We were at the pub last week with some people. Smart people. Accomplished people. But not technical people.
The conversation wandered over the course of a pint and - as it often does - landed on the way forward for AI in their industry. As I said, these were people who know what they are about and they have seen AI in businesses like theirs, but there was one question that came up more than once: “What is Agentic-AI and what the hell do I say the next time I'm asked?”
This, right here, is the moment we’re writing for. Not a pitch. Not a white paper. Just a clear explanation of what’s happening - and why it suddenly matters.
Let’s start with what you probably already know.
Most of the AI you’ve come across so far is what’s known as generative AI. You type a prompt - “Write me a job description” or “Summarise this meeting” - and the AI gives you something useful back. It’s clever. It’s fast and it saves time.
But there’s a catch: it always waits for instructions. It doesn’t take initiative. It doesn’t set goals. It’s smart, but passive. Like a calculator that knows the full train timetables in Mumbai.
Agentic AI is different. It doesn’t just respond to requests - it acts.
Give it an objective - say, onboarding a new hire or chasing up overdue invoices - and it figures out the steps, checks the data, nudges the right people, and gets the job done. Without being told what to do at every stage. It’s the difference between asking someone to bring you a sandwich and living with someone who’s already been to the shops, knows you hate coriander and has already cut off the crusts.
That shift - from passive tool to active helper - is why people are getting excited (and maybe a little nervous). Because it’s not just about saving a few minutes here and there. It’s about removing entire layers of friction from how businesses run. And when you scale that across departments, across workflows, across whole organisations - you can see why people like McKinsey and co. are calling this a generational shift.
But here’s the next catch. For agentic AI to actually do all this, it needs to know things. What systems you use. Where the data lives. Who does what. What it’s allowed to touch - and what it definitely isn’t. That’s where MCP comes in.
Your AI skeleton key…
MCP stands for Model Context Protocol. Stay with me. I know, it sounds technical.
Think of MCP like a universal plug socket for your business tools. Before MCP, if you wanted an AI to work across your calendar, HR system, payroll, Slack, Google Drive and CRM, you had to build custom integrations for each one. Painful. Expensive. Fragile. With MCP, those systems can expose just the right bits of information in a safe, standardised way - so the AI can get what it needs without breaking things or going rogue.
It’s a bit like how USB-C lets you charge your phone, connect your monitor, or upload photos - no matter the brand. MCP does that for AI and software tools. And it’s already catching on. There are now thousands of ready-made MCP “connectors” built by the community, so you don’t need to start from scratch.
Here’s what that means in real life. Imagine your company’s just hired someone new. In the old world, HR would send a checklist to IT. IT would set up accounts. Someone would forget to create a Slack login. Payroll would be missing a bank detail. And everyone would spend the first week chasing things down.
In the MCP-powered world, agentic AI notices the signed contract in Google Drive, checks the start date, pings IT via Slack to provision accounts, adds the person to payroll, sets up orientation meetings based on people’s calendars, updates the shared HR dashboard, and lets the hiring manager know it’s all sorted. No one had to remember. No one had to chase. It just got done and the hiring manager actually gets to finish their coffee that morning.
That’s the magic: not AI doing flashy new things, but AI quietly doing the things that usually slip through the cracks.
And it’s already happening.
So what does this mean for someone running a business today?
First, you don’t need to learn how to code. But you do need to be able to talk about this stuff - to your board, your team, maybe even the guy who sits next to you on a plane and has some investment funding you might like to get your hands on.
That means asking smart questions. Not “how do we install AI?” but “where are we wasting time on repetitive handoffs?” Not “can AI replace this role?” but “how can AI remove the grunt work so our people can focus on higher-value tasks?”
You’ll also want to check a few basics. Are your systems set up to talk to each other? Is your data in decent shape? Do you have clear policies about what decisions AI can make - and which ones still need a human? Because AI can’t help if it’s blocked from the information or stuck waiting for approvals no one thought to write down.
And maybe most importantly: is your team ready to work with this kind of tech? That might mean retraining people to spot exceptions rather than execute every step. It might mean updating job descriptions to reflect new kinds of collaboration. It definitely means having open conversations about what AI can do - and what it shouldn’t.
Yes, there are risks. Giving AI access to sensitive systems and a mandate to act without guardrails is a recipe for disaster. But that’s why MCP exists - to make those guardrails part of the system from day one. Every action can be logged. Every sensitive step can be set to require human approval. You stay in control.
This isn’t about handing the reins to a machine. It’s about letting go of the stuff you hate doing anyway - and trusting a system to handle it properly. With visibility. With accountability. And with a lot less faff.
We’re not saying you need to roll this out company-wide tomorrow. But we are saying it’s worth exploring now - before your competitors do.
So next time someone brings up agentic AI or MCP over dinner, you won’t have to change the subject. You’ll be able to say: “It’s basically AI that handles tasks for you - and there are now standards emerging for plugging it into business systems safely.”
And that, frankly, is the kind of answer that turns heads.
·- ··
At GiantKelp, we build AI tools which elevate your people and your business. Talk to us to find out how we can help you navigate the shift to agentic AI.
·- ··