Do I Need AI in My Product?

AI is everywhere. It’s in our phones, our cars, our homes, and increasingly in conversations about the future of connected devices. For many product teams, it now feels like a question of when to adopt AI, not if. But in reality, not every product needs it. The real question isn’t “How do we add AI?” It’s “Does AI actually solve a real problem?”

At Consult Red, we help product and engineering teams make that distinction early. The goal is not to chase a trend; it’s to make decisions that create long-term business value. AI can improve performance and open the door to new capabilities, but only when it addresses a clearly defined need.

Just as we discussed in The Hidden Costs of Choosing the Wrong Architecture, the smartest technology choices start with asking the right questions early. The same applies to AI.

What “AI in Products” Really Means

You could argue that the phrase “AI-powered” has become so overused that it’s lost some meaning. AI isn’t a single technology, but a collection of methods that help systems perceive, predict, and adapt. In the context of connected products, this can mean:

• Machine Learning: Systems that learn from data, improving things like energy optimisation or user-behaviour prediction.
• Computer Vision: Detecting objects, defects, gestures, or patterns.
• Predictive Algorithms: Forecasting wear, maintenance needs, or supply levels.
• Anomaly Detection: Identifying faults or irregularities in real time, often used in manufacturing and industrial IoT.
• Natural Language Processing: Voice control, chat interfaces, or smart assistants.
• Personalisation and Adaptive Experiences: tailored user interfaces, recommendations and product behaviour based on knowledge of the user.

In many industrial and regulated environments, AI plays a supporting role rather than making autonomous decisions. It is often used to surface insights, highlight risks, or optimise parameters, while humans validate and remain in control. This hybrid approach allows teams to benefit from AI without compromising safety, compliance, or predictability.

Most products use AI to enhance what already exists – helping systems respond faster, understand context, or support better decision making. And, as we explored in Build Smarter Without Starting from Scratch, evolution doesn’t always require rebuilding from the ground up. It’s often about extending the value of what you already have.

Start with the Problem, Not the Technology

Before committing to any AI initiative, it’s worth asking: what specific outcome are we trying to improve?

At Consult Red, we guide clients through a few simple but important questions, including:

• What business or user problem are we solving? (e.g. uptime, cost, usability, differentiation)
• Could automation, sensing, or simpler analytics solve it first?
• What measurable improvement would AI actually deliver? (speed, accuracy, cost, convenience or genuinely useful new functionality)
• Do we have the right data and infrastructure to support, train, and maintain an AI model over time?

These questions sound simple, but they reveal whether AI is truly needed or whether it’s a solution looking for a problem. Many teams fall into the trap of starting with the technology itself, as we highlighted in Key Mistakes Businesses Make When Evolving Their IoT Device.

As Tomasz Bodowski, VP of Industrial at Consult Red, puts it:

“In industrial environments, AI decisions often operate close to physical processes. This introduces constraints that don’t exist in purely digital products: deterministic behaviour, real-time response, offline operation, and coexistence with PLCs, SCADA, and safety systems. Any AI concept must be evaluated not only for accuracy, but also for predictability, latency, and operational risk in an OT context.”

When AI Adds Real Value

AI delivers genuine impact when applied to complex, data-rich problems where automation or prediction directly improves business performance. The strongest use cases start with a clear business outcome, not a technology ambition.

Manufacturing

AI has proven particularly effective in automating quality control. Computer vision systems can now identify issues on production lines faster and more accurately than ever before. In our Edge AI webinar, we demonstrated a visual anomaly detection solution that identified faults in pharmaceutical packaging in real time, without relying on the cloud. The result: faster feedback, higher throughput, and less waste.

Industrial IoT

Predictive maintenance is one of the clearest examples of AI creating commercial value. By analysing sensor data, AI can predict when equipment is likely to fail, and prompt intervention before disruption occurs. That reduces unplanned downtime and helps teams schedule maintenance more efficiently. The same approach can support environmental monitoring, logistics tracking, or energy management, enabling smarter decisions with continuous data.

In industrial environments, innovation rarely happens on a blank canvas. Most value comes from retrofitting intelligence onto legacy infrastructure. This might mean using edge gateways to capture vibration data from 20-year-old motors or adding visual inspection layers to existing assembly lines. In many cases, AI extends the life and usefulness of equipment rather than replacing it outright.

