Is your government department ready for an AI transformation?

Pressure is being felt by many government departments around Australia to become more automated and transform to include Artificial Intelligence into their strategy.

Understandably the benefits are exponential. That said, if a department hasn’t got the foundations right the project can easily run into a myriad of issues.

In this article, we look at some of the considerations Australian government departments need to make prior to embarking on a major organisational change to AI processes.


Why the change?

Understanding the objectives

Like all good business cases, a deep understanding of “why” and “to what purpose” should be first-off the rank when assessing a major new initiative.

Whilst the primary driver for switching to AI is to predict outcomes, forecast more accurately and aid in planning, the triggers for the switch can come from a multitude of reasons:

    • Service planning and policy setting
    • Structural change within departments
    • Keeping pace with technological advancements
    • Service improvements, speed of delivery and higher volumes of output
    • Outdated legacy systems and processes
    • Internal growth pressures
    • Departmental alignment, etc

Prior to going down the AI path, however, it’s important to ascertain what the modus operandi would look like if a transformation isn’t adopted – along with cost implications down the track.


What’s involved?

Four key areas that need to be front of mind

Thorough planning and agreement across the board needs to incorporate four key pillars:

 1. People

How are your teams operating now? Is information collected and recorded manually and kept on individual spreadsheets? Do the teams operate in silos, only collaborating when necessary?

A Change Management program will be essential for culture, morale, and ongoing productivity, which the Human Resource department will need to work closely with you on – particularly when it comes to messaging with internal communications, training for the new AI systems and working procedures throughout the transformation to AI.

A number of additional questions will also need answering:

      • How quickly can team members be brought up to speed?
      • Are specialised skills required, and what are the financial ramifications?
      • How can the change be conducted to avoid loss of productivity during implementation?
      • What cultural, communication and supportive elements would be needed for teams

2. Data

Since data is the lifeblood of any system, the management and application of it should be a business case in itself. Without reliable data, functionality is lost, reporting is unreliable, forecasting is incorrect and AI processing can be inaccurate. Data quality simply must meet the current and future business objectives that are set out in your AI transformation plan.

Considerations should include:

    • Where is your data currently kept? Is it available and accessible?
    • What is the quality of the data like? Is it up to date, complete and fit for purpose?
    • How should it be cleansed and migrated to any new platforms? Should this be outsourced?
    • What are the security and privacy implications around data collection and use?
    • Do you have Data Governance procedures in place; do they meet guidelines, regulations and departmental policies? Do they meet new business objectives?

Obviously, these are just a few of the questions that will need attention. We cover more data considerations below.

3. Technology

There are plenty of Artificial Intelligence platforms and tools on the market – all of which are highly configurable for departmental requirements. The big consideration is whether the department chooses a Deep Learning (or Machine Learning) framework or Narrow AI that requires human oversight? Ideally, your business objectives, legal obligations and planning process will dictate the most suitable technology to adopt.

Legacy technology should also be considered for incorporation into any new framework – if financially viable in the medium and long term.

Also, it’s equally important to think of future requirements. Can the framework be scaled to meet growing demands, or be adapted to avoid tech redundancy, for instance?

Whatever technology is agreed upon, it’s important to not get carried away with functionality and lose track of what your original objectives are.

4. Process

Buy-in from other business units will be imperative. Also, as with all Project Management, it’s prudent not to be overly ambitious with timelines. In fact, you may want to build in extra time and financial padding to accommodate holdups; a common occurrence when multiple areas are involved.

Effectively, your rollout process should look something like this:

1. Identify needs, objectives and any restructuring required

2. Audit your data and develop a roadmap:

2.1 Assess where it currently sits, who owns it, how it’s updated, how many versions exist, how a true source can be developed, whether it’s fit for purpose within the new AI platform etc

2.2 Develop Data Management processes to ensure existing and new data are automated, well-governed, consistent and meets standards and formats

2.3 Training of business users for collection and maintenance protocols – ensuring universal standards, procedures and definitions are adhered to

2.4 Consider a Data Management Department or specialised Data Manager

3. Design and build of the AI platform

4. Data migration to the new framework

5. Testing and fine-tuning

6. Training of the new system/s with teams

7. Ongoing maintenance, development and data governance.


Identifying obstacles

Plan for everything

Project Management should also take internal appetite for change into account. For instance, would it be more acceptable to conduct such a large project over an extended period in phases? Or should it be dusted off as quickly as possible?

If conducted over shorter increments in multiple phases, the opportunity to test and learn as you go is easier and can inform future rollout.

However, with some government departments, it could be wise to implement the initiative as quickly as possible. Especially if there’s a chance the project could be mothballed due to changes in policy or leadership.

Naturally, cost is another major consideration.

Can legacy infrastructure, equipment and processes be incorporated into the new modelling? What will the upheaval cost in terms of downtime? And then there’s the cost of ongoing management.

How do these costs compare with operational efficiencies further down the track? Is it a good investment? These questions need to be answered in a sound business case that includes a well thought out risk analyses.

Moving forward

To cover all bases above, bringing in an external team that specialises in Data Management and AI transitions could prove your least costly and stressful approach – either for collaboration on specific phases of the transition, or for complete oversight.

In our next article we’ll do a deep dive into why data is such a key element for any transformation to Artificial Intelligence including:

    • Why it needs to be accurate and complete
    • How it should meet business objectives
    • The importance of good governance

If you want to discuss your data requirements and technical processes – or find out more about transitioning to Artificial Intelligence – Data Agility can advise you about end-to-end data analytics solutions; from strategy development to oversight and implementation of the system transformation. For consultations, contact us today.


Submit a Comment

Your email address will not be published. Required fields are marked *

Most popular insights.

Effective use of technology and data in sports

Effective use of technology and data in sports

Sports across codes and the world expect decision makers, like referees, to make perfect decisions with the tech and data they have at their fingertips. The passion of the sports fan and the desire for the ref to make the correct ‘call’ is not going anywhere. So is technology to blame? Is the data inaccurate? Or is human error unavoidable? Let’s explore.

Data Lifecyle Management eBook

Data Lifecyle Management eBook

This eBook shows you the best practices on data lifecycle management; collect the data you need, store it securely while you need it, dispose of it when it is no longer needed.