Software - build or buy?

You have heard about the prolific powers of AI-enabled software. Perhaps you have made some of it yourself. You know that this is a paradigm shift for enterprise technology, but what does this mean in the context of your business? Let’s establish what has really changed in recent months and explore what to do with that information.

The time and cost of producing prototypes, mockups, and a functional minimum viable product has come down considerably. We like to use Claude Code and Claude Design (almost) exclusively for our projects, and are finding that with highly precise, and sometimes even sparse prompting, we can create something close to the idea that we’ve been marinating on, in an afternoon. A smaller set of capabilities is now required both to execute a working system. Furthermore, the bland corporate AI slop tool that we can all instantly recognize can be routed. Claude won’t just make things that are homogeneous and mean regressing; when well instructed, it can be tasteful, articulate, even exciting in its design.

The actual code writing is continuing to get faster and better at breakneck pace. In our most recent system build, we noted the frontier version of Fable has meaningfully improved not just at just executing, but inferring needs on all points of system development.

In theory then, building custom enterprise systems should be much cheaper. We believe it can be. But these developments may also mean that the out-of-the-box software supply will expand, pushing costs downward, and increasing variety. That much remains to be seen, but it is clear that where technology problems exist at your small business, there are more and better options for solving them than ever before, both in terms of buying and building.

Buying is a sound default starting point for small business. However, the search should still commence with a clear, articulated understanding of the problem to be solved. Consider the phrase “minimum viable product.” Minimum viable product required to do what?

If you’re earnestly considering building your own software, odds are you’ve spotted the potential for significant payoff. We think that qualifying this payoff through some version of a written vision-statement is well worth your time. We’re fans of Ideo’s design thinking practices for brainstorming and vision-setting; consider this simple method to consider for your first go around with your team (or by yourself). The objective is to establish an anchor to help you and your team stay focused on customer impact as your discussions become increasingly technical.

The question always emerges early on: what is it worth in dollar terms to go the custom build route? Efforts to concretely or reliably quantify the value of your vision can be difficult. The cost of delay method is popular in the software world, but when you’re talking about projects with revenue implications rather than cost savings, Cost of Delay estimations become less reliable due to the inherent challenges with forecasting revenue from product or service changes. If you’re considering a cost-focused project however, we think this approach makes sense.

It is also useful to ground expectations in some common outcomes for build vs. buy vs. extend from an execution and maintenance standpoint (take a look at the comparison table). But here is the point that matters most: underestimation of cost and timing is endemic in most organizations that lack technological and project management discipline, regardless of whether you’re building or buying.

Custom builds demand more from your team in the short run, and demand a certain rigor of participation. Hands on internal testing and validation of the system is critical. Frequent and high-quality communication is needed with the vendor to make sure business context is deeply understood. Tension between scope and cost is inevitable and will need to be continuously renegotiated, especially when you don’t know from the onset exactly what you’re building. Even if you build the right system, implementing them can be hard, especially when substantial training is involved and the product is customer-facing. A significant education burden, for both customers and team, accompanies any major software change. One benefit inherent to a custom build is that any performance gaps that education can’t, or shouldn’t, correct can be addressed with a system enhancement pre- or post-launch.

There is no prescription or formula. In many cases the option to buy for small business is still obvious. But you may still have that hunch that you need something uncompromising, and that, in spite of the practical challenges that lay ahead, a custom build needs to be considered. Here are our rules of thumb for the final layer of validation.

The simplest solution might not be a technology one.

Is it possible you need no software change at all to get the job done? Business processes, like the modern automobile, sometimes get too complicated. Your Tesla model 3’s oversized tablet is saturated with so many controls, knick-knacks, and doo-dads, in service of the spirit of continuous improvement and modernization, that you forget that people were once able to commute safely(ish) and on-time in a Volkswagen Beetle. If your business processes are more like Model 3’s, ask yourself what it would look like to make it more like the spartan, minimalist, but effective and cheap Beatle?

We recently went to Katz’s Deli New York City, arguably the most high-traffic sandwich counter in the world. Katz is able to churn thousands of customers each day using a fully analog, paper based ordering and queueing process; not a single fancy handled POS can be found on-site.

Make sure you are not an island.

Start with a simple definition of the key jobs the system performs, and for whom. Branch that outward: which other industries face similar conditions? Then devil’s-advocate your own matrix. Let’s say you’re an auto-shop franchise looking at a custom inventory management solution. How is the auto-industry similar and different from a fleet company in terms of parts carried, how they’re booked and used? If you find real overlap, you can likely repurpose or retool an existing product, even though it may not be marketed for you. Where processes and problems are common to many business types, chances are you’re looking for a commodity with tons of like copies in the marketplace.

Ask whether standing still is the riskier move.

You may find a product that is decent, or even good for the immediate term. If you decline the innovation leap however, are you just racing toward the red light? Is your competitor working on the same thing, and is there a risk you get steamrolled? At that point it hardly matters whether you have the in-house know-how to maintain and administer. It’s time to learn; time to act.

Start with a pilot project.

If you haven’t gone through a custom software development project before, start with a more bite-sized initiative. The strict economic pay-off might not be there on this individual project, but it will be an effective litmus test for you and your vendor, which could yield significant long-term returns if you’re contemplating a bigger build. We recommend staying away from something that your customers will interact with directly. Back office process optimization is a good bet, or maybe building a Custom AI Agent to make some of your front-line employees more effective with day-to-day tasks. Oftentimes, middleware can be built in conjunction with what you have already, and you don’t need to throw the baby out with the bathwater. We recommend that the project span at least two teams to stress-test team dynamics and coordination efforts. That convenience lowers your risk, and lowers your upside by exactly the same mechanism.

Building leaves room for incidental discovery, the system or feature you never anticipated, emerging from the combination of deployed and outlined ideas. Open development processes carry a wider margin for error, but also a larger creative space, which leaves greater room for positive upside to unfold at any stage in the project. Sometimes random ideas cascade and coalesce and show you a completely new path. Penicillin, sandwiches… all discovered and invented by accident!

Wherever you land on this spectrum, we’re happy to help. Not getting the most out of your CRM, want to make a change, and don’t have the manpower to configure and set it up? We can help with that. Need a custom integration to track how work order completion time at your hotel corresponds to guest rebook rates? Maybe you want a custom developed AI-agent to make sure your hotel continues to perform well in AI search (a practice referred to as AEO). We can set up custom integrations, model the data, and build agents to the job.

Lastly, if you’re looking to go all the way and build an broad-scale application for your customers, those are the projects we love the most. In these arrangements, our goal is to look beyond the build and help you take your vision all the way and support through project management processes, SOP development, and training.

Whichever path you choose to take, our commitment to you will be the same:

1. Partnership: We care about getting to know you, and your business, for its own sake. For us it’s not just about a successful engagement, it’s fun. We hope you call us on our cell-phones when there’s a problem, question, or whim.
2. Attention to Detail: We see our work as a creative act and apply that energy down to button placement and column names. It’s better to do it right the first time.
3. Price: We will not charge you for things you don’t want or need, and offer you hourly rates that bigger shops with high overhead cannot beat.
4. Flexibility: If the engagement isn’t working out, no locked-in fixed scope agreements, so you can step away whenever it makes sense.