The standard advice used to be simple: buy off-the-shelf for common problems, build custom for your edge. Then AI arrived and complicated things. Now every SaaS tool has an AI feature, every AI feature promises to replace a tool, and every sales call claims their product is “enterprise-ready.” Meanwhile, we keep seeing companies pay for software they don’t use, and build software they could have bought. Here’s the framework we use, updated for 2026.
Step One: What’s the Job?
Ignore the product for a moment and answer one question: what’s the job, and how often does it happen? Three buckets:
- Commodity jobs (payroll, email, spreadsheets, standard CRMs): the market has solved these well. Buying is almost always right.
- Differentiating jobs (the process that wins you customers, the workflow your competitors can’t copy): this is where custom software and automation earn their keep.
- Undefined jobs (you’re not sure how the process works at all): fix that first. Buying or building against a mystery process produces the same expensive result.
Step Two: Price the Whole Thing, Not the License
The license fee is the smallest number in a software purchase. The real total is license, implementation, integration, training, and the ongoing cost of the workarounds your team builds when the tool doesn’t fit. We’ve seen a $30/user/month tool cost a client six figures in a year once integration and rework were counted. We’ve also seen the reverse: a custom build that looked expensive and paid for itself in three months. Compare like with like, or don’t compare.
Step Three: Add the AI Twist
AI changes the calculation in three ways:
- It lowers the floor for automation. Workflows that were too small or too messy for a custom build now justify a thin automation layer. If the process is rule-bounded and high-volume, automate it before you buy or build anything.
- It raises the bar for “good enough” SaaS. If the off-the-shelf tool can’t expose its data or accept API calls, it’s a dead end for AI, automation, or anything else your business will need in three years.
- It makes “buy” cheaper to test. AI-powered tools have compressed the pilot from a procurement project to a weekend. The cost of being wrong has dropped. Use that.
What Companies That Get It Right Do
They buy the commodity, automate the repetitive, and build only what makes them different. Then they re-run the framework every year, because the market keeps moving and last year’s “buy” is this year’s “automate.”
If you’d like us to run this framework over your actual stack and processes, book a discovery call. We’ll tell you what to buy, what to automate, and what’s genuinely worth building.