Issue #11 - Dispelling the AI Hype Train
AI cannot be developed in a vacuum. Here is how to think of it holistically
Read Time: 10 minutes
AI is currently the bee’s knees, the cat’s pyjamas, the elephant in the room. It is all everybody wants to talk about, especially business executives looking to make an impact.
Yet I still haven’t talked about it in this newsletter. Until now…
Is AI the game changer it is marketed as?
I’m a big believer in doing things properly. If you don’t put in the work, you won’t realise the gains. The same is true when it comes to AI.
To answer my question above, I do think AI is a game-changer. I’ve used ChatGPT, perplexity.ai, and several other AI-enabled technologies. I have also seen it used in organisations to optimise supply chain stock picking, answer questions from relevant internal documentation, and improve pricing decisions given a multitude of complex factors.
That being said, while AI is frequently hailed as the ultimate solution for a myriad of business problems, this perception can be misleading. Misconceptions about what AI can do and how to get there is leading to countless misguided investments and strategies, resulting in wasted resources and missed opportunities. The most common misconceptions I see are:
Regular teams can build relevant AI solutions
It is worth investing in AI without a clear plan or strategy
You can do AI without setting up the right data foundations
Our company and workforce will automatically be more productive with AI
All of these perspectives take a simplistic and short-signed approach to Artificial Intelligence, which is very common when decisions are being made by extremely busy executives with no experience or time to think about the domain. So what do you need to know to plan accordingly?
The AI Vacuum
One of the significant dangers of the AI hype is when decisioning comes from the board or leadership group that doesn’t understand data. AI is treated as an isolated solution, detached from the broader organizational context, data & technology realities and strategic objectives.
When AI is thought of in a vacuum, companies risk simplifying an extremely complex digital transformation leading to inefficiencies, suboptimal outcomes, and even failures.
After a lot of digging, I finally found some proper data about the success of AI projects (don’t worry, this is not the 5-year-old 87% of data science projects never making into production stat). This survey showed that from a base of over 2,000 individuals, between 36-56% of AI and analytics projects failed. The three main reasons were:
Organisational Friction – Finding the right talent (75%), upskilling employees (58%) and getting the workforce to understand and use AI properly (35%) were the main detractors here.
Technological Friction – Legacy systems (33%) and the over-complex solutionisation of AI (63%) were stated as huge barriers from a technological perspective, alongside classic roadblocks like data quality or processing power.
Financial Friction – Cost implications is another overhanging consideration, with financial constraints (25%), question marks about the return (28%), and a high cost of implementation (33%) also deterring AI projects and investment.
What this all shows is that AI cannot operate in a vacuum.
Successful AI deployment needs to consider these things by finding the right use case, planning how the AI capabilities will integrate into organisational operations, and cultivating an AI-ready culture by evangelising the benefits throughout the workforce. All said, there are a lot of components you have to consider when thinking about AI.
AI in the Wider Data Ecosystem
Given the goal of this newsletter (to dispel the complexity of the data ecosystem), the subject of AI implementation fits right in.
Artificial Intelligence is one of the most complex and misunderstood parts of the Ecosystem right now.
What makes it so complicated is how pervasive AI actually is across the Data Ecosystem. An executive might think about it as one step on the journey, usually in the consumption, tooling, or solution phase of the data lifecycle. But in reality, considerations for AI have to made throughout, from sourcing data to processing & managing it to consuming and integrating it into the fabric of the organisational culture and operations.
Let me rephrase this: AI is not just an LLM Chatbot or a price optimisation tool; these are only the use cases that people are currently selling, using, and building. Instead, AI is a fundamental disrupter that will evolve how we work and impact every area of the data lifecycle (and organisational operating model) that we know.
What do I mean by this?
Business Model – Revenue streams and how value is delivered will change. AI will implant itself in how the business operates evolving the way the organisation works.
Data Sourcing & Generation – What data you source/ generate, how you source it, and why you source it will all evolve. For example, companies will consume more automated data, potential synthetic data, and will need more data to train different AI models.
Data Availability & Usability – AI requires a lot of data. Companies are going to feed these models everything it can get. Data engineers and analysts need to understand this and ensure the quality of data does not misrepresent reality when fed to these models (today, most companies have poor quality data which would do just this).
Use Cases – Today most use cases are business stakeholder driven and focus on lower-level analytics. With employees starting to use AI tooling and Chatbots, even lower-level descriptive analytics tools will probably have elements of AI for searchability or insights-enablement. Co-development of these use cases will be even more crucial.
Augmenting Data & Business Teams – Understanding how to use AI with existing employees and training them to take advantage of it. This could mean automating routine tasks, using it to get better insights, or doing more advanced forms of analysis.
Spend & Investment – AI investment cannot be solely borne by the data team. Given the high price of human, compute, and tooling resources, there has to be a more sustainable solution in order to justify the high cost and showcase the returns that AI can provide. Proving ROI has always been a weak spot in data teams, it will also become a huge issue for AI spend.
