The AI Consultant's First Rule: Never Start with the Technology
Why Most AI Projects Are Doomed Before They Begin
If you’re the one expected to bring “the AI plan” to the next leadership meeting — but no one can tell you what the actual business goal is — this is for you.
Because right now, in boardrooms across the world, senior leaders are making one demand:
“We need AI in the business. Now.”
No context. No clarity. Just pressure.
It reminds me of that scene in the movie Margin Call.
Jeremy Irons plays the CEO. He looks at the PhD risk analyst and says:
“Speak to me as if I were a small child… or a golden retriever. It wasn’t brains that got me here.”
That line has aged beautifully. It captures the modern AI dilemma.
The people demanding AI at the top don’t know what they’re asking for.
The people in the middle are stuck translating hype into PowerPoint.
The engineers at the bottom are asking, “Wait, what problem are we solving?”
This is why most AI projects are failing.
They start backwards. They begin with tech — not truth.
Today, I’ll show you a mental model I use with clients to break out of this trap.
It's the first rule of trusted AI consultants.
Never start with the technology.
Start with the consequence.
The Pattern of Flawed AI Strategy
Here’s your typical flawed approach.
Leadership feels pressure — from competitors, the media, or investors — to “do something with AI.”
A vague AI initiative is greenlit. Maybe it’s a chatbot. Maybe it’s “AI-generated content.” Maybe it’s “AI to speed up development.”
The company either buys into hype (like calling human labour “AI”) or rushes to automate visible tasks with no clear ROI.
Results disappoint. Teams get frustrated. The initiative is quietly shelved.
If you look around, there’s plenty dead bodies floating to the surface this year. It’s happened everywhere.
Builder.ai claimed an AI could build apps — until it was revealed humans were doing all the work behind the scenes. They went bankrupt.
Duolingo’s CEO said they were becoming “AI-first” and replacing contractors — and watched their most loyal users revolt.
Chegg ignored generative AI until it lost half a million subscribers to ChatGPT. Now it’s in survival mode.
Each of these companies tried to “do AI” without asking what the real problem was that they were trying to solve. Plus, the data tells us that 42% of businesses have scrapped most of their AI initiatives this year. Only 1 in 4 have made it past the pilot phase. And for the executives who followed the hype without the strategy? A tipping point is approaching.
Why technology first doesn’t work
Because leaders confuse the presence of AI with the creation of value.
That’s it. In one sentence. It’s that simple.
And I know this is right because I’m both at the table and in the trenches.
Exec is trying to build from tech, not intent.
They mistake being early for being ready.
And most fatally, they forget that AI is not a product — it’s a tool. A force multiplier. A capability. Without a job to do, it just burns money and trust.
There are patterns to these failures:
AI as marketing gimmick → Audience backlash (SheerLuxe, Lego, BBC)
AI as investor bait → Lawsuits and reputational damage (Builder.ai)
AI as disruption defence → Too late to save the business (Chegg)
AI as cost-cutter → Users leave, quality drops (Duolingo)
They’re all treating AI like a silver bullet.
It isn’t.
In my day job running Templonix, I have to constantly reinforce this in client meetings. Agentics in particular - it’s just a new classification of software.
It’s not Harry Potter magic. It comes with its advantages and constraints, and you need to have your eyes open to those facts to get the best results.
A better way to get results
Me personally, I take a different approach.
Instead of asking “What can AI do?”, it’s better to ask where the people in the business spend most of their time and energy. What decisions get made repeatedly that could be improved with better recall, faster research, or clearer communication? Where does human work feel mechanical, repeatable, or mentally exhausting?
More importantly - How do we use tech to unleash human potential?
I think too many people think of AI as a headline, and not workflow optimiser. This is a big mistake.
You don’t replace a sales team with AI. You replace the hours spent researching and preparing for contract renewal negotiations so that the account manager has room to think about the right moves and use the machine to do all the heavy lifting with the background work.
