Your renewal calendar is now a build-vs-buy list

Pull the renewals due in the next twelve months and sort them by notice date. A third of the organizations McKinsey surveyed have already replaced at least one line on a list like that with software they built themselves, and that changes what a notice date means. SaaS replacement used to be a war story. It is now a survey line, and each notice date is a decision.

Build wins some of those decisions. Renewal wins more of them. Either answer deserves six numbers on one page before anyone signs.

Why the run cost belongs on the same page as the license

Software spend is one of the few operating lines a mid-market company can move inside a quarter without touching headcount. A $120K contract renewed because the notice date passed is a capital decision made by a calendar.

The build side has its own bill. One in five organizations told McKinsey that AI operating costs already limit how much they deploy. A build that saves the license and doubles the run cost is a renewal with more steps, and the run cost arrives after the build is live, when the vendor contract is gone.

Six numbers before you build or renew

1. VENDOR PRICE. The renewal quote, plus the modules you would add at the next tier, plus the uplift the vendor has taken over the last two years.
2. BUILD COST. One-time engineering, including the spec work and the integration to the system of record. Price the internal hours too.
3. RUN COST. Model spend per month at current volume, plus hosting, plus the hours a named person spends reviewing exceptions. Then the same figure at double the volume, because usage grows once a system works.
4. MAINTENANCE. Who fixes it when the ERP upgrades or the model version is retired, and what that person costs per year. A build with no maintenance owner is a renewal with a delay.
5. WORKFLOW STABILITY. Does the job the tool does fit in a paragraph, and has that paragraph changed in the last two years? Stable jobs suit builds. Jobs the vendor keeps adding features to suit renewals.
6. PAYBACK. Vendor price minus annual run and maintenance cost, divided into build cost, in months.

Then the rule:

• Payback under 12 months and a stable workflow: build.
• 12 to 24 months: build only if the workflow is stable and the vendor is the one raising prices.
• Over 24 months, or a workflow that keeps changing: renew, and renegotiate with the login count in hand.

An illustrative example, one column apart

The numbers below are illustrative and do not describe a client. A $120K annual contract. Build cost $30K. Run cost $40K a year at double volume, maintenance $10K. Payback: 30 divided by (120 minus 50), about five months. Build.

Same contract, same build, but the run cost at double volume comes in at $110K. The annual saving before maintenance is $10K, payback is three years, and the model bill grows with usage. Renew, and take the login count into the negotiation.

One column separates the two answers. The 20% already feeling run costs priced that column after the build went live.

What McKinsey found

Source facts, from Dana Ellison's write-up of McKinsey's State of AI 2026 for Yahoo Finance. 32% of organizations decided against buying off-the-shelf software and built their own with agentic coding tools instead. Among AI high performers, the 6% of respondents who attribute at least 5% of EBIT to AI, close to half skipped a purchase in favor of a build; among their peers, 31% did. 20% of organizations already feel the pinch of AI operating costs. Among companies over $1 billion in revenue, 40% are scaling agents in one or more functions, up from 27% the year before. Technology leads the build decision at 41%, healthcare at 39%, professional services and energy at 38%.

My read, not McKinsey's: the survey shows that high performers build more. It does not show that building made them high performers, and I would not claim it. What a build does force is a written spec of what the workflow does, and that document is worth having either way. It is also the cheapest input to a renegotiation.

The $1 billion number I would ignore. A $60 million company has a controller, an ops lead and a renewal date. It can pick one contract, price the six numbers and be done inside a quarter.

The date is the deadline, the page is the decision

Most contracts on the list will renew, some at a lower price. One or two will get replaced. All of them should get the page, because the alternative is a calendar making capital decisions on the CFO's behalf.

If you have a renewal over $50K due in the next two quarters and want a second read, send me the contract line and a paragraph on what the tool does. I will tell you which of the six columns decides it.