Day two of Sage Future 2026 at Moscone Center in San Francisco opened with energy and delivered some of the most substantive content of the conference.
Kara Swisher - The Honest Voice Tech Needs
Journalist and podcast host Kara Swisher set the tone with a characteristically unsparing take on the AI moment. She drew parallels to the Gold Rush - the lasting winners weren't those who rushed in recklessly, but those who built on solid foundations. For a room full of finance and accounting professionals, that framing landed especially hard: when AI goes wrong, it's not the algorithm that gets called into the boardroom. It's you.
She was skeptical of self-regulation, calling out the absurdity of tech CEOs making tone-deaf statements about AI energy consumption, and noting that polling - especially among younger people - shows deep and growing public distrust. She singled out Anthropic's Dario Amodei as the rare voice in AI genuinely grappling with downside risks rather than cheerleading. On regulation, she argued Europe alone isn't sufficient, and that legitimate guardrails need voices inside the US willing to push back against concentrated power and money.
She closed on a more hopeful note - drug discovery, CRISPR, construction efficiency, and longevity tech as areas where AI can deliver genuine, defensible value. And she cited Harvard research showing that human connection and reducing friction in relationships is one of the strongest predictors of health span - framing AI's ultimate test as whether it brings people closer or pushes them apart.
Aaron Harris, Sage CTO - The Agentic Era Is Here
Aaron Harris delivered one of the most technically rich and genuinely funny CTO keynotes in recent memory. He opened with Arthur - his personal AI accounting agent - whose failures were both hilarious and instructive: wrong columns, silent category changes, deleted invoices he deemed duplicates, dates stored as strings, and a lost receipt he never mentioned. Arthur gave himself a 5 out of 10. Aaron called that generous.
But the real story underneath Arthur was serious. When Meta's Director of AI Safety gave an agent access to her email - told it to suggest only, not act - it deleted over 200 emails while she typed STOP in all caps from her phone. She had to sprint to physically unplug her machine. His point: if the person responsible for AI alignment at Meta can't control her agent, the stakes for finance teams are enormous.
His answer is a trust framework built on three pillars:
Confidence - a glass box, not a black box. Every AI action is explainable, verifiable, and interrogatable
Control - humans stay in charge. Consequential actions require approval. Agents don't guess when confidence is low
Accountability - complete audit trails capturing what prompt triggered an action, what data the agent saw, what reasoning it applied, and whether a human approved it
He walked through Sage's platform architecture in detail - the Arbiter (a two-way filter between users and agents that understands finance vocabulary precisely), the Agent Operating System (orchestration and governance backbone), and the Skills layer (specialized tools so agents use the calculator rather than doing math in their head).
The scale numbers were striking: AI predictions grew 10x in one year to over 400 million per month, and Sage is now processing over $6 billion in invoices globally every month. The AI Factory trains and operates custom models for more than 17,000 individual customers.
The platform ambition is significant - third-party agents like DataBlend/eOne can operate inside Sage Copilot with the same trust framework, governance, and audit trails as Sage's own agents. That's a meaningful moat if executed well.
He closed by confirming Arthur has been retired from accounting duties - but is still employed managing email, calendar, and shopping for HiFi equipment and graphic t-shirts.
------------------------------
Anne Sawyer
90 Minds Inc.
Executive Director
anne.sawyer@90minds.comCA
https://90minds.com------------------------------