Business First, AI Follows: Build Your AI Business Case First – and Value Follows
Ethan Mollick has posed an important question in his essay in The Economist (“The IT department – Where AI goes to die”):
How can the potential of AI be harnessed in companies?
In his answer, he rightly warns against treating AI like ordinary enterprise software and destroying its “weirdness” through de-weirding.
Instead, Mollick proposes an elegant sequence: Leadership, Crowd and Lab. His core idea is correct: leadership must step up, employees must be allowed to experiment, and a motivated team is needed to explore the possibilities of AI.
Still, a decisive step remains under-emphasized in Mollick’s thinking. He calls on leadership to develop “a vision of how AI changes the organisation.” That is the language of strategy consulting but not quite the hard language of business: concrete, measurable, rewarding business goals. Mollick’s formulation provides inspiration, but without translation it does not lead to action.
The old way of using IT technology in business
Companies are not creative playgrounds, but finely tuned, precise clockworks. Every process, every interface and every responsibility is so precisely balanced and aligned that the overall system works reliably and creates value, so that customers are satisfied and the company survives in the market. Introducing an unpredictable polymath like generative AI into this clockwork – a system that is emergent, sometimes hallucinatory and, moreover, controlled by external vendors – seems absurd at first glance. And at a second glance as well.
That is why the reflex of IT departments to “trim” AI, squeeze it into existing processes and minimise risks is completely healthy and sensible. The only problem is: this approach currently delivers too little real progress. Refusal is no strategy.
But the solution is not far away. It lies in the classic three IT questions that must be asked and answered BEFORE every technology introduction:
Where can the technology (the solution) really help us?
What exactly does it do there (the task)?
How do we bring it in reliably and scalably (the integration)?
These questions are not new. When ERP systems were introduced 25 years ago, they were exactly the same. ERP was introduced because material flows, orders, logistics and disposition needed better, faster and safer information. ERP could do that – reliably, faster and cheaper than before. Business benefited, the clockwork ran faster and could grow.
Why should it be any different today? Because AI is a completely different technology? Hm, let’s take a closer look.
The AI Use Case – AI Use-less Case Tragedy
AI vendors like to present spectacular use cases and demos. Many companies then start pilots – and watch them quietly die later. The reason: A use case without a motivating measurable business case is a “use-less” case. Even worse, it becomes a “lose-case” that drains resources without delivering ROI.
This is where “Question Zero” comes in — the question often hidden in the ERP era because the answer was obvious. Leadership had already stated the goal: “We want to turn a bigger wheel, so we need better information flow!” That was the implicit natural mandate for the IT department.
In the AI project this question is simply not asked – perhaps because AI can do so much? Maybe it can surprise us? Why restrain it? Perhaps it can do so much more for us. Mollick almost conjures up the creative power of AI. Let it unfold…
But how is that supposed to work? Without a business goal and clear specifications of its task and benefit, there is no case for an AI project.
A quick case study: The Bestseller
A chocolate manufacturer wants a new chocolate bestseller, Dubai-like. Normally this takes twelve months. But the competition does it much faster because they are smaller, more agile and have less legacy. So they ask: “Can we reduce our time-to-market from twelve months to four?”
Now that is a worthwhile business goal and a real challenge for the team. And for AI. Finally, AI has a job to do.
Business First, AI Follows
The decisive rule is: Business First, AI Follows. We decide what we want and then see how AI can help us achieve it. AI deployment is a business case first, and a use case only during implementation. Most projects today fail because they start with the second step.
AI as an Information Supplier
To find the right starting point, we must really get what AI actually does. AI is essentially an Information Supplier. It processes data using computing power and algorithms to produce actionable information. That’s it. Its output allows a human or a machine to make decisions and turn these into an action that eventually leads to a business outcome, like a bestseller.
Building a real AI Business Case
An AI Business Case is at its core a completely normal business case – independent of the technology.
The case starts with a clear, ambitious and SMART business goal, aligned to the existing business and organisation. Example: “Deliver a new bestseller in 4 instead of 12 months.”
Next: the “old” workflow steps must remain, or there is no way to achieve the new goal. Example: “Run the concept phase, design, prototyping, testing up to market launch, but faster, cheaper, and better with the help of AI.”
Last: Follow these three steps to bring AI into play and into the company in a controlled way.
Analyse the existing workflow leading to the business goal.
Identify the critical information, the bottleneck points for AI.
Test AI potential at THESE very points and introduce AI only where it actually and sustainably eliminates the bottleneck (and please without babysitting by the teams).
The advantages of this AI Business Case approach are evident:
Ambitious Business Goals: The “Why” is worth the effort.
Familiar Workflows: No risky experiments with the company’s core structure.
Expert Control: The workflow is managed by the people who understand the craft. And AI is forced to follow orders.
Reconcile the ages: Bridging the IT-AI Gap
So here are the answers to the three classic IT questions for AI.
Where to use AI? Only where information in the workflow is missing, too slow, or too expensive.
What is the task for AI? AI delivers better, faster, or cheaper information within the workflow.
How to integrate AI? The IT team and the workflow experts collaborate on a targeted project with clear KPIs.
Voilà. The technology is tamed, commissioned, and monitored. And the best part: The bestseller tastes like chocolate, not like an AI experiment.
The “Potential” Argument
Critics might ask: “Aren’t you limiting AI’s potential?” Well, no more than we limit the potential of any other high-performing employee. We hire experts to do a job. AI’s job is to deliver the information WE need, exactly where and when WE need it. Or stay outside.
And what happens next? Greater success in the market opens up new perspectives for a business. Once the ROI of a “tamed” AI is obvious, the company will naturally find ways to scale it. The next bestseller is closer than ever before.
Mollick’s sequence for AI usage in business — Leadership, Crowd, Lab — holds true, but the engine that drives it is the AI Business Case. If you want success, build an AI Business Case – and join those who get from AI exactly what they want, or even more than what they could have imagined.
Build Your AI Business Case First – and Value Follows.
About the author:
Silvio Gerlach is an an AI Integration Expert. His latest book "How Business with AI Enlarges the Pie: AI Business Integration in 6 Steps" tells the story of a chocolate company turning AI promise into real business value – a real bestseller.