Enterprise buyers are finally scrutinising the cost of AI add-ons

Software vendors spent two years pitching automated intelligence as an urgent upgrade. Enterprise buyers mostly nodded along during early trials. That initial phase of unvetted experimentation is ending.

Corporate purchasing offices are reviewing renewal invoices with new scrutiny. Line items for generative capabilities often double standard seat prices. Finance directors want concrete evidence of operational savings before approving those premiums.

The breakdown of per-seat bundling

Traditional software pricing relied on clear metrics like active users, storage volume, or server cores. Generative tools complicate this relationship. Vendors incur real computing costs per query, but they prefer to sell predictable monthly seats to protect their margins.

This mismatch creates friction during contract renewals. Customers find that only a fraction of their staff uses the added features regularly. Paying a blanket surcharge for dormant capabilities no longer makes financial sense.

Procurement teams now demand usage-based tiers or clear carve-outs for staff who do not need the tooling. Some enterprises refuse blanket bundles altogether, choosing to keep legacy tiers until vendors offer granular controls.

Measuring output against real cost

Assessing the business value of these additions remains difficult. Vendors cite internal productivity studies that promise hours saved per week. In practice, enterprise buyers struggle to tie automated text drafts or code suggestions to measurable bottom-line gains.

Most enterprise buyers now recognise that bundling an experimental tool into core infrastructure shifts technical risk onto the customer. When a tool generates flawed output or requires extensive human correction, the labour savings disappear. The procurement audit is therefore expanding to include operational quality reviews.

A slower, more defensive renewal cycle

Software suppliers are adjusting their posture as sales cycles lengthen. Where sales teams once closed multi-year expansions on broad AI roadmaps, they now encounter strict proof-of-concept requirements. Enterprise legal and compliance teams also insist on indemnities that vendors hesitate to provide.

The result is a return to traditional procurement discipline. Companies still want automation, but they are unwilling to subsidise vendor infrastructure trials. The era of unconditional software budget expansion has quietly closed.