Drucker's Five Questions in the Age of AI: A Test Most Organizations Are Failing

I received an email yesterday that took me back 30 years to when I did Management Science at university and specifically Peter Drucker’s five questions from a marketing strategy standpoint. What struck me is that the questions are global within themselves and can be used for any “strategy”. Hence, why I’ve applied them to AI, because of course that’s the buzz these days, if you haven’t heard!

So, for many organisations out there, I reckon your AI strategy is probably answering the wrong questions. Worse, you might not realise it until you’ve spent millions proving it. That’s the problem this article tackles, and how we can make sure these horror stories don’t support the job loss of the CDO or other people who are supposedly responsible for data, analytics and AI.

Right now boardrooms are awash with AI ambition. This ambition is more about platforms being procured, transformation programmes launched, and consultants engaged. The activity is feverish, the conviction is high and high-fives all around. But if you ask a simple question such as “which strategic problems are we actually solving?” the hoo-hah, high-fives and the supposed clarity evaporates.

Fifty years ago, Peter Drucker offered five questions to test whether an organisation truly understood its own strategy, and they had nothing to do with IT. They were about purpose, customers, value, results, and direction.

Most organisations failed them then and if I put my money on it, most are failing them now.

The difference today is cost and velocity. AI doesn’t just let you do the wrong things faster, it lets you do them at industrial scale, with the illusion of precision, until the damage is irreversible.

So, here are the five questions that might trigger a response in doing the right things and doing things right or just continuing to do the same thing over and over expecting a different result.

1. What Is Our Mission? (and is AI serving it or replacing it?)

Here are some examples of what what typically happens. A bank whose mission is to “support customers through life’s financial moments” launches an AI chatbot that deflects calls. A retailer committed to “creating exceptional customer experiences” deploys personalisation algorithms designed purely to maximise basket size. A public sector body dedicated to “serving citizens” automates processes that make services harder to navigate.

Oh of course the technology works, but the mission atrophies. AI has a super seductive quality and it feels like progress. Only this morning I was reading someone’s comment which stated “well the models and dashboards all work”. Of course ask anyone and they will tell you the dashboards update in real time, models are 99% accurate and efficiency improves. But the question that I keep asking is “efficiency towards what?”

Most teams and boards cannot answer that question with specificity. The latter approve AI programmes based on capability, not contribution. The result isn’t strategy, it’s silent drift, which is normally dressed up as innovation.

As I’ve always stated, the test itself is brutal in its simplicity. Question: if you stopped every AI initiative tomorrow, which strategic outcomes would you fail to deliver?

What is your answer? If the answer is “none,” you are not investing in AI strategy. You’re funding a very expensive habit!

2. Who Is Our Customer? (And Why Are You Still Guessing?)

I have this conversation over and over with organisations, because I see drift here too. Building stuff for the sake of it. In my canvas, I’ve even got a section which explicitly asks the questions about who the customer is and what their needs are. I didn’t put it in there because I think it looks nice! No I added it, because every single experience I’ve had with organisations gets stuck in this very area.

Organisations now have more data about their customers than any organisation in history. The challenge that I see is that most also probably know less about them than they think.

Let’s have a look at the evidence, because believe me it’s everywhere. Marketing runs campaigns based on demographic segments that were outdated five years ago. Product teams build features customers don’t want because the feedback loop is broken. Customer service handles the same complaints repeatedly because insight never reaches decision-makers.

What does this tell us? The fact that the data exists, but the understanding doesn’t.

Do you think AI has created this problem? Nope, I don’t think it has, all it’s done is expose it. Because when you try to train a model on fragmented, contradictory, or incomplete customer data, the model fails. But when the model fails the truth is, your organisation doesn’t have a unified view of who your customers actually are.

Some organisations respond by fixing the data. Most respond by lowering their ambitions for AI. That is a choice, but it’s not a strategy.

The question isn’t whether you have customer data. It’s whether you have integrated it, governed it, and built decision-making processes that can actually use it. If you haven’t, AI will only give you faster access to the wrong conclusions.

3. What Does the Customer Value? (Because You’re Definitely Optimising the Wrong Things)

This is where AI strategies fail the most and makes a massive hole in the budget! That’s a nice way of putting it Samir! Here are some examples:

• An insurer uses AI to speed up claims processing. Customers wanted empathy and transparency, not speed. Churn increases.

• A telco optimises call centre routing to reduce handle time. Customers valued resolution on first contact. Satisfaction drops.

• A retailer personalises promotions to maximise margin per transaction. Customers valued discovery and trust. Lifetime value erodes.

In every case, the AI worked perfectly, but the strategy was wrong.

Here’s the pattern I see. Organisations optimise what’s easy to measure for example, cost per transaction, time to resolution, conversion rate, because those metrics fit neatly into dashboards and business cases. What customers actually value is trust, ease, relevance, respect for their time and guess what these are harder to quantify, so it gets deprioritised.

AI doesn’t fix this. It makes it worse. Because once you’ve automated the wrong objective, reversing it is politically and technically painful.

The brutal question for any board is: can you name three things your customers genuinely value that you are not currently optimising for?

