Daniela Amodei Highlights Trust as Catalyst for AI Velocity

Enterprise leaders racing to deploy large language models (LLMs) at scale are increasingly standardising on Anthropic’s Claude – drawn by a safety-first philosophy that, counter to old assumptions, is proving to be a catalyst for speed. 

With Anthropic moving toward a public listing and fresh fundraising underscoring its momentum, the company’s influence on enterprise AI strategy continues to expand.

Speaking onstage during the opening keynote at Snowflake Summit (1-4 June), Anthropic Co-Founder and President Daniela Amodei argued that reliability is the foundation for true velocity. Safety, she said, isn’t a brake but the engine.

“Part of why we’ve chosen to primarily build for businesses and partner with Snowflake is the concept that trust is an accelerant. Trust is something that helps you go faster,” she said. 

“I’ve never had a customer meeting where the CEO said to me ‘I would love if Claude could hallucinate more.’ They’ve also never said it would be great if Claude was less predictable and better at producing bad outputs.”

A conversation onstage at Snowflake Summit

In a Q&A with Snowflake CEO Sridhar Ramaswamy, Daniela reflected on how dramatically the enterprise AI landscape has shifted in just 12 months.

“In AI, time is this crazy construct – a year ago feels like 10 years ago, right? Five years ago, nobody was using generative AI or LLMs in their daily workflows in their businesses and now every major enterprise says it is a foundational part of their workforce strategy,” Daniela said. 

“Every industry around the world is using AI… The part that is most mind boggling even to us in Anthropic is what are things going to look like a year from now?”

Daniela didn’t address questions about a potential IPO, but she did offer clear guidance for buyers navigating compounding capability gains.

Planning for the reality of scaling laws

A core challenge for CIOs and CTOs is future-proofing roadmaps as models improve every few quarters. Daniela pointed to scaling laws – the predictable relationship between compute, data and model performance – as the planning anchor.

“This idea that there’s this kind of predictable way that if you give the models more compute, more data, they get smarter, they get better,” she explained. 

Even for teams building the models, she noted, it can be hard to internalise the pace: gains arrive every year, six months, even three months.

Her advice is to set a bold target and build toward it, encouraging leaders to dream big and consider the absolute best version of their product or company. 

Because these models are developing at such a rapid clip, thinking about AI requires looking past today’s limitations to the largest version of what needs to be built, and then actively working toward that future.

What this means for enterprise leaders 

Navigating this fast-moving landscape requires shifting from temporary AI experiments to permanent infrastructure. 

First and foremost, enterprise leaders must treat trust as infrastructure by baking reliability, predictability, and governance directly into their stack early on. 

This solid foundation ultimately accelerates deployment, expands potential use cases, and builds vital customer and executive confidence.

At the same time, technical architectures must be designed for fast-forward adaptation. 

By using scaling laws to inform their roadmaps, leaders should assume material model upgrades will arrive on a three-to-six-month horizon, making it essential to design modular systems that can absorb these advancements without requiring major rewrites.

When executing this strategy, the smartest path begins with high-value, low-regret wins. 

Prioritising use cases where safety guardrails and deterministic workflows deliver immediate ROI allows an organisation to find its footing before graduating to more open-ended generation as controls mature.

To keep pace with this continuous model evolution, evaluation, red-teaming and monitoring loops must be established as permanent operational functions rather than one-off projects. 

Ultimately, today’s pilots should always align with a longer-term vision so that foundational investments in data quality, security, and application patterns compound over time, steadily building toward that “biggest version” of the enterprise goal, Daniela said.

As data platforms like Snowflake deepen ties with AI pioneers like Anthropic, the market is signaling a new mandate: don’t just build quickly – build on trust. 

With a reliable foundation, enterprises can pursue the largest version of their AI vision, confident that as scaling laws push capabilities forward, their architecture and governance will keep pace.