in the simplest version, nobody loses a seat. Everyone just becomes a manager. An accountant still owns accounting while agents do the spreadsheet work. The accounting team doesn’t shrink, because rising productivity elsewhere in the company creates more demand for accounting than before.
The context bottleneck, not the model bottleneck
A separate a16z piece on Databricks makes the case that smarter models are not actually the binding constraint for most companies. The bigger problem is context: agents don’t know your internal processes, your unwritten rules, or the judgment sitting inside your best employees. Databricks is building a shared system of record to give agents that context, along with permissions and feedback loops. The competitive advantage this creates isn’t about which model you run. It’s about how well your company captures what it learns so that knowledge compounds even when you switch models.
Services eating software, and what comes after
Companies spend roughly $6 on services for every $1 on software. That ratio is why AI startups are moving from selling tools to delivering the work itself. But selling the work might only be the first move. Kirkland & Ellis is putting $500M into its own AI stack, which signals that large professional services firms won’t stay software buyers for long. The follow-on play: give the service away and monetize the transactions or data flowing through it. Rockwell runs a free benefits concierge for employees at mid-market companies so it sits directly in front of their healthcare spending.
The 90% gross margin era is over
AI companies need a new way to talk about unit economics. Serving customers with AI costs meaningfully more than serving them with traditional SaaS. The recommended framing for investors: separate gross margin by customer cohort and by pricing generation. The goal is to show that customers on your newest pricing model have higher margins than those on older tiers, and that the gap is widening over time.
One real data point on the AI-native company
Spawn is a 10-person company that spends 30 times its payroll on AI tokens. The company reports shipping 77,000 substantial changes in the last 30 days, putting its output close to Anthropic’s pace. Its work runs through a fleet of permanent AI agents, each owning a defined slice of the business. The team reports that each new model improvement has increased both speed and coherence.
That last detail matters. The compounding effect from better models only kicks in if your agents have enough context and task ownership to use the improvement. Which brings it back to the Databricks point: the bottleneck is context, not compute.
I’m founder of BizStack at Cagri Sarigoz LLC and a passionate advocate for entrepreneurs.
With over 14 years in tech, marketing, and AI, including my role as Head of SEO at CitizenShipper and co-founder of TaleBot at Intale AI, I’m dedicated to sharing genuine, useful product insights and tips.
At BizStack, I aim to cut through the digital noise to provide clear, actionable advice.
And more than all else, I’m a father to a (always) little girl and a husband.
Contact me at [email protected] for assistance.
