Dear Favorite Business Leader: A year ago, most of our conversations on AI in the workplace centered on chatbots, those supercharged search boxes that could answer almost any question intelligently … provided you asked it the right way.
Well, as you might expect, AI technology has moved on.
The conversation now is about agentic AI, and if you’re a business leader who hasn’t yet wrapped your head around what that means, read on.
Here’s the difference. An AI chatbot answers. An AI agent acts.
Give a chatbot a question, and it returns only words in response.
Give an AI agent a goal — something like, “find our three best suppliers for this component, compare pricing and draft a shortlist email” — and it plans the steps, pulls data from your systems, uses tools, checks its own work, revises if necessary and delivers a finished result. It doesn’t need hand-holding at every turn.
My own agentic shift
I’ve spent the past 18 months deep in this transition, both in the classroom and in the trenches as a consultant.
I earned a certificate from MIT Sloan in planning and designing AI agentic workflows. I’m able to identify and map business processes for an AI agent (or group of agents), establish governance, build pilots, identify relevant ROI and build human cohorts.
Some of my best education, though, has come from a separate “vibe coding” program, where I used AI and a coding tool to build basic apps myself (and I’m no traditional programmer). This gave me additional insight into when, where and how AI agents outline tasks, how they handle getting confused, how they rethink their approach, and how they go back and correct mistakes, all on their own.
I’ve built a few basic apps. I’ll probably never be as good as my talented programmer friends, but I encourage you to keep watching this trend. Vibe coding might quietly reshape how companies build, or buy, their next piece of software.
The secret sauce (context engineering)
That brings me to the primary differentiator behind good agentic AI, “context engineering.”
Prompt engineering is about the words you type. Context engineering is about the entire information environment you hand the AI agent — your CRM data, past decisions, tools, policies, institutional memory — structured so the agent can actually use it, not just skim past it.
An AI agent without good context is like a brilliant new hire with no onboarding. He or she is capable but has to guess how to handle the work. Without context and guiding documentation, an AI agent that impresses you once in a demo might struggle to live up to that first impression.
Identify a workflow candidate
Thinking about agentic AI for your business?
Here’s a starting point any leader can use: Don’t approach the decision thinking about where you can best add an AI agent. Instead, look for a workflow.
Ask your team for input. Try to identify a single, high-value workflow that could be redesigned, as a real test, so that humans and AI agents work it together.
PwC’s 2026 AI Business Predictions supports this point, too. Technology delivers only about 20% of an initiative’s value. The other 80% comes from redesigning the work itself, including the governance and, frankly, how we measure ROI in the first place.
Buying a tool isn’t the project. Redesigning the workflow is.
Tracy Pratt, Brandfully Yours columnist
Examples to inspire you
Here’s where this is already playing out, with real numbers behind it:
Customer service triage. Instead of a bot that reads a script, an agent pulls order history, past tickets and customer sentiment, then resolves a basic level of requests and routes the complicated remainder to a human, with the full context already assembled. No “please press 3 to hear these options again” (that one drives me crazy). DXC Technology, an IT services firm, applied this to its security operations center, cutting ticket acknowledgment and triage time by up to 80% while moving its analysts into more strategic, judgment-heavy work.Sales prep and follow-up. An AI agent briefs a rep before every call (account history, recent activity, likely objections), then afterward updates the CRM and flags at-risk deals on its own. Uber for Business’ own revenue operations leadershipshared an example on LinkedIn of AI agents augmenting, not replacing, its sales staff: 17 AI-powered sales use cases tracked, more than 10,000 hours of rep time saved, and measurable gains in win rate, deal size and email response rates.Content operations. This one is close to my heart, since I cut my teeth in a newsroom. It isn’t about replacing writers, it’s about clearing the clutter around them. The New York Times’ own AI Initiativesteam built an internal tool called Echo that summarizes and repurposes Times content from a single, journalist-written prompt. It’s paired with a companion framework, called Stet, that scores AI-generated copy against the paper’s own editorial standards, then routes a random sample to human editors for review before anything ships. It’s a genuine agentic workflow, at a media company with every incentive to protect its writers, and it still leans on humans for the final call.
Notice the pattern. In each case, the AI agent doesn’t replace the human. It removes the friction points around the person so their judgment shows up where it counts.
Agents are more than automations
PwC also flags something worth thinking about.
Many 2025 AI “agent” projects didn’t deliver value because, under the hood, they weren’t doing agentic work at all. They were automations wearing an agent costume.
There’s a real difference between a system that can act, review, correct itself, and learn to do better next time, and one that just runs the same script only faster.
I hope I’ve given you a lot to think about, but keep this in mind. In this new season of AI transformation, the businesses that come out ahead won’t be the ones that piled on the most AI. They’ll be the ones that picked one real process, redesigned it with real governance, and figured out how to measure success in a way that actually matters to their bottom line.
It’s the difference between cramming for a test and actually learning the material. Only one of those sticks around after graduation.
Brandfully Yours,
Tracy
Tracy Pratt, a Cedar Rapids marketing professional with expertise in communication, consumer behavior and AI, believes in blending data with storytelling to help businesses build stronger relationships. Message her on LinkedIn: linkedin.com/in/1tracypratt