A practical example is our Smart Demand Vehicle Charging Solution. Working with a client, we helped design a system that uses AI to balance energy load across a fleet of chargers. This made the solution more efficient, more reliable, and easier to scale commercially.

Media and Broadband

In digital media and connected home services, AI enables smarter recommendations, contextual search, and adaptive user experiences. Operators also use it to detect quality issues, optimise bandwidth, and predict churn, improving satisfaction while reducing support costs. If you’re exploring how AI is reshaping this space, our Media AI page provides more examples and practical guidance.

Across these industries, the benefits may look slightly different, but the value is the same: cost reduction, operational efficiency, and better user experiences.

When AI Doesn’t Make Sense

AI is powerful, but not always the right answer. Some of the most effective product decisions come from knowing when simpler, rule-based solutions will deliver the same value with less complexity. Being selective about where you apply AI is what separates successful projects from expensive experiments.

There are situations where AI adds more overhead than impact:

• When automation or fixed logic can solve the problem reliably.
• When there isn’t enough clean data to train or maintain models.
• In safety-critical systems, where predictability matters more than adaptability.
• When small production volumes make the economics unviable.
• Resource-constrained environments (e.g. battery-powered wireless sensors) where AI solutions are unfeasible given the constraints on power and compute resources.

In industrial and safety-critical environments, it is crucial to distinguish between control and optimisation. Deterministic, rule-based logic should always handle the safety stops. AI is typically best used as a ‘co-pilot’, helping operators optimise or predict faults, without directly taking over critical actuation paths where determinism is non-negotiable.

In regulated industrial domains, AI must coexist with functional safety standards (e.g., IEC 61508, ISO 13849). Black-box models that cannot be validated, bounded, or explained may be unsuitable for control paths, even if they perform well statistically.

Finally, AI doesn’t compensate for poor mechanical design, unstable processes, or missing instrumentation. If the process itself is not under control, AI will amplify variability rather than fix it.

Understanding these boundaries is a strength. It allows product teams to focus resources on the areas where AI creates measurable impact, rather than adding intelligence simply because the market expects it.

Laying the Groundwork: AI Readiness and Product Evolution

Before any model is trained or integrated, a product needs the right foundation. Adding AI is not simply a software upgrade; it touches data, hardware, security, and lifecycle planning. Product teams should start by assessing:

• Data quality and governance: Is the data structured, reliable, and compliant with standards like EN 18031 and GDPR?
• Hardware and scalability: Can the device support inference, connectivity, and model updates?
• Lifecycle and sustainability: Models drift over time; do we have processes for retraining and secure updates?
• Cross-functional collaboration: Hardware, firmware, and data teams must align from the start.

This doesn’t always mean starting from scratch. With the right architecture and platform in place, AI can often be added to existing products through firmware updates, new edge modules, or smarter data pipelines. As we covered in Build Smarter Without Starting from Scratch, evolution beats reinvention, and AI can be part of that same journey.

For industrial teams, Explainability (XAI) is often as important as accuracy. If an AI model rejects a batch of products, the operator needs to know why. Was it a colour deviation or a dimensional error? Building systems that provide insights, not just binary decisions, is key to adoption on the factory floor.

How Consult Red Helps, and Why Smarter Products Start with Smarter Questions

For companies exploring AI, the first challenge is deciding where to focus. There are many possibilities, but not all create value. Consult Red helps teams cut through the noise and prioritise what’s technically achievable and, most importantly, commercially sound.

We approach AI adoption with our customers in four connected stages:

1. Strategic assessment: Identify where AI adds measurable ROI and clarify the problem the product is trying to solve.
2. Proof of concept: Validate ideas with rapid prototypes before committing to full-scale development.
3. Integration: Combine embedded engineering, connectivity, cloud, and data expertise to deliver reliable AI integration.
4. Lifecycle support: Maintain performance, security, and compliance across the product’s lifespan.

Because we work across the full stack, from chip to cloud, we understand both the technical realities and commercial pressures that shape product evolution and AI adoption. Whether the product is an industrial controller, a consumer device, or a media platform, our goal is the same: help clients introduce intelligence in a way that strengthens their offering and supports long-term strategy.

From experience, we know that the most successful AI projects don’t start with algorithms; they start with the right questions. AI can be transformative, but only when it serves a clear purpose and delivers value that matters to both the business and the user.