Artificial Intelligence (AI) strategies are essential for your organisation’s innovation and evolution. An AI strategy is a strategic plan of how to systematically incorporate AI into your workflow to reap the benefits of its innovations. Here’s what you need to consider before implementing one yourself.
Unless you’re a big tech company, you will likely rely on an AI consultancy like Boston Consulting Group or IBM to assist with your AI strategy. Before you get to that point, though, it’s useful to think through your organisation’s needs, goals, and limitations in conjunction with AI adoption, deployment, monitoring and ethics so that you can develop an AI strategy that is well-thought-through.
Why your organisation needs an AI strategy
AI has been adopted and used across the globe as an opportunity for nations, organisations, and individuals to remain part of the continuously evolving global economy that implements state-of-the-art technology and innovation as part of the 4IR.
To put this into perspective, the need for AI solutions is ever-increasing to a point where nations and big-tech companies like Amazon, Google, Tesla, and Meta are spending billions on AI research and development.
AI strategies contribute towards making profits, lowering costs, and providing value to the customer and business alike.
While many people recognise the importance of AI, they may not know where to start.
What to consider before implementing an AI Strategy
Align your strategy with your business’s needs and goals
Before running with an AI strategy, you must consider whether your strategy aligns with your organisational needs and goals to add value to your clients and your business.
It’s pointless innovating for innovation’s sake when it doesn’t add value. For example, you have a plan for autonomous AI implementation, however, you don’t know if you have the capacity to fully integrate it into your business operations.
Rushing into an AI strategy that isn’t linked to your core business value proposition can lead to resource wastage.
Here are some useful questions to ask yourself when thinking about a strategy:
- Does the AI strategy add to your organisation’s digital transformation? For example, does it improve annual revenue?
- Does it allow my customers to make better-informed decisions as a result?
- How do the outcomes of the AI strategy improve conditions for the client, organisation, staff, and society?
- What needs are you trying to address? For example, are you implementing an AI strategy to create new products and services to remain competitive within the industry? Or, is it to attract new customers or generate new avenues for revenue?
- What goals and objectives the organisation wants to achieve by developing and implementing an AI strategy? For example, are you trying to reduce time and labour costs, improve quality control, increase revenue, or improve decision-making?
Establishing the links between your organisation’s needs and goals and your AI strategy ensures that it directly addresses these needs. It also enables you to be more explicit when executing your strategy.
Consider how AI affects your people
One of the biggest hindrances to adopting AI is your organisation’s people. Employees can see AI as a threat to their jobs, and you should approach this sensitive area with thoughtfulness.
Augmentation and automation are the two standard methods of implementing AI—there are other caveats, such as assisted intelligence. Augmented AI assists humans in making better decisions, while automation requires minimal or no human input.
While AI has sometimes replaced human jobs, it does not apply to all instances. Roles that require completing repetitive tasks are prime candidates for AI automation. Non-routine tasks, such as unstructured physical tasks or non-routine cognitive tasks, are areas in which humans still play a vital role.
For example, AI models can be used in medicine to provide diagnostic predictions a doctor can use to make informed decisions. The AI requires the human to be in the loop and is a form of AI augmentation.
On the other hand, self-driving cars are intended to be autonomous to some extent. The same goes for robotics. Both of these are autonomous approaches to AI.
Organisations should see AI as a complementary innovation that can free employees of routine tasks and allow them to focus on higher cognitive tasks that require creativity.
The first step would be to educate your staff on the benefits and challenges of AI automation and augmentation within the organisation and what employees can do to improve their opportunities for development and growth. AI consultancies like IBM or Boston can help provide insight and education for your staff.
Education is vital for the development of your organisation in AI advancement. In other words, ensure your relevant employees at all levels keep up to date with the latest trends, technologies, and developments.
Choosing your approach to implementing an AI strategy
The next step is to think about how you’ll implement AI. How do you actually do the augmentation or automation? For example, you could have a long-term vision for the AI solution(s).
However, you could start small and create smaller iterations considered most important for your AI strategy. From there, you can build a timeline for AI innovation deliverables and features that work in iterative cycles. Your approach is essential here because it affects the project’s schedule, cost, and quality.
Determine how you’ll acquire data
A critical part of a successful AI strategy is data acquisition. Data is the ‘new oil,’ and AI can be seen as ‘electricity’. It helps if you know how you’ll acquire data before creating your AI strategy. Knowing this before implementing your strategy enables you to set realistic timelines and budgets.
For example, data could come from external providers, such as third-party companies willing to ‘sell’ the data at a fee. It could also come from governmental departments and NGOs ready to make it available for the benefit of society. The data could also be self-created through the company’s own data-gathering techniques, whether via the client’s data or data collection through, for example, R&D.
Another essential resource to consider and plan for is the human skills required to develop AI models that ingest data. There are different ways to acquire ‘AI talent’. One example would be to search for AI specialists available in the market. This way can be expensive due to the high demand for these skills. In addition, retaining talent is another challenge. However, this challenge exists regardless of whether these individuals are acquired eternally or not.
