Blog post: Useful AI starts where your business knows the problem best

A lot of businesses are asking how to "do AI".

The better question might be: where do we know the problem well enough to build something useful?

That was the starting point for Scentia, a house of education brands that teaches around 40,000 to 45,000 people each year across leadership and management courses, graduate programs, MBAs and executive coaching.

For Scentia, AI was not treated as an abstract transformation project. It began with a specific question: what does the future student expect from learning?

For Matt, Scentia's Head of product technologies and AI, the opportunity was mass personalisation.

"The best way to learn would be with a one-on-one tutor or an individual teacher focused solely on you," he says. "We can't do that to the fullest extent, but what we can do is mimic that in key spots."

Those key spots mattered.

A student working on an assessment late at night wants to know whether they're on track. A student with 30 minutes to study wants to know what to prioritise. A student revising for a degree subject may need help pulling together the right content from articles, videos, activities and past weeks of learning.

These are not futuristic problems. They are everyday moments where students need clearer, faster, more personalised support.

That's where Scentia chose to begin.

Start where the organisation has the most knowledge

One of the most useful lessons from the Scentia story is that the best AI use case is not always the most obvious one.

Scentia looked at different opportunities across the business, including sales and support. But it decided to start with the student experience because that is where the organisation had the deepest expertise.

"We thought the best place to start was going to be with our core business," Matt says. "We know education extremely well. We've been doing it since 1941."

That domain knowledge mattered because AI does not remove the need for expertise. It depends on it.

To build something useful, Scentia needed to define what good feedback looked like, what students needed in the learning journey and what academic standards had to be protected. It needed to test AI outputs against the organisation's own tacit knowledge.

This is one of the most important points for any business considering AI.

If you start in an area you only half understand, it is difficult to know whether the AI is doing the right thing. But if you start in the part of the business you know best, you can describe the problem clearly, shape the requirements properly and recognise whether the output is genuinely useful.

The problem with generic AI

One of the tempting assumptions around AI is that a general tool can solve almost anything.

In education, the limitations become obvious quickly.

A student can upload materials into a generic AI tool and ask for a summary or quiz. But that does not mean the response reflects the institution's academic standards, assessment requirements or teaching approach.

As Matt puts it, "A generic Google searched answer isn't going to cut the mustard."

Scentia wanted students to use AI in an environment shaped by the institution: with the right course materials, rubrics, exemplars and guardrails. That way, AI could support learning without weakening the integrity of the course.

The aim was not to stop students using public AI tools. Matt is realistic about that. Some students will keep using tools like ChatGPT, Claude or NotebookLM, especially if they already use them at work or in their personal lives.

The point was to give students another option: one designed around Scentia's learning environment and expectations.

It is a useful distinction.

The goal is not to ban the generic tool. The goal is to build a better, more trusted version for the job that matters.

What Scentia built with Decidr

Working with Decidr, Scentia developed three agentic apps focused on the learning experience.

The first is an Always Online Course Tutor. It answers basic student questions, from assessment due dates and enrolment timing to face to face class information and introductory academic concepts. It gives students a place to go when a small question is blocking progress.

The second is an Always Online Assessment Tutor. Students can upload a draft assessment and receive immediate feedback based on the rubric and selected exemplars. It does not write the answer for them. It helps them see where they are strong, where they need to go further and how their work is tracking before the due date.

The third is a Personalised Learning Generator. A student can ask to revisit a topic from a particular week and receive a mini learning package that brings together relevant content, summaries, reflection questions and application activities.

The same capability also supports learning designers. Instead of using AI to write the learning content itself, Scentia is using AI to automate the build of interactive modules. In the demo version, Matt says this reduced build time from roughly three days to about 30 minutes.

That distinction matters.

Scentia is not trying to replace the educator's expertise. It is using AI to remove friction around the learning experience, support students more immediately and give staff more time for higher value work.

Why Scentia needed a partner

Scentia knew education. It knew its students, its content and its assessment expectations.

But, as Matt says, "We're not a tech company."

The organisation needed a partner that could help turn education knowledge into working AI tools at enterprise scale. It also needed a partner that could think beyond the first agentic app.

This is where Decidr came in.

Scentia was not looking for someone to build one agentic app, deploy it and move on. It wanted a longer term partner that could understand the journey across AI use cases and build the first tools with that broader direction in mind.

"The more holistic approach gave us confidence that they're understanding the problem space more completely and that they're also on board with us longer term," Matt says.

Decidr also gave Scentia a platform agnostic path, built on DecidrOS.

The team had seen AI tools emerging from learning management systems, but it did not want to be locked into one platform. Scentia wanted the flexibility to pivot, connect across systems and keep building as its needs evolved.

That is a useful reminder for any larger business. The first AI use case matters. But the foundation matters too.

Build the wrong thing in the wrong way and you may end up with a disconnected tool that works for one moment, then creates more complexity later. Build with the bigger picture in mind and the first use case can become part of a wider agentic organisation.

AI is not just a model problem

Matt is clear about one of the biggest implementation lessons.

"The AI component is usually the smooth and easy part," he says.

The harder part is connecting systems, testing assumptions and making sure people understand what the tool is meant to do. This is the work of orchestration, and it's where most AI projects succeed or fail.

That is where many AI projects become real. Not in the demo, but in the connections between the AI and the business: the data, the platforms, the workflows, the people and the expectations.

For Scentia, implementation involved the authoring environment, the learning platform and the AI component. It also involved change management, helping staff understand that the tools were not generic chatbots. They were designed to do a specific job with specific data sources.

That specificity is what makes the AI useful.

The future: personalisation beyond the first tools

The first tools are focused on teaching and learning, but Scentia's ambition is broader.

Matt describes AI eventually touching five functional areas of the business: what Scentia sells, how it sells, how it teaches, how it supports students and how its teams work.

In learning, the bigger goal is to bring "skills at scale" to a more personal level.

That might mean helping students understand the skills they have, the skills they need and the learning pathway that will help them reach a particular career outcome. It might also mean helping businesses understand the skills their workforce will need next.

The vision is simple, but ambitious.

"Every single student gets a personal teacher and every single staff member gets a personal assistant," Matt says. "You can't do that with humans, but you can try to do that with AI."

The real lesson: useful AI is built with the business, not around it

Scentia's story is not really about putting AI into education.

It is about what it takes to build AI that survives contact with the real world.

Start with the part of the business you know best.

Find a problem people already feel.

Build around the workflows, standards and data that make the business what it is.

Choose a partner who listens, but can also bring technical judgement, delivery capability and a bigger view of where AI is going.

That is what Decidr helped Scentia do.

Not a generic AI tool.

Not an experiment for the sake of it.

Practical AI, built around a real learning experience, with a path to grow. See more agentic AI use cases or explore Decidr's product.