If you can’t, your AI strategy isn’t customer-centric. It’s operationally convenient for you!

4. What Are Our Results? (And Why Are Your Dashboards Lying to You?)

AI has made it easier more than ever to measure things. It has also made it easier than ever to measure the wrong things and doing that with confidence.

Walk into most boardroom updates and you will find dashboards being presented that track everything, such as model accuracy, processing speed, automation rates, cost savings, user adoption. These numbers are very precise and guess what, the graphs trend upwards. Everyone nods and high-fives all over again!

But, no one asks whether any of it matters.

Because here’s what those dashboards typically don’t show: whether revenue quality improved, whether customers stayed longer, whether the organisation can now do things competitors cannot, whether strategic position strengthened.

Efficiency metrics dominate because they’re easy to capture and politically safe to report. Strategic outcomes are harder to define and often reveal uncomfortable truths, so they’re quietly omitted.

The result you are measuring activity, calling it results, and wondering why competitors who spend less on AI are still outpacing you.

Drucker’s question remains the only one that matters: What are your results?

Not your outputs. Not your activity levels. Not your cost reductions.

Your results, your outcomes. I created a very powerful measurement framework from strategic objectives to results with feedback loops, which was done pre “GenAI” and I reckon it will still hold strong and true today. We must look through the lends of competitive position, customer value, and strategic advantage.

If your “AI reporting” cannot answer this question clearly, you are not measuring performance, you are simply measuring noise.

5. What Is Our Plan? (Or Are You Just Calling Hope a Roadmap?)

Ask most organisations for their AI plan and you’ll receive a document. It will have phases, timelines, budgets, and use cases. It will look credible.

It will also probably be worthless.

Because what’s typically labelled as a “plan” is actually a collection of initiatives that someone thought sounded promising, arranged in rough chronological order, with dependencies ignored and trade-offs unacknowledged.

A real plan does something fundamentally different. It makes choices.

It says: we will do this, which means we will not do that. We will prioritise growth over optimisation. We will build capability in these areas and accept constraint in others. We will measure success by strategic impact, not project completion.

Most AI roadmaps I’ve seen, don’t do this. They try to do everything, which means they deliver nothing that matters.

Here is the test that I would apply: show your AI roadmap to someone with no context and ask them to identify your strategic intent. If they can’t, you don’t have a plan. You have a wishlist with Gantt charts.

But, from experience, most boards approve these plans anyway. Because rejecting them would require difficult conversations about what the organisation is actually trying to achieve, and whether AI is the right way to get there.

Those are exactly the conversations you should be having.

Why Drucker Matters Now

These five questions were never about technology. They were about something more fundamental: whether an organisation actually knows what it’s doing.

Drucker understood that most don’t.

What’s changed in the age of AI is not the challenge, it’s the cost of getting it wrong.

Poor strategy used to fail slowly. You had time to notice, correct, adjust. AI compresses that timeline. It allows you to implement bad decisions at scale, with conviction, before anyone realises the strategy was flawed.

By the time the metrics start declining, you’ve often automated the problem into every process, embedded it in every customer interaction, and trained your people to defend it.

That’s not transformation, it’s industrial strength failure.

The organisations that succeed with AI aren’t the ones with the most sophisticated technology. They are the ones that can answer Drucker’s five questions clearly, and then align their AI investments accordingly.

Everyone else is just spending faster.

A Final Question for the Board

So, here’s what happens next.

You approve another AI programme. It gets added to the portfolio. Someone builds a business case showing a 3-year payback. The steering committee signs off. Work begins.

Six months in, you realise the problem it was meant to solve isn’t actually a priority. Twelve months in, the technology works but nobody’s using it. Eighteen months in, it’s quietly shelved. The business case gets filed. Everyone moves on. No questions, because, another shiny toy emerges!

This pattern repeats itself across hundreds of organisations, thousands of initiatives, billions in wasted capital.

The reason is always the same: nobody asked Drucker’s questions before the money was committed.

Nobody asked whether the initiative served the mission, addressed real customer needs, optimised for things that actually mattered, measured strategic results, or fit within a coherent plan.

Instead, boards asked whether the technology was proven, whether the vendor was credible, whether the budget was reasonable, and whether the risks were manageable.

Those are procurement questions. Not strategy questions.

So here’s the choice.

You can continue approving AI programmes the way you’ve been doing it, on the basis of capability, vendor reputation, and comparative benchmarking. You’ll get exactly what that process delivers: a portfolio of disconnected initiatives that look defensible in isolation and achieve nothing in aggregate.

Or you can do something most boards won’t.

Require every AI proposal to answer Drucker’s five questions explicitly. Not in the appendix and not as an afterthought. As the first and primary test of whether the initiative deserves funding.

If the team presenting can’t draw a direct line from their AI programme to mission, customer understanding, customer value, strategic results, and an explicit trade-off against other priorities, send them back.

Because if they can’t answer those questions before you spend the money, they definitely won’t answer them after.

Can your organisation pass Drucker’s test?

If not, it means your current AI portfolio probably shouldn’t exist.