Data & AI Tooling Consumption – Every tool is now “AI-enabled”. Discerning what your company needs and what they don’t to get past the marketing hype will be crucial. AI makes a complex tooling landscape even worse, especially given the previous point about poor ROI.
Organisational Culture – AI will evolve faster than your company’s culture. How can people keep up, feel valued, and work with instead of against progress? Leadership needs to clearly articulate “why” they are going on this AI journey and what the benefits are for certain individuals.
Right now we are just seeing the potential of AI. The waters are murky and only a few things are coming to the surface. Without considering the different implications and route to cultural adoption or value delivery, there is a huge potential for organisations to fall flat when it comes to AI implementation.
How to Approach AI from a Holistic Perspective
Okay, let’s dig in. First thing I want to make clear is to not just skip all the data fundamentals and lower-level analytics and jump to AI. This is my biggest callout and what I see a lot of companies doing wrong. Check out my previous articles on data investment and determining business stakeholder needs to better gauge how to deliver data use cases in a sustainable, logical way.
If you do think you are ready, great!
With the expanse of the Data Ecosystem in mind, here are the six considerations you need to make to deliver against your AI ambitions properly:
Set the AI Strategy/ Direction – Companies often jump straight to building tools. This is short-sighted, expensive, and just plain dumb. Given the magnitude of AI and its potential impact, companies need to define clear objectives aligned with their business strategy and goals. Understanding how other industry competitors are using AI and mapping out short- to long-term milestones creates a focus for the data and AI teams to track against. Moreover, the strategy should have KPIs that help leadership evaluate the benefits and ROI of any AI initiatives, which can justify and inform investment.
Use Cases – With the AI strategy/ direction set, use cases become a lot easier. Talk to business stakeholders, assess existing pain points, determine the feasibility of delivery, and calculate potential ROI aligned to KPIs. With that baseline, match it up to potential AI use cases, while considering stakeholder readiness, data foundations and analytical tools that already exist. This isn’t rocket science, or a new approach in any way. But teams don’t spend enough time on this. By taking a week to identify, scope and prioritise AI use cases, you can save months of wasted time (and tons of money) in the future.
Understand Business Model Implications – I mentioned this above, but it is worth reemphasising. Companies are rushing into AI investment and use cases without thinking about how AI should and will change business processes, revenue streams, value propositions, and the overall operating model. Considering implications at the outset (like how teams will work in the future, what we will be selling, or how AI will evolve the supply chain/ operations), especially aligned to prioritised use cases, will help teams build more sustainably and approach AI from a proactive, rather than reactive, mindset.
Cultural Change Management – This is an area most companies are blindly ignoring. Leaders need to think about how to engage the business, communicate AI goals, and understand how to build collaboratively instead of just forcing AI down the throats of the unsuspected. This helps create an acceptance mindset rather than a “is AI going to replace me” worry, leading to adoption, improved productivity, and further innovation. Another huge element of this is establishing the guidance around doing AI ethically. Tech likes to work with the build-fast and break-things mentality, but there are too many risks to this approach with AI and part of building an AI-enabled culture is considering and pre-empting what could go wrong and what shouldn’t be done.
Data Foundations (Data Model, Quality, Tooling) – Data is the lifeblood of AI, while the underlying data tech makes up the key organs that allow it to run. Companies need to invest in the necessary platform/ technical infrastructure and data quality/ governance teams to ensure these components are healthy and productive. Some example steps here are to model the data so it aligns with new business AI processes, support these data flows with proper quality controls, and establish the right governance and tooling to underpin it along its lifecycle before feeding into AI models. These systems/ tooling rely on relevant and accurate data to function correctly, and without the foundations, the only output companies will get from their AI are data silos and inconsistencies.
The too real reality of where most companies focus… Teaming & Training to Support the Journey – If data & tooling is the backbone of AI, the organization’s teams and people are the brain that is directing it. Having people who understand data science, AI, domain expertise, and project management needs to be top of mind for companies, as complexities will arise that derail projects and make outputs useless. Coupled with these resources, is training to ensure the rest of the workforce can use and take advantage of any AI tooling. Having business-oriented AI pod leads or evangelists can help with this journey.
AI holds tremendous potential and the hype will be justified, but it’s essential to approach it with a balanced view and understand how it fits in with the Data Ecosystem. Remember, AI is just an extension of your data team and successes.
Note we haven’t touched on development, deployment, or implementation. We will leave that for a future article, where we can get into more of the specifics of actually delivering successful AI solutions.
Next week I dive into a super interesting topic, which is the three biggest problems I see in data today. This will be followed by a deep dive into each subject, an analysis you shouldn’t miss!
Thanks for the read! Comment below and share the newsletter/ issue if you think it is relevant! Feel free to also follow me on LinkedIn (very active) or Medium (not so active). See you amazing folks next week!
Such a holistic and balanced approach is exactly what's needed - thank you for laying it out so clearly!