The default move when a team needs a new internal tool is to subscribe to one. Calendly for scheduling, Notion for docs, Retool for dashboards, Mixpanel for analytics. Six months later you are paying $400 a month for tools you barely use, integrated badly, and you still cannot get the one report your CEO actually asks for.
Here is what changed for me last year: with Claude as an engineering partner, the cost of building a small custom tool dropped from "two weeks" to "one weekend." That math reshapes the build-vs-buy decision. If a tool only needs to do your version of something, and the surface area is small, it is now genuinely cheaper to build it.
The tools I have built in the last six months that replaced SaaS line-items: a lightweight CRM for my own pipeline (replaced a $40-a-month plan), a private analytics dashboard against Postgres (replaced a paid analytics plan for one product), a receipt generator with print templates (replaced an accounting add-on). None of these are good enough to sell. All of them are perfect for the one team using them.
The advantages compound. You own the data. You can change the schema in an afternoon. You can wire it into your own stack in ways the SaaS vendor will never expose. You stop budgeting for tool subscriptions you forgot you authorized.
When should you still buy? When the network effects are the product (Slack, Linear), when compliance is the cost (most billing tools), or when the surface area is honestly bigger than a weekend (Stripe, Sentry). Anything else — at least price the build.
My personal pattern: keep a tools backlog alongside the product backlog. When something I'm paying for irritates me twice in a month, it becomes a candidate for a weekend rebuild. Half the time the build is cheaper to maintain than the subscription was to ignore.