Buy or build? How HR leaders can make smarter tech decisions in the age of AI

So an old question is back on the table. Should you buy your HR tech, or build it yourself?

Anders Holmberg, Chief Revenue Officer at Eletive, opened our AI in HR 2026 digital summit with exactly that question. He has spent more than 15 years in SaaS (software as a service), working closely with product teams. Here are the key takeaways from his session.

Building has never been easier

The numbers are hard to ignore. According to GitHub’s 2025 report, 121 million new software projects were created in 2025. That’s 230 every minute. AI-created projects grew by 200% in a year.

Early this year, that shift worried investors. People started talking about a “SaaS apocalypse”: if everyone can build their own tools, why would anyone buy software? Valuations for some of the biggest SaaS companies dropped sharply.

Companies didn’t stop buying software, though. The public SaaS market has recovered by 40% since the dip in February. Software spending is still growing by 15.5% a year (Gartner, 2026), and the average company still runs around 300 tools (Zylo, 2026).

So it’s probably not buy or build. It’s both, side by side. The real question is when to do which.

When building makes sense

Four signs building could be the right call:

•  The logic is unique to you. Your competency framework, your job architecture, your internal mobility rules. The moment a vendor generalises it, it stops being yours.

•  The scope is well defined. One workflow. One team. One question. This quarter. Build projects fail when they quietly turn into platforms meant to last for years.

•  The value sits in data you own. If the clever part is your data and the AI is only the interface, you’re closer to building than you think.

•   Ownership is clear. If you can’t name the person and the budget line for 2029, you’re not building. You’re prototyping.

What could that look like in HR? Maybe a micro-survey. How was the kick-off? Who wants to join the Christmas party? Who is employee of the year? There’s no need for anonymity, trend lines or mapping to your org hierarchy, so a simple build does the job.

Maybe it’s a board and reporting pack: an AI agent that pulls data from several systems into your quarterly report, in your own format, every time. Or a policy Q&A assistant trained purely on your handbook, your agreements and your local rules.

All three work for the same reason. The value is in your data, not in the model.

When buying is the better call

Six signs buying is the smarter option:

•   The problem is the same everywhere. Payroll, hiring, compliance reporting, performance reviews. Someone has probably already solved it better than you will.

•   You need a benchmark. You can’t build comparison data. It’s an asset you either have or you don’t. In engagement, that’s often where the real insight sits.

•   Your regulatory and data exposure is wide. You can build a GDPR-compliant app today. Keeping it compliant as laws change is the hard part. If you risk breaching GDPR or employee anonymity, you want a vendor whose full-time job is keeping up.

•  The models underneath keep changing. Sometimes the output drifts quietly, and your monthly report looks a little different each time. Vendors handle those migrations for you.

•   The expertise matters more than the software. Buy it, and someone else carries implementation and adoption. Build it, and you become the expert. Permanently.

•   You want someone else to answer the phone. Every internal build has a launch party. Very few have support on a Monday morning.

What building really looks like: an example from Eletive

This summer, our revenue team built an AI tool that helps sales reps choose which companies to contact, and drafts the first message.

It connects tools we already pay for: HubSpot as our CRM and source of truth, Apollo for company data, n8n for workflows and Claude on top. Every morning, the data updates based on recent events and our past conversations. The AI picks the companies that make most sense to contact and drafts a tailored message. A sales rep fine-tunes it before it goes into the email queue.

This is likely the shape many internal builds will take from now on. Not “we built our own system”, but “we connected four things we already own to solve a problem nobody sells a product for.”

The first prototype took a handful of days, and it was impressive. It felt 90% done. The last 10% is where the real work is. What happens when a contact is about to leave their company? When two companies are linked? Exception handling, data quality, response times, loading data properly every time (not most times) and training all take time.

So far, it’s taken hundreds of hours and two people working on it almost full time. More than we’d hoped. It’s still worth it. The goal is 25% higher output per sales rep. It only made sense because we had the resources and the data was unique to us.

The lesson: getting to the first win is easy now. Getting to the hundredth is exactly as hard as it always was.

Missed the session? Watch the full session here.

The tools got easier. The job around them didn’t.

Building is about much more than code. You become a project manager: scope, sequencing, exceptions, testing, rollout. You own your data architecture: where the data lives, what the source of truth is and what happens when two systems disagree. You also become a product manager, spelling out exactly how it should work, including every edge case you’d rather not think about.

AI will build exactly what you ask for. The hard part is knowing what to ask for.

Neither side is one line item

If you build, count hosting and model costs, security reviews, legal sign-off, testing and change management. Count maintenance when the API shifts underneath you, and who handles support. Count the “bus factor” too: what happens if the person who built it leaves? Maybe the biggest cost of all is opportunity cost. If you run a burger restaurant, your time is probably best spent on the burgers.

Buying has hidden costs too. Integration, implementation, data migration, adoption and training, admin time, vendor management and switching costs if it doesn’t work out.

There are costs on both sides. The real difference isn’t the money. It’s who carries the expertise and the risk.

The future is buy and build

Most HR teams will likely buy tools for their core work: engagement and performance, recruiting, core HRIS, time management and payroll. Those tools give you the interface for day-to-day work, data handling and compliance, and vendor expertise and support.

On top of that sits an AI layer that can safely access the data. On top of that again, you build your own in-house apps and an analytics layer that connects the systems. Many companies are already moving in this direction.

Where to start on Monday morning

A few practical tips from the Q&A:

•   Check what you already pay for. If you run on Microsoft or Google, you likely already have AI tools you can experiment with.

•   Start with the problem, not the tool. Describe your problem to an AI assistant and ask how you might solve it.

•   Look at workflow tools. Moving data from one system to another is often where the first quick wins are.

•   Keep your first project low risk. Choose something with no personal data. When personal data is involved, bring in your legal team early.

Missed the event? Watch the full replay of AI in HR 2026: From People Insights to Business Impact