The Importance of an AI Strategy
AI is not just a technological trend, it’s considered the 'single biggest impact' technology for differentiating companies, and is essential for your business strategy. If you’re not considering adopting AI for your business, beyond say, a Chatbot, I can relate. Many long established business leaders don’t see the benefits of intelligent technologies, or find a cost/benefit ratio that excites them.
It is the larger organisations that are shifting from their original digital transformation, and becoming 'algorithmic enterprises'; leveraging data and complex algorithms to drive competitive advantage. They have the funds, the technical know how, and the budget for external consultants; and these people are not only scarce, if they know what they’re doing, they’re also expensive.
If these established SME’s don’t at least plan for intelligent technologies, will they run the risk of falling behind?
The technology is not going away anytime soon. In the future there will potentially be more technical people to approach; many will have far more experience under their belt also; after several early project failures, people tend to learn what works and what doesn’t.
Many of the current 'experts' will tell you that if you don’t adopt AI, or have an AI Strategy in place, you risk falling behind competitors who are already seeing higher profit margins from early AI adoption. So it may be prudent to analyse your competitors, not just for marketing strategies, or their recent hiring, but for their approach to adopting intelligent technologies.
I have heard so-called 'experts' saying how, “It’s too early to do everything, and it’s too late to do nothing.
But if you long established competitor is biding their time, what is the rush?
Perhaps, they are waiting for you to make the first move?
Either way, early adoption can certainly reduce costs in certain areas, it can also take time to recuperate the original cost of implementing the technologies, along with training, and a bedding in period for staff, to get to grips with using it in real-time.
With the advent of the EU AI Act, and future regulations over AI adoption, you will find that there will always be a demand for basic AI training for staff. It is possible that the powers that be see us all working from hime, with AI systems providing the middle management tasks.
Of course, with the emergence of Generative AI many jobs have also become obsolete. The young have the opportunity to retrain, those like me do not; time seems to have caught up, and learning new technical things just doesn’t have the appeal it once did.
I can take my current skillset, and experience and apply it to emerging technologies, as well as legal Acts, regulations, guidance etc., perhaps many can do the same, after all, leadership, people management, project management, process management, and all the support services are not going to change, but possibly the number of people required in the organisation will.
Transferable skills and experience are always valuable, at least until you reach 60-years, then you become invisible; your resume is like a hot potato and quickly dropped in the proverbial bin. But I digress …
Let us look at the goals of adopting an AI Strategy, just in case you decide to cover your bets.
Warning: I am apt to digress at times, so please forgive me—
A well-defined AI strategy should aim to achieve several high-level benefits for your business. I will list them here, in a format that can be used as a reference point:
1. Eradicating Repetitive Tasks: AI, often through Robotic Process Automation (RPA). This can automate mundane, knowledge-based tasks that require little cognitive effort, taking the 'robot out of the human', so to speak. This will allow you to reduce costs, human errors, and prevent staff dissatisfaction, freeing them up for more creative work; like polishing their nails, or checking their FB page.
2. Generating Insights and Predictions: Machine learning and deep learning systems can automatically extract previously undiscovered knowledge and insights from vast amounts of structured and unstructured data. Which is great until you realise you don’t have vast amounts of data, and your current servers could not cope with vast amounts of data, if you could indeed gather it and store it.
Note: Machine Learning, or ML, you will know as AI models, like ChatGPT, and there are many more. Whereas, Deep Learning is a subset of ML, and where the real work begins; delving into the data to provide true insights. Deep learning focuses on utilising multilayered neural networks. Think 'Terminators'—
Using ML, the insights can enable better business decisions in the areas of acquiring new customers, increasing revenue, reducing customer loss, and optimising your chain; supply chain that is.
3. Amplifying Human Intelligence: Intelligent technologies can provide true contextual knowledge and support to your staff and customers. They — being the egg-heads — call it 'augmenting human capabilities', and sell it as a way to keep staff rather than replacing them with a machine.
The upside for management, is the potential for what 'they' call 'business clairvoyance'. This is where you can anticipate and prepare for situations, enhancing human performance and decision-making. Simply put, it is 'predictive technology, which is what Generative AI, like ChatGPT has the ability to do; depending on the data input. Same analogy applies “Shit in, shit out.”
So what are the challenges of Implementing AI without a clear strategy?
You might struggle to move beyond your pilot projects, or your 'proofs of concept’ with AI. You will be hitting several barriers, and have hoops to jump through. I will list some examples here for your reference, and to scare the be-jeebies out of you.
• Hype vs. Reality: There is a real lack of deep understanding of the different types of AI; plus a lot of new terminology to grasp. This can lead to confusion about the actual capabilities of AI, its true value, and the all important ROI. These things alone can hinder adoption.
