Discover more from TeamCraft
Delivering Data & AI Strategy
Part 1: Understanding your dire situation
“We want to be data-driven.”
A timeless classic 🙃
Especially hilarious when it comes from the CEO/the-powers-that-be.
Because you know, they are more likely to achieve the outcome than a CDO/Director/Head of Data. Finance tends to have their data in order because if they don’t, you are looking at a big-time fine.
What’s preventing the same happening to marketing, sales, customer success, product data etc.? The decision-makers simply don’t hold them to the same standard—because they can get away with it. Even though this type of leniency is what ends up killing your company in the future—or ensures that your company will never realise its true potential.
One of the most difficult endeavours every company needs to undertake to go from good to great (amazing book BTW) is digital transformation.
I guess lucky for us data leaders, senior management still tends to hire folk like us to transform the company to be more data-[insert theme of the day].
<start rant> On a slight tangent, I used to argue we shouldn’t aspire to be data-driven, but data-informed. My thinking was that blind data-driveness leads us to astray; data needs context to be useful. The example I used to give was the number of inmates in US prisons by race. Looking at the numbers shows that historically, more black people being sentenced compared to other groups.
If you are merely data-driven, you can might as well stop there—go formulate policies based on this fact. Being data-driven doesn’t incentivise you to understand the root causes.
However, the right policy surely depends on not a statistical correlation, but an underlying causal mechanism. Being data-informed makes you ask the next question—why are there more black people sentenced in the US? And the next question, and the one after that—until you identify an actionable root cause.
I since stopped correcting people; as it is not about being data-whatever, but delivering a successful cultural transformation <end rant>
Nowadays, data leaders are asked to wear two hats: A lame Data & Analytics helicopter cap, and an elegant AI fascinator hat.
For clarity, Data & AI strategy can be: Data & Analytics, Data Science & Analytics, Analytics & AI, AI & BI (?) etc. The important point being there are two domains:
One pertaining to data infrastructure, data governance, analytics, BI. It is about ensuring valid and quality data and data processes.
Another focusing on statistical modelling, data science, machine learning, and AI. It is about utilising labours of the first stage to bring value.
It is easy to be enchanted by the AI hype. The story is not so dissimilar to the Data Science hype circa 2010 when DJ Patil coined the term while leading the data team at LinkedIn. Statistical models and statisticians became ‘algorithms’ and ‘data scientists’ overnight. When I pushed back against this (i.e. whining to my boss), his advice was to keep my mouth shut and ask for a bigger paycheck 🤷🏻 Now we all happily accepted the narrative of ‘AI’ and ‘AI researchers/engineers’ 🤑
So what’s issue with being tasked to lead a Data & AI function?
Well, the reality is that for the vast majority of companies out there, you need to absolutely nail down the Data & Analytics part before you can start dreaming about the AI bit. Bummer, I know.
Let’s consult the ever-useful data science hierarchy of needs by Monica Rogati. Was this page always called the AI hierarchy of needs? 🤔
You need to get the foundations right, before realising value from the more niche applications of DS/ML/AI. It is absolutely true that ML and AI can enable avenues that are previously not available to your company, and that you won’t be able achieve those with just descriptive analytics or a robust data infrastructure.
However, if you don’t have a robust data infrastructure and satisfy the data & analytics requirements of the product squads and business units, you won’t be alive to reap the benefits of AI at that company. You can probably clock-out at a respectable 18 months, which nowadays seems to be average tenure of a CDO. Yikes 🤮
But you want to flourish in your role and lead the AI transformation of your company.
You need to get your Data & Analytics game in order first. And this is not a technology challenge, but a people and culture one. What you will be doing is change management, and let me tell you, it sucks. People don’t like change, even when it benefits them. Imagine the typical startup data mess:
No data culture—aka we grabbed the first BE we ran into and they put together a database n years ago
Bad history with data function—broken trust because of unfulfilled promises
Shadow data sources—not trusting the central data function, other parts of the business are forced to solve their own problems, creating parallel data sources and dashboards while paying governance no mind
Fixation on tech that nobody wanted or uses—data function demoes something and all end users go ‘cool but who asked for this?!’
Over the next several posts, I will cover how to set and execute a Data & Analytics strategy in your first three months at a job that sets you up for success when the time comes to realise value out of your AI initiatives.
Hope you are up to the challenge 😉
Thanks for reading TeamCraft! Subscribe for free to receive new posts and support my work.