The Smarter Way to Build AI-enabled Decision-Support Systems

Empower your decision-makers with meaningful new insights within weeks. Calibrate Consulting explains how to get results now, while de-risking your strategic change programmes. 

The Smarter Way to Build AI-enabled Decision-Support Systems

Knowledge is power, as the saying goes. In business, this typically means having the right insights in front of the right people at the right times, to help them make informed decisions quickly. 

Like most organisations, you’ll probably be looking to improve your data-driven decision-making, by putting higher-quality information in front of people at all levels of the business. This might be with the help of artificial intelligence (AI), such as machine learning (ML) or deep learning (DL). 

Replacing the spreadsheets

It’s common for organisations to use spreadsheets at the core of their operational and strategic decision-making. This is perfectly understandable: spreadsheets are easy to build and share, and can be made to do quite sophisticated things. Indeed, these will have built up in abundance over many years, resulting in multiple layers of spreadsheets upon spreadsheets, often reaching the point they are regarded as business-critical applications. However, there comes a point when organisations outgrow these spreadsheets and manual processes. Contributing factors include the growing business-importance of the insights they provide, the need for faster decision-making, the increasing complexity of the data, and demands from different parts of the organisation to explore advanced analytics and AI. The next big step is typically to seek to replace these ways of working with more robust solutions that aim to make it simpler to deliver AI and advanced analytics, and to comply with regulatory legislation, such as the Sarbanes-Oxley Act (SOX), or policies such as around End User Computing (EUC). 

A catalyst for such a change is often the upgrade of a core application used in a particular domain, such as finance, HR or procurement. With a major transformation programme required to change such a key, transactional business system – and that application being a significant source of decision-making data – it’s seen as the ideal opportunity to tackle the challenges associated with related spreadsheet-based decision-support tools. 

The problems with making the wider transformation programme a dependency for any front-line informational change 

The usual approach is to build a new data layer as part of the transformation programme, and to do this before considering the wider, integrated use of the data. This layer can be either physical or virtual, and will sit on top of the new application. It will bring together data from this new system and other parts of the business, before eventually making all of this available to reporting and analytics software, perhaps coupled with some kind of AI capability. These decision-support tools and the visualisations they display to your decision-makers are typically also specified and delivered as part of the overall transformation. 

However, there are two significant drawbacks to building your business-important decision-support systems in this way. 

The first is the risk involved. For many organisations, this will be the first time they’ve stepped away from spreadsheets to create enterprise-grade analytics and AI capabilities. While it’s an incredibly exciting prospect, there’s a lot of unknown associated with it. Will the insights people think they need be the ones they actually need? Will there still be the same needs in 12-24 months’ time when the programme completes and the insights are finally made available? 

If the insights don’t meet people’s needs, adoption will be poor, resulting in low returns on your investment – coupled with further proliferation of spreadsheets and manual processes. 

The other drawback, which we alluded to above, is the time typically taken for a large-scale transformation programme to produce usable insights. Your decision-makers ideally need these within weeks, not years. 

Getting the insights sooner

The question then becomes whether there’s a way to get those important insights sooner, while also reducing the risks associated with building enterprise-grade analytics for the first time. 

Something we’ve done successfully for our clients is to build the analytics and reporting layer as the first step of the journey, rather than the last. 

To do this, we create a temporary data layer within this new analytics application, and initially use this to feed data from your existing spreadsheets to your new visualisations. In subsequent phases, we replace the temporary data layer with the permanent one, and phase out the spreadsheets and manual processes in favour of more robust solutions and data sources. 

This phased approach can deliver your first analytics results within a small number of weeks, including the discovery workshops, source-identification, modelling, data standardisation and initial visualisations. You then iterate on these based on user feedback, and later connect them directly to your new data layer, since they’ve been built with this precise architecture in mind. 

Driving down overall transformational risk

The great thing about this approach is that as well as giving decision-makers much faster access to an expanding range of insights, the early-stage work supports and de-risks your wider transformation programme. 

This is because the analysis and iterative build you do upfront to understand your users’ analytics needs, coupled with the feedback you glean from user-testing the early reports and visualisations, enables you to specify your new domain application and business processes with much greater confidence than you could otherwise do. This means they’ll be built from the start to meet your decision-makers’ actual insight needs, rather than just the operational/ transactional needs. 

Time to change your approach?

So, if your organisation is using a major transformation programme as the vehicle for delivering new and improved decision-support tools, based on enterprise-grade analytics and AI solutions rather than spreadsheets, we recommend starting by building analytical applications on the data you already have, then iterating on the visualisations and reports your people will be using. 

Having access to these as soon as possible means your business benefits faster from better-informed decision-makers. And in addition, the learnings you gain will drive down risk in your flagship transformation, and ensure this high-profile programme helps deliver maximum value to your organisation. 

Don't miss these stories:

Do you have a project in mind?
We combine expertise and creativity to solve highly complex data and analytics problems.