Another option is equipping your staff with the skills necessary to contribute to AI solutions, although the time horizon for this option might be longer.
An alternative, which many big-tech companies have done, is to acquire the skills through acquisitions. This happens when they buy out the company/competitor intending to acquire the extant AI solutions and the AI skills.
Other factors to consider as part of your strategy
AI is regulated at various levels. This can create limiting conditions on your strategy depending on the regulations of your particular location. For example, at the state level, you need to know the rules, processes, and policies to which you must adhere, especially around privacy, which is expanded on below.
Further, if you are a multinational organisation, you must know different countries’ and regions’ governance laws. This affects how you manage your data within your organisation. It affects whether your data will be stored in the cloud or on-premises, who will have access to it and when and which technologies will be used to manage it.
For example, depending on how long you must store the data, you will need to choose either a data lake, a data warehouse or distributed transactional databases. There might also be further limitations depending on the industry you’re in. For instance, financial data might need to be stored for a certain amount of time due to auditing.
Figuring this out beforehand helps you to plan by knowing the limitations that exist due to regulations.
Another important aspect that needs your attention is the security of your data and AI models. This is on top of general information security procedures such as network security, penetration testing, intrusion detection and prevention methods.
For example, when you collect data, you need to consider the potential threat of data poisoning or back door attacks. Data poisoning involves hackers altering the data to affect your model’s predictions adversely.
Similarly, backdoor attacks involve malicious hackers altering the data just enough to cause a different prediction than expected.
From the model’s security perspective, you could have model inversion attacks or model stealing attacks. These attacks are incursive methods whereby the attacker tries to ‘reconstruct’ the data or the model through ‘querying’ methods.
Because this can have serious implications for your business on multiple levels, you need to be adequately prepared and make sure you’re familiar with the risks while also trying to protect yourself as best as you can.
Data privacy issues are a massive concern for governments, organisations and individuals implementing an AI strategy.
Governments can create legislation that slows down or deters corporations and groups from inappropriately gathering, using, and sharing individuals’ data, which often can be used to identify individuals.
That led to a big drive to create data protection laws such as the European General Data Protection Regulation (GDPR) and the South African Protection of Personal Information Act (POPIA).
In addition to this, individuals and companies are hyper-aware of protecting their personal data, so you need to protect your customers’ or users’ privacy. In general, when developing your AI strategy, the more sensitive the data is for the client, the more valuable it tends to be.
Real-world data improves your models’ performance, making them less prone to failure. However, with privacy being a challenge, using methods such as differential privacy lowers the risk of data being personally identifiable. Whatever you do, you need to carefully consider how you protect user data to comply with the law and to ensure your customers don’t lose trust in you.
How to plan for AI deployment
Before you deploy your AI, you need to plan for it.
For example, are you deploying it on the cloud as AI as a service (AIaaS)? If so, you need to consider challenges such as model prediction latency.
Will the model building happen in real-time or pre-built? Real-time slows the process of predictive and prescriptive insights. This is why some neural network AI models are pre-built to expedite the process.
Infrastructure and security always play a massive role in deciding the preferred deployment method.
In addition, you should think about how the deployment will happen. Will you deploy with or without human intervention? Like a DevSecOps pipeline where you can deliver (with human intervention) or deploy (without), you have the option here. Generally, you want the human in the AI loop to improve the chances for success.
Deployment considerations are essential before implementing a strategy so you can figure out timelines, business impact and goals, and how your staff may need to be employed in different areas.
Monitoring the progress of your strategy
When your strategy is implemented, the process needs to be monitored continuously to measure impact and refine the strategy. In other words, the same way you manage a typical company project.
Similarly, an AI strategy implementation must be managed to ensure it is on track. Therefore, a CTO, CIO, CDO, or any team leader would need to ensure that the strategy is on track to being fulfilled.
You can use different project management frameworks to do this. However, the ‘leader’ making sure the AI strategy is on track needs specific leadership skills. These skills include being a good coach for the team members, being able to execute the strategy, collaborating with stakeholders, and reflecting on progress made and whether to pivot.
Thinking about this ahead of time is crucial to prepare staff who may need to take on additional responsibilities, or to ensure you hire the right people and skills.
The concerns demonstrated in previous sections can be seen as an AI strategy framework for your organisation. The following framework can be used as a guide when developing your AI strategy.
With an implementable AI strategy, you are giving your organisation a competitive edge in this fast-paced and ever-changing economy.
So the right time to start is sooner rather than later.
- How Tangent Solutions Used RPA and AI to Speed up COVID-19 X-Ray Screening
- Quick Guide: Introducing AI To Your Company
- Ashi Krishnan, Apollo: Robots and Their Future in Our World
Ridewaan Hanslo is a lecturer at the School of IT, Department of Informatics, University of Pretoria. He has more than two decades of industry experience and has been teaching since 2011. A typical day includes providing value-driven insight into aspects affecting project outcomes. Ridewaan regularly contributes to initiatives in Science, Engineering, and Technology.