The Buy It Guide: when to buy, when to build, and the HR AI tool landscape

Every week brings a new AI tool “built for HR.” Some of them are excellent. Many are a thin layer on top of a general model you already pay for. The hard part isn’t finding options. It’s knowing which problems deserve a purchase, which deserve a build, and which deserve to wait.

This guide gives you a way to decide, and a map of the landscape so you know what you’re looking at.

The short version

Situation	Usually the right move
The problem is common to every company (screening, scheduling, help desk, document search) and the stakes of a wrong answer are moderate	Buy, probably from a vendor you already use
The problem is specific to your organization (your competency model, your policies, your process) and a good answer depends on your context	Build, usually on a platform you already have
The problem touches decisions about people (hiring, pay, performance, discipline)	Slow down. Buy or build only with a clear human decision point and a review from Legal
Nobody on the team has tried to solve it with the tools you already have	Wait, and run a two-week experiment first

When to buy

Buy when the problem is generic, the volume is high, and the vendor has solved it many times before. Signs you should buy:

Someone has done this a thousand times. Resume parsing, interview scheduling, benefits questions, knowledge search. Your version of the problem is not special, and that is good news.
It lives inside a system of record. If the capability belongs in your HRIS, ATS, or LMS, the vendor who owns that data will usually win on integration alone.
The risk is in the plumbing, not the judgment. Security, access control, audit logs, uptime. Vendors are good at plumbing.
You need it everywhere at once. A purchased tool comes with support, documentation, and someone to call.

When to build

Build when the value comes from your context: your policies, your language, your leveling, your process. Signs you should build:

The answer depends on documents only you have. A policy assistant is only as good as the handbook behind it. Yours is the one that matters.
The workflow is yours. Your onboarding, your calibration process, your HRBP case notes. No vendor’s template matches it.
It’s small. Many of the most valuable HR builds are 20 percent of a product: one workflow, for one team, done well. See the builder mindset guide for 25 of them.
You already have the platform. Most organizations are already paying for a place to build: the assistant and agent features inside Microsoft 365 or Google Workspace, ChatGPT or Claude for business, or a low-code automation tool. Building on an approved platform is far easier to get through Security than a new vendor.

When to wait

Wait when nobody has tried the simple version yet. Before any purchase, run a two-week experiment: can one capable person, with the tools you already have, get 70 percent of the way there? If yes, you have learned what you actually need, and your vendor conversations get much sharper. If no, you now know why, and that reason becomes your requirements list.

The HR AI landscape

A map of what you will encounter. Categories matter more than names; the names change monthly.

1. AI inside your core systems

Your HRIS, ATS, LMS, and engagement platform are all adding AI features: drafting, summarizing, search, recommendations. Strength: already integrated with your data and already approved. Watch for: features that are a demo rather than a product, and pricing that moves them into a new tier.

2. General assistants and platforms

Microsoft Copilot, Google Gemini, ChatGPT, Claude, and their enterprise plans. Strength: the most capable models, the broadest use, and increasingly the place where your team can build its own assistants and agents without code. Watch for: uneven adoption. Without fluency, most licenses go unused or produce slop.

3. HR point solutions

Tools built for a single job: candidate screening and matching, interview intelligence, skills inference, survey and comment analysis, compensation intelligence, employee help desks, performance and feedback assistants. Strength: depth in one problem. Watch for: anything that makes or shapes a decision about a person. Ask how the tool is audited, what data it was trained on, and where the human decision point is.

4. Builder and automation platforms

Agent builders, low-code automation, and workflow tools that connect your systems. Strength: this is where HR teams become builders. Watch for: security defaults, access to people data, and who maintains what gets built.

5. Learning and enablement platforms

AI fluency courses, simulation and practice tools, and coaching assistants. Strength: the fastest way to raise the floor across a workforce. Watch for: generic content that doesn’t reference your tools or your work. Fluency sticks when it’s personalized. See the AI Fluency demo.

Ten questions to ask any vendor

Which model powers this, and what happens when it changes?
Where does our data go, who can see it, and is it used for training?
What is the human decision point? Show me, in the product.
How do you test for bias and adverse impact, and can we see the results?
What does an audit log look like for a decision this tool influenced?
What does the product do when it doesn’t know?
Which of these features are live today versus on the roadmap?
What does it integrate with, and what does that integration actually sync?
What happens to our data and our configurations if we leave?
Can we talk to a customer who has used it for more than a year?

The decision on one page

Before you sign or build, write down:

The problem, in one sentence, with a number attached (hours, days, cases, dollars).
The two-week experiment you ran with what you already have, and what you learned.
The human decision point, and who owns it.
The data the tool will touch, and who approved that.
What “working” looks like in 90 days, and how you’ll know.

If you can’t fill that in, you’re not ready to buy or build. You’re ready to learn, and that’s a fine place to start.