Enterprise AI expands into higher business layers
Enterprise AI is advancing through three distinct layers at a pace that has surprised many observers.
Where humans still draw the line
Editors Zack Miller and Sara Khairi discussed which tasks they would assign to AI and which they would keep for themselves in their first podcast episode.
Miller reserves creative work for humans. He uses AI for planning, logistics, and as a sounding board—posing questions and refining ideas. The final call remains his responsibility.
Khairi refuses to let AI make decisions. She allows it to compare options for a costly laptop but insists on making the purchase herself. That small difference—AI assisting rather than deciding—is growing more significant in financial services.
For both editors, keeping humans involved is not just a preference but a core principle. AI can prepare the groundwork, present choices, and explain reasoning. When money is involved, however, someone must still take responsibility.
Credit Karma’s math-first approach
Intuit’s newest Credit Karma tools put this principle into action. The company introduced assistants for paychecks, refunds, and debt, though the real change lies in their design.
The tools share financial context, giving them a fuller picture of a user’s situation rather than working in isolation. Their process separates analysis from explanation.
Related: Markets wrap up a volatile week
Credit Karma’s engine first evaluates income, loans, interest rates, and goals. It runs simulations and settles on a strategy. Only then does AI translate those results into clear, personalized advice. The final decision stays with the user.
This structure suggests where financial AI might be headed. Instead of forcing users into conversations with bots, AI could appear within existing workflows—offering recommendations, requesting approval, or prompting action when needed.
A problem remains. Chatbots handle simple issues well, but complex problems often lead to frustrating back-and-forth exchanges. Banks celebrate high interaction volumes, but those numbers don’t guarantee success. The real measure is whether customers solve problems faster, make better choices, or engage more deeply.
Headline figures don’t always reflect meaningful progress.
Agentic AI is moving faster than expected
Agentic AI was mostly theoretical just months ago. Now financial institutions are rolling out agents rapidly.
Miller initially thought the shift would take longer.
Banks operate in silos; customers don’t. A decision-making layer acts as control, seeing across systems, understanding the user, and determining the next step. That coordination layer is now essential. It provides shared context, assigns tasks, and keeps everything aligned. This approach shapes enterprise architecture and forces the industry to define what machines should handle—and what humans must never surrender.
