You’ve built something useful with ChatGPT. It works on your computer. But how do you make it available to your staff and customers, securely and without spending a fortune?
AI has made building software surprisingly accessible.
You don’t necessarily need to be a developer to turn a business idea into a working prototype. With the right prompts, some patience and a bit of trial and error, business owners are now building tools that would previously have required a development team.
And I think that’s fantastic.
But there’s something AI doesn’t always make quite so straightforward.
What happens when you’ve finished building it?
The application was built. Now what?
I recently worked with a client who had built his own quoting system using AI.
The idea was simple and clever.
Instead of having staff manually prepare every quote, customers could use the application to generate quotes themselves.
Less back-and-forth, less administration and a much easier experience for everyone.
The client had done the hard part of understanding the business problem and building something that could solve it.
But the application was running locally on his computer.
Now he needed to make it available to staff and customers.
That’s where things became a little more complicated.
How do customers access it? How do you protect the information? Where do you host it? How do you manage logins?
And perhaps most importantly, how do you do all that without turning a relatively simple business application into an expensive IT project?
Building an app and running a business application are two different things
When an application runs on your laptop, you’re generally working in an environment you control.
Once you put it on the internet, the situation changes.
You need to consider authentication, permissions, hosting, data protection, backups, updates and ongoing maintenance.
Customers shouldn’t be able to access each other’s information.
Employees should only have the permissions they need.
And the application needs to remain reliable and maintainable as the business grows.
None of this is particularly glamorous.
It’s certainly not the exciting part of building something with AI.
But it’s the difference between a clever experiment and an application you can confidently introduce into your business.
So what did we do?
In this client’s case, we didn’t need to start again.
We took the application he’d already built and worked through what was required to get it ready for deployment.
First, we moved the development project into GitHub.
That gave us a proper home for the source code, version history and a more manageable way to handle future changes.
Next, we set up a Virtual Private Server (VPS) to host the application.
Then we used Cloudflare as part of the security and internet-facing configuration.
For identity and access management, we incorporated Microsoft Entra ID.
The client was already using Microsoft 365, so there was no reason to introduce an entirely separate identity platform where the existing one could meet the requirements.
The idea was to use the right tools for the job, reuse what the business already had and avoid unnecessary complexity.
It is the same principle I keep coming back to: understand the business problem before choosing more technology.
Of course, securing an application involves more than choosing a hosting provider and enabling logins. The application itself, its permissions, data handling, updates and backups all need appropriate attention.
But those requirements don’t automatically mean you need an elaborate or expensive infrastructure.
And the hosting cost?
Here’s the interesting part.
The VPS costs approximately $20 per month.
That’s the server hosting cost, not the entire cost of developing, securing or maintaining the application.
But it illustrates an important point.
A custom business application doesn’t automatically require hundreds or thousands of dollars in monthly cloud infrastructure costs.
Depending on its requirements, a straightforward hosting environment can be perfectly suitable.
Some applications genuinely need more sophisticated infrastructure, high availability, managed databases or additional security services.
Others don’t.
The trick is understanding the difference.
And that’s where a little independent technical advice can save a business from unnecessary spending.
You don’t always need another expensive development project
One thing I find particularly exciting about AI-assisted development is that people who understand their business can now build tools to solve their own problems.
They don’t necessarily have to explain every detail to a software development company and hope the final product matches what they imagined.
They can experiment, build, test and improve their ideas themselves.
But when the application is ready to become part of everyday business operations, getting some experienced technical help can make a huge difference.
Not necessarily to rebuild the application.
Sometimes it’s simply to review what exists, identify what needs improvement, put the right foundations in place and help get it deployed.
That might involve a few technical changes, or it might uncover issues that need more substantial work.
Either way, the objective should be to build on what’s already there rather than automatically starting from scratch.
The bigger opportunity for New Zealand businesses
I suspect we’re going to see a lot more of these applications.
A sales manager building a quoting tool.
An operations manager creating a scheduling system.
A business owner developing a customer portal.
An office manager automating a process everyone has complained about for years.
These aren’t necessarily people who want to become software developers.
They’re people who understand their business and finally have tools that let them do something about its inefficiencies.
That’s a great thing.
But businesses also need to recognise when a useful experiment is becoming a system that staff and customers will depend on.
That’s the point where security, reliability and sensible technical decisions really matter.
AI has made building software easier. Getting it into production shouldn’t have to become the expensive part.
Sometimes you just need someone who knows how to connect the dots.