“Agentic AI” - how it can be deployed in #intled workflows

“Agentic AI” is one of those terms that seems to have just landed out of nowhere - and only about a year or so after “vibe-coding”…

Most of us have been experimenting with Generative AI platforms such as Claude, Gemini and Chat GPT, of course, but Agentic AI takes this to the next level.

So, following on from completing the MIT Sloan School of Management and the MIT Schwarzman College of Computing Implementing Agentic AI programme recently, I wanted to share some insights that could be relevant for our sector. So for the first in a series of three blog posts, today I wanted to explore what agentic AI is and how it could be used.

An agentic AI model uses those same LLMs above but has architecture around it to allow it to do so much more. Crucially, it can call on external tools, use memory and can reason to plan, complete and refine goal-focused workflows. This is also not just automation - agentic AI can work out the best way to do something and then iterate to improve the process!

Of course, understanding the impacts of how this technology can fit with, enhance or even completely redefine our business workflow processes is critical. It has become clear though that the way to go about this is to prototype, develop, deploy, test and iterate rather than spending months attempting to evaluate the capabilities of a platform that can quickly become obsolete. 

In consultation with industry, the MIT programme identifies Three Adoption Realities (1):

“First: Speed of iteration matters more than perfection of plan. The technology landscape is moving so fast that a six-month evaluation cycle is itself a risk.

Second: The biggest barriers are organizational, not technical. Models are mostly capable enough; organizations are mostly not ready.

Third: Existing metrics will not capture the value agentic systems create. The KPIs you have today were likely built for a different unit of work”.

So for us here in #intled, the priority is to identify discrete processes that are, in a sense, quick wins. Workflows that are time-consuming and repetitive but that may not be easily automated could be ideal candidates for deploying agentic AI solutions.

In International Student Recruitment, for example, this type of workflow might include: qualifying and converting leads; onboarding education agents; compiling & collating audit materials; and; analysing recruitment data - and many more, I’m sure.

When selecting an ideal candidate for your AI agent/s to handle, you need to consider a number of key questions including: which systems will need to be called on; what milestones need to be established; defining the level of autonomy and human escalation points; and your overall approach to governance. A discrete process that is repetitive but requires some variation and input based on reasoning could be an ideal candidate.

Here at AskEd, we’re working on agentic solutions for the TOF - to engage, qualify, convert and nurture leads in the most compelling way. We’re also looking at other parts of the process in our development roadmap. If you’d like a demo or to discuss any of the above please reach out.

Next time, I will focus on architecture and autonomy (sounds like an OMD album from 1981!) - and, of course, the “human in the loop”.

(1) MIT Sloan Implementing Agentic AI: Building Your Organizational Playbook program 2026-06-08