Data & AI Strategy: Defining Your Stratagems
Part 3: Selecting your moves
This is part 3 of the Data & AI Strategy series. Check out the below posts for the previous entries:
Assuming you have something in place for your Data & Analytics approach to hold back the four horsemen of data apocalypse, now itโs time to think about your Data & AI strategy.
The term strategy as we know it today goes back to 5/6th century Byzantine Empire, stratฤgia, which roughly translates to โart of troop leadingโ (i.e. generalship).
As middle managers and above, we care about strategy because:
Our resources are limited
There is usually more than one obvious approach/solution to our problems
We add value to the business via delegating work to others
Our line reports focus and operate on the tactical/operational level (by definition not concerned much re: the big picture)
Without going full Clausewitzian, letโs approach the challenge of setting a strategy from a combination of big picture and troop-leading perspectives.
The Big Picture
At this level, the company pays you because you have a vision for your department/functional area. It is assumed that you deeply understand your domain, challenges, solutions, as well as speculating about future innovations that might affect the status quo.
But the big picture is also the company positioning (not just marketing but ethics etc.), the competition, the industry, the effects of technology (e.g. โAIโ) on the industry and so forth.
However, it eventually all comes down to money in, money out.
The execs can and will throw unrealistic expectations your way. Perhaps your business shouldnโt be โAI-enabledโ, because it has no need for AI. This is rarely an obstacle for execs and investors. Or maybe, your top exec (CEO) says they are investing in AI; now go make it happen middle manager responsible for AI. Then you ask how much we are actually investing in AI, and they go ๐คท๐ป
Actually holding execs accountable for such internal investment promises can be tricky. But only because they have the power asymmetry in their favour; not because their arguments are superior.
In that vein, the term โinvestmentโ sounds like a good thing but is treated like a veiled threatโyou better deliver, because we are investing in your department.
Are they, though?
In my experience as a data science/machine learning/now โAIโ leader, what that investment looks like is merely headcount: Here, we are building a data science team, go hire 4 people. Thatโs the investment.
Which is obviously a type of investment; they could have hired a bunch of marketing/sales people for that money. But are you actually given a real budget to accomplish your departmental goals? Of course you get headcount; whoโs gonna do the actual work?!
There is a simple and quick test to see if your exec has what it takes. I have been asked a lot of times by execs to recommend a singular course of action to make the company more data-driven. I always provide the same suggestion:
Calculate the cost of doing nothing (not making data-driven decisions): find some budget numbers and do some Fermi-style back-of-the-envelope calculations to roughly quantify x% less churn, y% higher utilisation etc. Itemise this list.
Sum these efficiencies as an annual total (the benefit of making informed decisions). This is usually in the range of high six-digits to millions.
Take some percentage of the sum and transfer it into a pot.
Now, you have several options:
Say, the benefits sum up to $1M/year, and the pot is $50K/year (1/20). We offer this sum as a bonus to the employee(s) that demonstrate making data-driven decisions every year. Perhaps at every six-month reviews, so two $25K prizes to be won every year.
Or, establish a company-wide bonus structure based on making data-driven decisions. Add an area to everyoneโs performance reviews, and give 1-2% pay raise if they demonstrated high data-driven approach to their work.
Voila! The absolute best strategy to incentivise your employees to be more data-driven: appealing to their intrinsic motivations.
Iโm not saying everyone cares about more money (and some will try to game it a la Goodhartโs law), but it helps. It definitely helps more than the CEO engaging in cheap talk at company all-hands or people team writing a Notion doc that no one bothers reading.
But the execs canโt seem to commit do it. Itโs too out there. โWhy would we throw away money like that?โ one said to me. Mind you, these people are supposed to have vision. They rather spend another couple of million on inefficient outbound marketing to see a pitiful percentage increase (maybe, not even guaranteed) than walking the walk when it comes to being data-driven. But they cannot justify spending $50K/year on an initiative. When you think about a full-time tech hire costing more than thatโฆ
When it comes to AI strategy, now that all execs are hyper-fixated on this, ask your top exec the following:
What are your expectations re: the AI investment?
What percentage of our revenue are we investing into Data & AI?
Make sure to ask the questions in this order, so that they have ample time to dazzle you with their high expectations (outcomes, time frames, market share increase etc.).
Then ask the next question.
If they need help, say that the investment needs to match the expectations; you can deliver low investment <> low expectations, or high investment <> high expectations. They pick the left-hand side, and you deliver the right-hand side.
This talk can range from awkward to uncomfortable, but it will absolutely be eye-opening for you. I suggest committing this to paper once it has been discussed. When you write a strategy proposal, quote this text in the very beginning to set the stage. When talking to the wider company, mention it (โWe are investing this much in AI to doโฆโ).
The Art of Troop Leading
The first part is about removing the shroud on the โinvestmentโ. Doing so should arm you with the knowledge and the clarity on where you and your boss stand.
However, this will be unknown to the managers and ICs under you unless you explicitly choose to align them with what you know. It sounds obvious to note, but do we actually communicate the extent of the pressures on us to our reports?
Many middle managers donโt. Some think itโs their burden; others think it could be seen as a weakness. Thereโs also a segment that thinks line reports shouldnโt be privy to such matters; I suppose these middle managers feel โimportantโ and โmysteriousโ ๐
I suggest the following: make the pressures on you absolutely, obviously, explicitly crystal-clear to your team.
I usually โreport backโ to my team every month or so, using a template similar to below:
The CEO/board/investors expect [this] by [then]
Some CxO really needs [this] and donโt care about [xyz]
We donโt need to care about [other execs]
This [middle manager] also needs [this] done, and we should deliver it because itโs a win-win for us
Ignore the rest
Everything in this list important; not just things need doing but also things the team needs to ignore because doing those tasks wonโt move the needle. The finite resources equally applies hereโyour function/department/team can only do so much. Unless you tell them what/who to ignore, they will either be haphazardly pulled into various directions (opposite of strategic), or their tasks will be dictated by the loudest (Finance? Sales?).
It is your privilege to be in the middle, between the strategic and the tactical worlds. You are uniquely positioned to understand the company dynamics, politics (yes), and bureaucracy. A line report may not know how to gauge and prioritise doing [this] for Marketing vs. doing [that] for RevOps. You, on the other hand, absolutely should.
You need to be agile. Such lists rarely stay static. If your company employs OKRs (they probably should), you are looking at 3 objectives with 3-5 key results each in every quarter. Some of the objectives will be cross-functional, given the traditional role of Data functions as a support function (read: being a cost centre). In part 4, I will lay out a blueprint to turn data functions from cost centres to profit centres.