Build vs Buy: That is the question

Overview

"Couldn't Co-pilot do that?"

If you've said this, or heard it in a leadership meeting, you've already entered the build vs buy conversation. Your organisation will be making many decisions like this over the next few years. And every single organisation is asking the same question.

Take Sumday. We have a team of engineers who could build a customer contract management system. Should we? They could build an invoicing tool with their eyes closed. Should we? As AI becomes operational rather than experimental, organisations are being rewired around the software they adopt, the agents they build on top of models like Copilot or Claude, and the custom tools they choose to develop internally.

The question isn't "could we build it?" It's "should we, and how quickly can we?"

Could vs should

If a use case is worth pursuing, it's usually because it delivers significant time and cost savings, or a meaningful lift in quality. That means every month it isn't implemented is expensive. Very expensive. If you have the internal talent, time and budget, you can build almost anything right now. The real question is whether that's the right use of your organisation's capacity, and whether there's a compelling advantage in scoping, building, maintaining and improving it yourself.

The biggest mistakes we see

The AI committee

Leadership knows the organisation needs to move on AI, fast. So a committee gets stood up, often with the best intentions, and a monthly calendar invite goes out.

Two problems immediately emerge.

First, you're relying on a workforce to surface use cases when most people have barely scratched the surface of what these tools can do.

Second, the business cases that come through are often inconsistent and high-level. The committee isn't making clear calls backed by a well-scoped case, forecast ROI and a game plan quickly.

It gets noisy and overwhelming fast. One approved use case consumes all the available capacity, two if you're lucky.

What you actually need is an engine for innovation, fuelled by the people closest to the work, producing business cases with enough quality to unlock faster decisions and proper governance. Why would an AI use case forecast to save $50k a month or 20 hours a week need to sit in someone’s Outlook for 4 weeks because they missed the ‘AI Innovation Committee’ meeting yesterday.

IT confuses "could" with "should, and quickly"

IT has a critical role to play, but the trap of asking them to build everything never works well. People are busy. BAU doesn't stop unless you actively create the conditions for it to. The tools are new, the bugs are real, the testing is intensive and the maintenance is ongoing. A use case that looked clear-cut suddenly takes six months, and all momentum is gone. That's not IT's fault. It's the consequence of a poor plan, or no plan at all.

Leadership hasn't seen what's possible

At best, most senior leaders are using Copilot for basic prompting. If you haven't put real time into curiosity and had an aha moment, it's genuinely hard to understand why a company like Canva would ask 5,000 people to stop work for a week and experiment with AI tools at a paper cost of $17 million. The answer is that the ROI is extraordinary.

Leaders who aren't close to AI native take a long time to progress experiments and drive change, which is fine until someone above them has that aha moment and suddenly the organisation is thrown into turmoil. Something that could have been progressed at a pace the organisation could tolerate gets pushed at a pace it can't. So much for change management.

On the business case for build vs buy

Building internally is not free. The costs are real:

Manager time to scope and oversee the work. IT time to build and integrate. AI token costs to run it. Ongoing maintenance. Team upskilling as the technology evolves. Continuous updates, at minimum every few months.

On the buy side, most vendors will support a pilot and most will work on a 12-month contract. So even if you're planning to build internally, you can be learning while using, seeing the benefits and the shortfalls and upskilling your team in parallel. If you reach month 12 and your internal build is ready to go, you didn't waste money. You bought time and knowledge.

Buying makes sense when the problem is well-defined, a proven solution already exists, and the cost of delay is real. That last point matters more than most leaders realise. If a $50k tool eliminates the need to hire an FTE, but your team couldn't build the equivalent for 12 months, you didn't save $50k by waiting. You spent 12 months at full labour cost and still ended up with the hire.

The opportunity cost of indecision is very high right now.

The organisations moving fastest aren't necessarily the ones with the biggest budgets or the most technical capability. They're the ones making clear-eyed decisions quickly and treating AI adoption like any other business case.

The problem is that most organisations, particularly in the public sector, don't yet have a reliable read on how long or how expensive either path actually is. Without that, decisions get made on instinct, vendor pressure or whoever made the loudest case in the last executive meeting.

How can Sumday help?

We're on a shared mission to help you make better decisions, faster. We help you get a list of the highest ROI use cases, produce the business case, prioritise, build or buy, evaluate and scale.

If you're sitting on a build vs buy question right now and you feel like you should have a bigger picture plan, talk to us.