Generating a new approach to business

For William Flanagan, director at digital transformation specialists OpenSky Data Systems, the most interesting technology organisations are today adding to their digital transformation checklist generative artificial intelligence (AI).

Indeed, OpenSky itself uses generative AI in its own business for a variety of tasks.

“We use it from sales to business analysis and support, software developers and QA [quality assurance], who each have use cases and have been able to drive things forward. That was our first use case when we started working with customers,” he said.

Most of us today are at least familiar with generative AI in the sense of having dipped a toe in the water by testing the likes of ChatGPT or Google Gemini. These free tools, however, are far from the cutting edge of AI.

“You don't have to work very long with the publicly available ChatGPT 3.5 to see the hallucinations appear. It's a generalist and you ask it questions and it will do the best it can. But the paid models, like GPT 4 and Claude [from Anthropic], are performing better,” Flanagan said.

It is not a theoretical debate, either. OpenSky has worked on real use cases for clients, and while many of us have been wowed by the parlour-trick aspects of AI, the very same technology can be put to work processing and summarising documents.

“There are organisations processing a large volume of forms that [currently] require a human being to trawl through thousands of pages just to ensure compliance. That is a use case for generative AI: to remove that labour, and allow staff to perform high-value tasks. The model is trained specifically to check for certain compliance checks,” he said.

Another example use case is staff and candidate grading, with candidates being assessed for certain examinations.

It’s very important that the human makes the decision. This is AI-driven automation supporting the decision-making by human beings

In both cases, human labour is greatly reduced by the application of AI, but the human dimension should not be forgotten, as AI is, at its core, a data-based automation technology intended to assist human decision-making.

“You keep the human in the loop, at the top. It’s very important that the human makes the decision. This is AI-driven automation supporting the decision-making by human beings,” he said.

The implementation process sees OpenSky, which is a Microsoft partner and externally-audited, train the AI model and integrate it into an organisation’s systems. An important aspect of this is data governance, partly because automation demands good data, and partly because of tightening regulatory frameworks.

“Let’s say you have got a document repository in your business and had it for ten or 15 years, it's a mistake to just sit your GPT [AI] on top of that.

“There could well be a whole bunch of documentation that was not approved or is outdated, for instance, so there's work, in terms of data quality, that needs to be done,” he said.

“It’s very important that you apply a level of data governance, which you should have in your organisation already, but it becomes more important with AI and we have NIS2 and the EU AI Act coming down the track.”

Of course, AI in itself is not the goal of digital transformation – indeed, no technology is – but its application not only helps organisations achieve the desired efficiencies, it also, by being implemented in a thorough and compliant way, prepares the ground for ongoing digitalisation.

“You might have someone who knows they have a legacy overhang, let's say Excel sheets sitting on a fileshare, containing very critical info. By bringing in data governance rules, the first thing you do is ensure they are not able to be emailed outside the organisation,” he said.

However, it can be intimidating for organisations to think about digital transformation if they have a significant reliance on legacy technology. This is understandable, but digital transformation is more a process than an event, Flanagan said.

Tools such as Microsoft’s new data platform Fabric support making moves step by step or project by project.

“Microsoft Fabric is a consumption-based model so you can take an agile and incremental approach, built out with a zero-trust model,” he said.

The end result is that, as the transformation programme progresses, data from all core systems become available to the users who need it, which in turn drives the results that matter to the organisation.

“Your ERP, your CRM, a bunch of access databases, some Excel spreadsheets, they can be moved over and have the governance rules applied. Then we can build data models on top of that, and do it in an agile way,” he said.