Would You Rather Your AI Agent Said "I’m Not Sure" or Guessed At an Answer?
In a regulated business a confident wrong answer is far worse than no answer, which is why a good agent hands the hard case to a person.
Read on Substack →
Book a callWhat we are learning as we deploy agents into real companies, written by the people doing it. New posts land here automatically.
Thanks, you are on the list. We will be in touch.
No spam, unsubscribe any time.
In a regulated business a confident wrong answer is far worse than no answer, which is why a good agent hands the hard case to a person.
Read on Substack →Most teams measure AI by the time it saves, while the organizations getting real returns measure what the work now delivers.
Read on Substack →The big banks are not saying no to AI agents, only to agents that sign in as you.
Read on Substack →Most of the value from AI agents depends on how the organization is set up, not on how skilled each employee is.
Read on Substack →Nobody would hire one person to run credit, treasury and social media, so why expect it of one AI?
Read on Substack →Almost every company now runs AI agents, yet almost none of them manage those agents the way they manage people.
Read on Substack →The person who fears an agent the most is the person who knows how the work really gets done.
Read on Substack →When an AI tool can open files your employees cannot, it will hand them to anyone who asks.
Read on Substack →Nobody hacked anything because it was only a search, yet what it exposed was a permissions problem.
Read on Substack →Monitoring tells you the agent is running, while managing tells you the role is being done.
Read on Substack →Chatbots are the most widely used AI tool at 47 percent, while agents that actually carry out the work reach roughly 20 percent.
Read on Substack →You do not install someone into a role, you onboard them, so an AI agent holding a seat is no different.
Read on Substack →Escalating costs, unclear business value and inadequate risk controls, which is governance under another name.
Read on Substack →This shouldn’t be measured on tokens, which are used to gauge what it costs to run the technology, when the question that matters is what the role is worth to you.
Read on Substack →You did not hire the keystrokes, so stop pricing your AI agent as though you did.
Read on Substack →One of them needs you to explain the situation every single time you use it, while the other one already knows what its job is.
Read on Substack →The difference between the agent you built yourself and a governed one is the difference between a helper and a hire.
Read on Substack →The instinct is to point an agent at your hardest problem. That is exactly where pilots stall. Start with the work that quietly drains your best people
Read on Substack →Here is the difference between the two and why it actually matters and changes what you should expect from the technology
Read on Substack →Gartner says the future of work is augmentation and that cutting heads does not pay. Here is what that means for how you deploy AI agents
Read on Substack →Every post is published on Substack. This page updates on its own as new ones go out.