• Bad Data: AS mentioned, your friendly little AI needs a lot of high-quality, and relevant data. You might face the issue of 'dirty data'; missing values, incorrect information, duplicates, etc., or simply you don’t have sufficient data and a way to collect it? A lack of data will derail your projects going forward. Insufficient, or poor data will lead to inaccurate predictions. Data cleansing and robust data pipelines are crucial, but often overlooked and the cost will astound you.
For an SME this might be the biggest hurdle to adoption. If you want to have a system that provides clear and current outputs, then you need clear and current input. I would suggest your earliest project is focused on data collection, and storage.
Just ensure you have your Data Protection in place.
• Disparate Use Patterns: In a nutshell, different departments will more likely have unique needs in relation to AI, but similar underlying algorithms or data structures could serve multiple purposes. This simply means, data sharing. The AI will do all this for you, so no alarms should be going off just yet.
Obviously, without a comprehensive strategy, you might implement disconnected solutions, which will lead to duplicated efforts and higher costs; so share that data in-house as much as possible, design it in.
• Complexity and Emerging Technology: Intelligent technologies are inherently complex, and the little bast@^*$ are constantly evolving. Of course, it is difficult for SMEs to keep up. There are enough issues choosing the right tools as it is, and all this results in AI solutions being implemented in what the egg-heads call, Silo’s.
• Scarce Talent: As mentioned above, there is a significant shortage of skilled AI practitioners. This is a world-wide issue, making recruitment challenging to say the least. One cannot Balme the top talent seeking the best roles; environments with strong teams, diverse problems, clean data, and robust AI platforms.
Which pretty much rules out most SMEs right? Like I said, in time they will be in abundance. Let’s hope the cavalry arrives in time.
• Misaligned Execution: Another egg-head term. So, even with a competent team, AI projects can fail if there isn't alignment among business users, IT, and AI experts. If you have managed to find one; they are as common as dragon eggs at the moment. It is no surprise that with a lack of understanding, in relation to business priorities, insufficient IT support or cumbersome data governance processes, can prevent AI models from being deployed. Obviously, if they’re not deployed, they cannot be adopted.
So what are the components of a comprehensive AI Strategy?
In case you manage to find an expert wandering the streets, or falling around in your favourite watering hole, then I can tell you that a successful AI strategy will require your 'expert-led' team to define several key deliverables:
NOTE: Don’t ever forget the regulations, these need to be carefully applied at all stages of an AI project, depending on location, and reach of your intended business services and products.
1. Goals and Vision: Clearly articulate the purpose of AI to the team, whether it's for sustained competitive advantage, new revenue opportunities, or even cost reduction. Make no mistake, this will involve a 'holistic AI transformation', supported with an element of digital business innovation, to say the least.
Of course, your expert is sobering up by now, so lock the doors.
2. Use Case Identification: Use cases are important, and define potential short-term and long-term ways AI can drive business goals. Create a "use case catalog”. This will help the team understand the overall business case, the required AI solutions, the information management requirements, and the user groups and processes affected by the project.
3. Architecture: Design the technology components and platforms to support AI, and its vast data requirements. This includes how to make existing business-wide architecture AI-centric. You will nee do cover the processes for data gathering, storage, data analysis, AI modelling, visualisation, the expected user experience, careful model management, deployment, and integration.
4. Data Strategy and Readiness Plan: You need to establish rules, policies, and standards for managing data. This ensures relevant data is available, it is of high quality, and timely. This includes data monetisation (using data internally for value), and potentially data commercialisation (creating new revenue streams from data products/services). Which of course, is the whole point. No brainer.
5. Organisational Capability: Make sure you define the structure, talent, and processes needed to execute AI projects at scale. Management need to pull their fingers out, and make decisions on how AI teams are structured (centralised, decentralised, or federated), and also how to manage talent and skills.
Check the 'expert' has not escaped; tie the fu*^#@ up if you need to.
6. Governance and Change Management: You possibly already have systems for decision-making, so in this case ensure to consider AI model safety, as well as managing enterprise-wide deployment. Like you don’t have enough on your plate.
Simply put, this includes defining project ownership, standards, value measurement, prioritisation, and addressing market analysis, competitive assessment, and vendor evaluation. All standard stuff for a good PM, eh?
In this case, change management has never been so crucial. You need to ensure management and staff engagement, and reduce resistance to AI adoption. Make no mistake, you are not rolling out a new operating system, or some software tools. Adopting Intelligent Technologies is a complete overhaul, an enterprise-wide transformation. Once you commit, there is no going back; possibly too many staff would already have exited, or seen their roles evaporating before their eyes. I don’t want to discourage you, but this strategy will affect lives.
Let’s look at the steps required to develop an AI Strategy. It might not be news to you that the process of developing an AI strategy is systematic in nature:
1. Educate Executive Team: Of course, the leadership will by now have a collective understanding of the potential for AI. They would have read enough headlines to enable them to separate the hype from reality, and establish consensus on focus areas and objectives.