Analysing quarterly reviews from CRM notes
Contract documents
Conducing sentiment analysis
Running web searches to do discovery on a customers current business circumstances
You shouldn't promise AI-first. It should be problem-first, and let AI earn its place in the solution.
That’s how you avoid hype traps and build something that survives past the prototype.
How to Build an AI Strategy That Won’t Explode on You Later
It’s only June and there’s already a lot of bodies in the AI graveyard. Want to stay above ground a bit longer? You do this by being disciplined.
Below is a simple 7-step checklist you can use to pressure-test any AI initiative before it becomes a costly PR crisis or a dead-end pilot.
But before you even get to the checklist, here’s a mechanism I use to very quickly classify opportunities in 2025 that deserve your attention.
✅ 1. Start with a real problem, not a trend
Ask: What are we trying to fix, speed up, or scale?
If your AI project doesn’t replace an existing painful workflow, save time, or improve decision-making — pause. Hype isn’t a use case.
Also, in my experience, the most value with AI is unlocked from scaling. If you can find repeatable processes within a workflow that releases human capital to focus on higher value tasks, this is the force multiplier you want from AI.
✅ 2. Map the workflow first
Before you build anything, sketch the manual version of the task:
Who does what?
What inputs are required?
What decisions get made?
With what data?
Where are the bottlenecks?
You can't automate what you don't understand.
✅ 3. Decide where AI fits — and where it doesn’t
AI should enhance human strengths and handle repetitive or pattern-heavy work, not blindly replace people. Identify the agent’s role: Is it summarising Drafting? Prioritising? Conducing research?
Also, make sure you give the agent a persona that fits with your use case and organisation. This vastly improves output quality.
✅ 4. Define success before you start
Choose one clear metric that proves the agent is delivering value.
This is the part that hype-driven AI strategy bypasses that results in failure.
Following through on establishing what you expect good to look like keeps your project honest. Even if it’s as simple as “Reduce the time to produce a forecast report by 50%”, you have something to aim at.
Plus, from there, you can not only measure success, but you then have the first part of the ROI equation - if you know that that the 50% equates to say, 5 hours, you can then apply the labour cost on the other side and all of a sudden you have a $ value cost benefit to put in from our your CFO.
✅ 5. Build a narrow pilot — and review it ruthlessly
Don’t overengineer form the beginning. AI projects, especially agents, require iteration and improvement over time.
Build a low-risk version first. The first version will be crap.
Also, let real users try it. Watch what breaks. Don’t be afraid to kill or pivot.
Most AI success comes from iteration, not vision.
✅ 6. Communicate clearly and early
People fear AI because of bad framing. Be honest about what it will and won’t do.
Show how it will help humans, not replace them.
Bad comms derailed Duolingo. Don’t repeat the mistake.
✅ 7. Assign a human owner
Every AI system needs a real person responsible for it. Someone who monitors performance, gathers feedback, and is on point for adjusting the workflow as needed.
Not fully understanding the amount of work the Human in the Loop has with an AI project is one of the main reason why project budgets burst the banks once the project is in-flight. Don’t treat this as an afterthought, this needs up-front consideration.
Despite what the hype-merchants say, AI is not “set and forget.”
What you can do today
If you’re in a business that’s currently fumbling toward “AI strategy,” here’s what action you can take to move you forward.
Start a list of the high-effort, low-value tasks being conducted. Things that you and your colleagues spend time on, but see little joy or return.
Then, think about how quickly you believe those tasks could be conducted by an AI agent. Be conservative and not too ambitious. Sandbag it a bit.
In a world chasing AI headlines, this will give you an edge, clarity and the foundations for ROI that could support a business case.
Overall, don’t build with AI. Build for something. Let AI prove it belongs.
Until the next one,
Chris
🧰 Whenever you're ready, I might be able to help you.
Ready to up-skill and transform your career in the AI economy?
Join our paid subscribers to unlock the complete AI Agent Business Translator Toolkit - 18+ frameworks, tools, and templates. 7-day free trial included. Visit the Toolkit page to find out what’s included.
Great advice here, thanks!!