Much of the buzz about AI is focused either on the models themselves or on how existing companies are adding AI to their products.

But there is another important piece here that I don’t think most people are talking about: using AI and other recent innovations, new and flexible SaaS startups can out-compete the old guard.

Most existing software products are monolithic, not always in service architecture, but in the product itself.

You can choose a specific tier of usage and features, but the SaaS is the SaaS, and customization is either a few toggles on a settings page or a multi-million dollar endeavour, and there’s usually no in-between.

I recently met with a startup that's building a new product, in a market where they are definitely not the first. But that doesn’t mean they don’t have a competitive advantage.

Competing products were built the old way, designed for one assumed happy path, with little to no option to deviate.

By building their product to be flexible from the start, that startup will be able to offer potential customers a tailored solution that their competitors can't match.

That's because adding flexibility to existing software is rarely feasible. Those existing products weren’t designed to allow it, and their current customers and processes often stand in the way of that level of change.

So how do you create that flexibility? And just as importantly, how do you make sure that you are building for flexibility without over-engineering?

Here are some ways:

1. Build your application and data structure to enable flexibility in inputs and outputs, with users able to provide instructions through both natural language prompts and more traditional interfaces.

   The results they get should be available in multiple formats, such as paragraphs, point form, or JSON. It isn’t essential that all of this flexibility is available from the start, but that your current offering doesn’t block that future flexibility in inputs and outputs.

2. Use infrastructure-as-code and CI/CD pipelines that expect and allow for variation in product offerings, tailored to customer needs.

   Multi-tenant isn’t dead, but will likely be on its way out as customers start expecting more control over their data and applications.

3. When building, keep it simple.

   Avoid anti-patterns that restrict flexibility, such as Big Ball of Mud (too many interconnections and interdependencies) and Not Invented Here Syndrome (always re-inventing the wheel).

   Software is good when it is as simple as it can possibly be, while still fulfilling its function.

For software engineers that are building or learning to build, it’s a good idea to gain a solid an understanding of managed services, infrastructure-as-code, and prompt engineering with schemas and other enhancements.

I may be over-indexing on flexibility, but I am sure these will be a big part of what’s next.

This article was originally posted on Regan's LinkedIn.