Practical Deployment Lessons When Scaling AI At The Edge

AI at the edge is often portrayed as a future ambition reserved for organisations with deep technical expertise and significant budgets. According to Neil Cresswell, Co-Founder and CEO of Portainer, that picture is increasingly outdated. Advances in containerisation, edge computing and AI tooling are making it easier to deploy, manage and scale intelligent applications across distributed IoT environments.

What is AI at the Edge?

AI at the edge refers to running AI models and decision-making processes close to where data is generated, rather than sending all data to a central cloud platform for analysis. As edge hardware becomes more capable, devices can now run applications, store and synthesise data, and perform AI inference locally.

This approach is particularly valuable when:

• Decisions need to be made quickly
• Connectivity is unreliable or intermittent
• Large volumes of data would be impractical to send to the cloud
• Local action is required in near real time

As Neil explains, edge intelligence enables organisations to make fast operational decisions without relying on a cloud round trip.

Why Containers Matter in IoT and Edge Computing

One of the central themes of the conversation is containerisation and its role in simplifying software deployment across large fleets of connected devices.

What is a container?

Containerisation packages an application together with the files and configuration it needs to run. This creates a more consistent environment between development, testing and production, reducing the risk of unexpected differences when software is deployed in the field.

For software vendors and organisations managing IoT deployments, this creates significant operational advantages:

• Applications can be tested once and deployed consistently
• The same image can be distributed across large device fleets
• Updates can be delivered remotely
• Software versions can be governed more effectively
• Devices can maintain consistency regardless of location

The challenge of scale

Managing a handful of devices is relatively straightforward. Managing hundreds, thousands or even hundreds of thousands of devices is a different challenge altogether.

Some devices may be connected continuously. Others may only come online periodically through cellular or satellite links. Some may have missed multiple update cycles. Keeping software versions consistent across such an environment requires a dedicated control layer.

Portainer’s role is to help organisations distribute, run, update and govern containerised applications across these distributed environments, reducing the need for specialist Docker, Kubernetes, networking or storage expertise.

As Neil puts it, the goal is to answer a simple question:

‘How do we take skills and experience out of the equation?’

Why AI Is Moving Closer to the Source of Data

Traditional IoT deployments often focused on collecting information and sending it elsewhere for analysis. Today’s edge environments increasingly support local processing and decision-making.

The reason is practical rather than theoretical.

In many use cases, the value lies not in collecting data but in acting on it quickly. Waiting for cloud processing can introduce latency that makes the outcome less useful. Edge AI enables organisations to analyse information where it is created and respond immediately.

This shift also reduces the need to backhaul large quantities of data, such as video streams, while improving responsiveness in environments where connectivity may be limited.

Real-World Edge AI Use Cases

The episode highlights several examples that demonstrate why edge intelligence is becoming increasingly valuable.

Agriculture: AI on the Tractor

One example involves agricultural applications where computer vision systems identify weeds while machinery is moving through fields.

The challenge is straightforward: a tractor cannot wait for a cloud response after it has already passed the target. Data must be analysed and acted upon quickly enough for the equipment to respond in real time.

This is precisely the type of scenario where local AI inference provides clear operational value.

Food Processing and Safety

Another example involves detecting unexploded ordnance before it enters a potato-processing production line. While unusual, the use case illustrates an important principle: edge AI can identify hazards and trigger action before they create operational or safety risks.

In both examples, success depends on making fast local decisions tied to a specific operational outcome.

The Biggest Mistake in AI and IoT Projects

While much of the conversation focuses on technology, Neil repeatedly returns to project execution.

His central argument is that organisations often struggle because they attempt to tackle too much at once. Rather than focusing on a specific business outcome, they try to transform entire operations simultaneously.

Don’t boil the ocean

The projects that succeed tend to start with a single problem.

That problem might involve:

• Reducing production waste
• Detecting process abnormalities
• Preventing equipment damage
• Solving a clearly measurable operational issue

Achieving one successful outcome creates organisational trust, demonstrates value and provides a foundation for broader deployment.

As Neil explains:

‘The ones that I see successful are trying to solve just one problem.’

Has the Build-vs-Buy Equation Changed?

Another notable theme is the falling cost of experimentation.

Capabilities that once required significant investment can now be explored using commodity hardware, affordable edge computers and lightweight AI models. This doesn’t eliminate the need for robust operational deployment, but it dramatically lowers the barrier to entry for proving whether an idea has value.

Neil advocates a practical approach:

1. Identify a focused use case.
2. Build a low-cost prototype.
3. Test the idea.
4. Measure the outcome.
5. Scale only when the evidence supports further investment.

This mindset reduces risk while helping organisations learn quickly. Not every experiment will succeed, but failures become affordable lessons rather than expensive mistakes.

Why Innovation Is Becoming a Competitive Requirement

Beyond technology and project management, Neil argues that organisations must actively create room for experimentation.

As the cost of testing new ideas falls, more businesses could explore innovations that were previously out of reach. The companies that encourage experimentation are more likely to discover operational improvements and competitive advantages. Those that do not may find themselves overtaken by competitors willing to test and learn.

His advice is refreshingly direct:

‘My advice is just do it.’

Key Takeaways

• Containerisation provides a repeatable and scalable way to deploy software across distributed IoT environments.
• Managing IoT at scale is primarily an operational challenge requiring governance, version control and remote lifecycle management.
• Edge AI enables faster decision-making by processing data close to where it is generated.
• Successful AI and IoT projects start with a single measurable business problem rather than attempting broad transformation from day one.
• Affordable hardware and AI tools have significantly reduced the cost of experimentation and proof-of-concept development.
• Organisations that create space for experimentation will be better positioned to innovate and compete.

FAQ

What is edge AI?

Edge AI is the practice of running AI models and inference close to where data is generated, allowing devices to make decisions locally without always relying on cloud processing.

Why are containers useful for IoT deployments?

Containers package applications with the files and configuration needed to run consistently across environments. This simplifies deployment, updates and governance across large fleets of connected devices.

Why do many IoT and AI projects fail to scale?

According to Neil Cresswell, projects often struggle when organisations attempt to solve too many problems at once. Starting with one measurable use case creates clearer outcomes and a stronger path to scaling.

About the Guest

Neil Cresswell, Co-founder and CEO of Portainer, brings nearly 30 years of IT experience and works closely with organisations deploying containerised applications across edge and IoT environments. In his episode of the IoT & AI Leaders Podcast, he shares lessons from a wide range of real-world deployments and explains how businesses can turn AI and edge computing from experimentation into operational value.