Small, Strategic Steps To Get AI Benefits Faster

By Peter Swart – Qlik presales consultant at iOCO

We commonly hear from customers: ‘My boss says we need to use AI, but I don’t know where to start.’ or ‘I’ve read all the horror stories of failed AI projects, so I’m afraid to throw resources at a risky project.’

Fortunately, you don’t need a big bang approach to AI. The most effective way to dip your toe into AI waters is to take small, strategic steps to solve real business problems.

Is AI worth the hype?

In a nutshell, yes! AI is worth the hype because it delivers easy-to-use capabilities that other tools can’t provide. It may even be seen as a game-changer, since it allows us to implement solutions that only humans could have done before – only faster, and more precisely.

We all know the question-answering, chatty capabilities of the popular LLMs, but what else can AI do? Firstly, it can make unstructured data – that which lives in all the PDFs and Word documents that are scattered around the folders of your shared drives – into valuable, searchable knowledge bases.

Getting started

Costly mistakes occur when AI projects are too ambitious or where they throw  technology at a problem without fully understanding what the problem is. Don’t start with the technology. It is important to start with a business problem that you need to solve.

Define the problem, then choose the tool.  Choosing AI first is like picking up an electric drill, and only afterwards deciding that you needed a saw to cut the wood.

To derisk AI projects and get the most value from them, start small. To move faster, buy a tool or capability – don’t try to develop one yourself.

You might use an LLM just to index documents, export data from one of your databases and get that indexed, and then finally bring the whole thing together without a lot of very technical effort on your part.

It’s also crucial that you have the right data and can trust your data.

Harnessing RAG capabilities

A RAG (Retrieval-Augmented Generation) application enhances a LLM output by combining its generative capabilities with an external information retrieval system. This turns your company’s private unstructured documentation into searchable knowledge bases.  Your private information stays private, while putting the power of internet trained LLMs to work to assist your staff in wading through what would have been a pile of paper or PDFs.

The RAG system will index your data, and because RAG based LLM solutions have been designed to read documents, and are not always good at looking at databases, you may need to reengineer structured data to be understood by a RAG system. The only coding you will have to write is the report that dumps your database into documents.

Such a system could be deployed to make your company HR policies or factory maintenance documentation available via chat bots, to save employees hours of work.

At iOCO, our internal HR policies AI project took just two weeks to gather the documents, and an afternoon to get the LLM to answer its first question, with links to the source documents for verification. This implementation has saved our HR team many hours.

Success in AI projects like these is not always easy to quantify. But if individual users are using the system, or time-consuming queries that previously went to humans start tapering off, the project could be considered successful.

At iOCO, we offer tools and services to help organisations take their first steps into successful AI implementations, including Qlik Answers – the RAG based LLM tool of choice, which is quick and easy to implement.