The head of AWS's partner ecosystem just turned her own engineering team into a live demo, and the results are a blueprint for every company trying to figure out what "AI-assisted work" actually looks like when it ships.
The Summary
- Ruba Borno runs AWS's Global Specialists and Partners org and converted her AWS Marketplace engineering team into an AI-first unit: 88% higher shipping throughput, 21% faster time to production, 76% of code now AI-assisted
- AWS Partner Central now uses autonomous agents that qualify sales opportunities, route deals, and match resources without human input, adopted by 1,000+ partners since launch
- Her mandate to her teams: "We must use AI every day in nearly everything we do"
- The unlock isn't the model, it's the data plumbing: label it, control access, start with one use case and scale from there
The Signal
Most companies talk about AI pilots. AWS put its own Marketplace engineering team through one and published the receipts. The results matter because they're specific, measurable, and tied to a real production environment, not a sandbox. 76% of production code is now AI-assisted. Shipping throughput jumped 88%. Time to production dropped 21%. Rollback rates held steady, meaning quality didn't tank when velocity increased.
The model Borno describes is division of labor, not replacement. Humans define requirements. AI generates code. Autonomous agents triage incidents and push fixes without waiting for a prompt. This is Web4 operations: people set direction, agents execute, the work compounds while you sleep.
"Humans define requirements. AI generates code. Autonomous agents triage incidents and push fixes without waiting for a prompt."
The partner-facing work is where this gets interesting for anyone selling B2B. AWS Partner Central deployed AI agents that qualify inbound sales opportunities, deliver tailored guidance, and route each deal to the right team. More than 1,000 partners adopted the system since launch. That's not a beta, that's distribution. For partners, it means deals move faster and resources get matched sooner. For AWS, it means the sales funnel runs itself.
The Marketplace reinvention is the consumer-facing version of the same shift. Agent mode lets customers ask questions conversationally and get routed to the right product. Express private offers deliver personalized pricing in minutes instead of weeks of back-and-forth with sales. Both moves strip friction out of buying, which is what happens when you let agents handle the parts of commerce that don't require human judgment.
Key moves Borno made to get adoption:
- Mandate daily AI use across the org
- Start with one use case, not a full data warehouse overhaul
- Build the plumbing to connect, label, and control data access
- Deploy agents in production, not just demos
Borno's framework for getting worker buy-in is anti-hype. The AI tools need access to the right contextual data. That means infrastructure work: connect the data, label it, organize it so it's findable, set access controls. You don't unify everything on day one. You work backwards from a specific use case, get that data right, and scale from there. Without the right data, the agent is just expensive autocomplete.
The Implication
If you're leading an engineering or sales org and wondering whether AI agents are real or vaporware, AWS just gave you a case study with numbers attached. The playbook is clear: pick one high-friction process, feed the agents the right data, let them run in production, measure the throughput gains. Start with incident triage or lead qualification, not your entire stack.
For partners and B2B sellers, the shift is already happening. Buyers expect instant personalized pricing and conversational guidance. If your sales process still requires three meetings to deliver a quote, you're competing against agents that do it in three minutes. The question isn't whether to deploy this, it's how fast you can build the data infrastructure to support it.