This is a given, most of the people I meet in any position of decision-making have done their due diligence. Possibly one reason they were in the bar in the first place, eh? I wonder why the expert was in the bar?
You don’t think he has issues with talking to people, and females in particular? Conversations with machines can possibly take its toll on your relationships, eh?
2. Form an AI Team: It is obvious you need to assemble an initial team, it worked for the Avengers, right?
Your team will need to combine internal talent with external collaborators, and new hires.Do you still have the expert Locke-up? Time to let him out me thinks, because your team will need him to drive 'use case discovery' and 'define architectural needs'.
3. Identify Specific Areas of Opportunity: You need to align on critical drivers for AI adoption. This simply means, you need to examine potential 'use cases', determine business objectives, your anticipated returns, and impacts on existing processes and systems.
So it is time to go fetch your highly competent PM from the bar.
Make sure to prioritise based on your business value and viability. Due diligence all the way.
4. Determine Readiness: Once you have assessed your preparedness across technical and non-technical dimensions, including data, technology, structure, governance, and supply chain, you need to Identify any gaps, and define a 'transformation path'.
Simply put, make sure you have thought of everything; removing any barriers to project success.
5. Develop a Strategic and Operational AI Roadmap: Don’t you just love these terms? Okay, ensure you create a high-level plan with priorities, milestones, and timelines. From above, you will have addressed any identified gaps, and have remediation plans in place.
6. Create a High-Level Business Case: Here is where the leaders will be really needed. You must conduct a qualitative, or order-of-magnitude assessment, for the costs (staff, training, software, hardware) and benefits (operational efficiency, revenue generation, competitive advantage, customer satisfaction). No pressure. If you are like many leaders I have ever met, you did this before day one of the project. Possibly, moments after you discovered the 'lone expert' wandering around the bar?
7. Engage the Broader Organisation: Conduct educational 'roadshows' to get all the staff left over comfortable with AI, fostering an AI and data-driven culture, and then mobilising them to leverage the AI systems. Not easy, as many would have seen their colleague bolt for the recruitment pages at the mention of artificial intelligence. Of course, they hadn’t thought it through.
How long before all businesses are utilising intelligent technologies?
Okay, so what are the 'Key Enablers for Scaling AI'?
Thinking beyond strategy, there are some fundamental elements that are crucial for scaling intelligent technologies, there are also many AI Centres for Excellence to help you on your path. I will share the obvious, but there are many more across the world.
ELISE - European Network of AI Excellence Centres: This is a network of AI research hubs, which are based on the highest level research, it spreads its knowledge and methods in academia, industry and society. The network invites all ways of reasoning, considering all types of data applicable for almost all sectors of science and industry.
NOTE: ELISE has been instigated by those unelected people, based in Geneva.
AI Centre of Excellence (AI CoE): This is the organisational entity responsible for executing the AI plan. It defines and maintains strategy for the AI platform, data, and talent across all business units. A federated model (hub-and-spoke) is often recommended, allowing local innovation while maintaining central coordination and knowledge sharing.
AI Platform: This obviously comprises of the hardware, software, and tools that accelerate the full lifecycle of AI projects at scale. It is a robust platform which increases productivity, enables faster experimentation, and reduces costs by providing self-service access to AI technology, standardising workflows, and allowing rapid deployment and monitoring. It should have an evolutionary architecture to adapt to changing technologies, and business needs.
Key components typically include a Data Minder (for data management), a Model Maker (for experimentation), an Inference Activator (for deployment), and a Performance Manager (for monitoring). These will either be new hires, or re-assignments of existing talent; trained and willing to get involved, because these are all responsible positions in the face of looming regulations. By understanding and addressing these concepts, SME leaders and business executives can establish a robust framework for any AI adoption. They are take bold steps, and transforming their business towards being an 'algorithmic enterprise' that leverages AI for sustained competitive advantage.
The biggest decision now, is whether you release the accosted expert, or keep him under lock and key?
NDA anyone?
This article is one of a series of articles, in preparation for the publication of my Substack newsletter: AI in Practice: Guides for SME Leaders
Along with my published books on Intelligent Technologies, I will be increasing the volume of my articles and online posts, with the intention of bringing clarity in relation to artificial intelligence; focusing on strategy, ethics, governance, and legislation.
With a background in government Statutory Compliance, within the UK construction industry, and years of experience carrying out Technical Risk Auditing for insurance underwriters, I’m turning my focus to emerging legislation and guidance surrounding Intelligent Technologies, such as the EU AI ACT (2024), the National Institute of Standards and Technology: NIST AI RMF 1.0 - the Artificial Intelligence Risk Management Framework, plus the companion resource, NIST AI 600-1: Generative Artificial Intelligence Profile.
I interpret and translate emerging laws and guidance in the AI space, to provide SME leaders and business executives with a clearer understanding of the requirements. My articles are provided in plain language, for easy assimilation by those who lack the time, which of course is your most valuable asset.