Build, buy or connect: how to choose your CRM now that AI can build one

Until recently, a small company choosing a CRM had one real question: which one to buy. Building your own was for companies with a development team.

That changed. With AI tools, people who have never written code are getting a working first version of a CRM on screen in days. So a second question is now on the table: should we buy one, or build our own?

There's a third option most teams skip, and it's often the best one. This guide walks through all three, what each really costs over time, and how to decide for your company.

First: what a CRM actually has to do

Before comparing options, write down the jobs. In most small companies, a CRM has to:

• Keep one record per client. Who they are, what they bought, every conversation, in one place.
• Show the pipeline. Every open opportunity, its stage, and what happens next.
• Log email and calls automatically. If it depends on someone remembering to copy it in, it won't happen.
• Remind people to follow up. Tasks with an owner and a date.
• Answer basic questions. How many leads came in, how many became clients, where they came from.
• Control access. Who can see which clients, and who can export the whole list.
• Keep the data safe. Backups that someone has actually tried restoring.

The first three are what people picture. The last two are what people forget, and they matter most on the day something goes wrong.

Option 1: Buy

You pay a monthly fee per user for a tool someone else builds, hosts, secures and updates.

• Good: it works from day one, the boring parts (hosting, backups, security, updates) are included, and anyone you hire later may already know it.
• Watch out for: the entry price is rarely the price you end up paying. The features many teams want sit in higher plans, and the jump between plans can be large. Paying for features nobody uses is also common.

Option 2: Build

You create your own tool, now often with AI writing most of the code.

• Good: it fits exactly how you work, there's no per-user fee, and you control every detail.
• Watch out for: getting it on screen is the start, not the end. From then on you own the hosting, the backups, the security, every fix and every change. Someone in the company becomes its support team.

Option 3: Connect

You keep a standard CRM as the base and build small things around it: an intake form, a report, a connection between two tools that don't talk to each other.

• Good: you get the fit of building where it matters, without owning the whole system. If one small piece breaks, the CRM still works.
• Watch out for: each connection still needs an owner and a short note on how it works. Ten undocumented connections become their own mess.

What each option really costs

Compare them over the years you'll actually use them, say three, not over the first month. Put these lines side by side:

• Buy: the price of the plan you'll really need (not the entry plan) × users × months. Plus the setup hours.
• Build: hosting + backups + AI usage per month × months. Plus the hours to build it, the hours to fix and update it every month, and what those hours are worth to you.
• Connect: the base plan × users × months, plus the hours to build and look after each small piece.

The line people leave out is the last one: hours. In a small company, those hours usually come from the owner or the most capable person on the team, and they come out of selling or serving clients.

How to decide

Building makes sense when:

• you have a technical person who will own it for years, not just launch it;
• nothing on the market fits how your industry works;
• the three-year numbers still work after you count the hours.

Buying makes sense when:

• your process is close to how most companies sell;
• nobody on the team wants to be the system's support desk;
• you'd rather spend the hours on clients.

Connecting makes sense when:

• a standard CRM covers most of what you need, but two or three things are specific to you;
• your tools already work but don't talk to each other.

In the companies we look at, it's usually the third: the tools are there, they just don't talk to each other. And whichever you choose, the same rule from what to automate first applies: start with the step that repeats every week with the same steps.

The data decides more than the tool

A CRM is only as good as what goes into it. If today nobody keeps client data clean, a new CRM, bought or built, will be just as messy within months. Before switching anything, agree on three things:

• who enters what, and when;
• which fields are required, and which nobody needs;
• who checks it once a week.

And if your CRM holds clients' names, emails and conversations, find out which privacy rules apply to you. Ask a lawyer, not an AI tool.

A one-hour exercise

• List everything your team does with a client from first contact to payment. One line per step.
• Next to each step, write which tool covers it today, or "nothing", or "a spreadsheet".
• Circle the steps where something gets lost, copied by hand or forgotten.
• For each circled step, write buy, build or connect, and who would own it.

Most teams find that the answer isn't a new system. It's two or three gaps between the tools they already pay for.