April 4th, 2026 at 10:43 am
By Ronak Shah, Founder at Nordstone
A few years ago, most founders I spoke to were asking a simple question:
“Do we need an app?”
Today, that question has evolved into something far more complex:
“Where does AI actually make sense in our product?”
And this is where many businesses get it wrong.
AI has become the most overused — and misunderstood — concept in product development. Founders feel pressure to “add AI” because competitors are doing it, investors are asking about it, and the market expects it.
But after working closely with startups and scaling products at Nordstone, I’ve seen a consistent pattern:
The companies that win with AI are not the ones who adopt it fastest —
they are the ones who apply it with clarity.
This article is not about hype. It’s about how founders should practically evaluate AI opportunities — based on real-world experience, product outcomes, and what actually works.
Start With the Right Question: Where Does AI Create Real Value?
The first mistake I see founders make is starting with the technology instead of the problem.
They ask:
- “Can we use AI here?”
- “Should we integrate a model?”
Instead, the question should be:
“Where in our product is there friction, inefficiency, or untapped data?”
AI is most effective when it solves one of three problems:
- Reducing manual effort
- Improving decision-making
- Personalising user experience at scale
If your use case doesn’t clearly fall into one of these categories, AI is unlikely to deliver meaningful ROI.
Identifying High-Impact AI Use Cases
In my experience, strong AI opportunities usually exist in specific layers of a product.
1. Repetitive Decision Systems
If your product involves repeated decisions — whether from users or internal teams — AI can often automate or optimise them.
Examples include:
- Recommendation engines in eCommerce
- Fraud detection in fintech
- Content ranking in media apps
These are high-impact areas because even small improvements scale quickly.
2. Behaviour-Driven Personalisation
If your product collects user behaviour data, there is an opportunity to personalise experiences.
This could mean:
- dynamic content
- tailored onboarding
- intelligent notifications
We’ve seen this drive significant improvements in engagement and retention when implemented correctly.
3. Data-Rich but Insight-Poor Systems
Many businesses already have data but don’t fully utilise it.
AI can unlock value by:
- Identifying patterns
- Predicting outcomes
- Generating actionable insights
This is especially relevant in industries like healthcare, finance, and logistics.
My Rule of Thumb
When evaluating a use case, I often ask:
“If we removed this feature tomorrow, would users notice?”
If the answer is no, it’s not worth building — AI or not.
Data Readiness: The Real Bottleneck
One of the biggest misconceptions about AI is that models are the hard part.
They’re not.
Data is the hard part.
Before considering any AI implementation, founders need to assess data readiness across three dimensions.
1. Data Availability
Do you actually have the data needed to train or power the system?
Many early-stage products assume they do — but the data is often:
- Incomplete
- Inconsistent
- Unstructured
2. Data Quality
Even if data exists, it must be reliable.
Poor-quality data leads to:
- Inaccurate predictions
- Irrelevant recommendations
- Broken user experiences
3. Data Flow
AI systems are not static.
They require continuous data pipelines to:
- Update models
- Improve predictions
- Adapt to user behaviour
Without this, AI systems degrade over time.
A Practical Insight
At Nordstone, we often delay AI implementation not because it’s difficult — but because the data layer is not ready.
Fixing data first leads to significantly better outcomes later.
ROI Analysis: Is AI Worth It?
AI is not cheap — not in terms of cost, time, or complexity.
This makes ROI evaluation critical.
Where AI Delivers Strong ROI
AI tends to perform best when:
- The use case affects a large number of users
- Small improvements create measurable impact
- The system operates continuously
For example:
- A 5% improvement in recommendations can significantly increase revenue
- A small reduction in churn can improve lifetime value dramatically
Where AI Struggles
AI is often not worth it when:
- The use case is too niche
- There is limited data
- The feature is rarely used
In these cases, a rule-based system may perform just as well at a fraction of the cost.
Think in Multipliers
AI works best as a multiplier.
If your core product experience is already strong, AI can amplify it.
If the product is weak, AI will not fix it — it will expose the weaknesses faster.
MVP Strategy: Start Smaller Than You Think
One of the biggest mistakes founders make is overbuilding AI systems too early.
They try to create:
- complex models
- fully automated systems
- large-scale infrastructure
This increases risk significantly.
What an AI MVP Should Look Like
A strong AI MVP is:
- Simple
- Measurable
- Focused on one use case
For example:
- A basic recommendation engine
- A simple prediction model
- a rule-based system with learning capability
The goal is not perfection.
The goal is learning.
Why This Matters
AI systems improve over time.
Starting small allows you to:
- Validate assumptions
- Gather real user data
- Refine the model
This approach aligns closely with how we guide clients through early-stage AI adoption at Nordstone.
When to Scale AI (And When Not To)
Scaling AI introduces new challenges:
- Infrastructure complexity
- Increased costs
- Model maintenance
- Data pipeline management
Many teams attempt to scale too early.
In reality, you should only scale when:
- The MVP proves value
- Data pipelines are stable
- The business impact is clear
Otherwise, you risk building expensive systems without clear returns.
Lessons From Working With Founders
Over the years, working with founders across industries, a few patterns have become clear.
The Best Founders Focus on Problems, Not Technology
They don’t chase trends.
They identify real user problems and apply the right solutions — AI or otherwise.
Clarity Beats Complexity
Simple AI systems that solve real problems outperform complex systems that try to do too much.
Speed Matters
Launching early and learning quickly is more valuable than building the “perfect” AI solution.
Data Is the Long-Term Advantage
Companies that invest in data infrastructure early gain a significant competitive edge over time.
AI is one of the most powerful tools available to founders today.
But it is also one of the easiest to misuse.
The goal is not to “add AI” to your product.
The goal is to:
- Solve meaningful problems
- Improve user experience
- Drive measurable outcomes
When evaluated correctly, AI becomes a strategic advantage.
When applied without clarity, it becomes a distraction.
From my experience, the difference always comes down to one thing:
How well founders understand where AI truly fits in their product.
FAQs
How should founders start evaluating AI opportunities?
Start by identifying areas where AI can reduce effort, improve decisions, or personalise experiences. Avoid starting with the technology itself.
Do all startups need AI?
No. AI is only valuable when it solves a real problem or improves a key metric. Many successful products operate without it.
What is the biggest challenge in AI adoption?
Data readiness. Without structured and high-quality data, AI systems cannot deliver reliable results.
How do you measure AI ROI?
Focus on metrics such as conversion rate, retention, efficiency improvements, and revenue impact rather than vanity metrics.
What should an AI MVP include?
A simple, focused solution that tests one use case and provides measurable outcomes.
When should startups scale AI systems?
Only after validating the MVP, ensuring data pipelines are stable, and confirming clear business impact.
