Teresa Carlson says AWS had to define the government cloud market itself, since agencies had no way to buy cloud, no contracting mechanism and no security regime.
AWS won a classified CIA contract in 2012 or 2013 that required building the first classified environment for the intelligence community, bid by a team of four against IBM and Microsoft.
Carlson argues government and regulated industries are adopting AI without a definition or agreed framework, so the General Catalyst Institute is pushing Washington for a federal AI framework.
AWS's government playbook offers a guide for AI startups
AWS won government by defining a market that had no buying path, no security rules and no references. Carlson sees AI in the same position, growing faster than cloud while government lacks a definition or framework for it. Her General Catalyst Institute works at the industry level to supply that policy groundwork for startups in regulated fields.
Takeaways
AWS had to invent the government cloud market. Carlson started with two people and no revenue. Government had no way to buy cloud, no contracting mechanism and no security regime. AWS ran demos, built a partner ecosystem from scratch, and grew the business to about 10bn in ten years.
The CIA contract proved cloud could be classified. Won in 2012 or 2013, it required building the first classified environment for the intelligence community. Carlson's team of four beat much larger teams from IBM and Microsoft. She calls it a shot heard around the world.
Government standards pulled regulated industries into cloud. NIST defined cloud computing and the General Services Administration created the FedRAMP security model. Healthcare and financial services later adopted that certification approach, so government rules became a template for other regulated sectors.
AI spending will follow cloud's optimization pattern. Carlson says AI is growing faster than cloud did. She expects customers to cut token spend as cloud customers once cut usage costs, and companies to run separate models for different use cases.
Government and regulated industries need an AI framework. Carlson says they are adopting AI without a definition of it or an agreed framework. The General Catalyst Institute believes the federal government needs one, and she cites Anthropic's work with defense as a case where clearer rules would help.
Startups earn legitimacy through customers who speak publicly. Customers using a product and calling it good gives startups credibility. Carlson says Microsoft once avoided publicized references, which she calls the Washington Post test, while Andy Jassy pushed her to get AWS customers to speak out.
The Institute lobbies for industries, not single companies. Inside General Catalyst, it serves about 1,000 portfolio companies in healthcare, defense, financial services and energy. It brings founders to Capitol Hill, the Pentagon, Brussels and London, and urges them to offer solutions rather than problems.
Operating lessons from AWS
Sales shifted to consultative account management, since cloud customers paid only for usage.
AWS launched services fast, from 23 in Carlson's first year to about 2,800 later.
Tap a name to go there · tap a line to see why they are linked
On this episode 2
Show 1
Amazon leaders replaced PowerPoint with written documents.
Bezos and Jassy taught the one-way and two-way door framework for decisions.
Carlson's hopes for the year
A federal AI framework.
Faster permitting for critical minerals, energy and data centers, so the country can compete with China.