AI Speeds Up Tasks, Not Processes , and Enterprise Buyers Are Starting to Notice

8. AI Speeds Up Tasks, Not Processes , and Enterprise Buyers Are Starting to Notice

A blog post by Frederick van Brabant, published May 15, 2026, reached 331 points on Hacker News within hours of posting. The core argument: AI tools accelerate individual tasks inside a process, but the process itself , with its handoffs, approvals, waiting periods, and coordination overhead , stays the same. Cutting the time a worker spends drafting a document from 40 minutes to 4 minutes does not shorten a two-week approval cycle. The bottleneck moves, or more often, it does not move at all.

This is a direct challenge to the ROI framing that Salesforce, Microsoft, and ServiceNow have built their enterprise AI sales motions around in 2025 and 2026. All three have published productivity metrics anchored to task-level time savings: Copilot for Microsoft 365 cites hours saved per user per week; Salesforce Agentforce pitches reduced handle time per case. Neither metric maps cleanly to process throughput or revenue cycle compression, which is what CFOs actually sign off on. If buyers start asking for process-level evidence rather than task-level benchmarks, the current generation of ROI calculators breaks. Vendors that cannot show end-to-end cycle time improvements will face harder procurement conversations in H2 2026.

The pattern here is not new. It mirrors the early cloud productivity debate, where per-seat license savings looked compelling until IT leaders realized total cost of ownership included migration, retraining, and integration debt. AI is now entering that same scrutiny phase. The next move to watch: whether enterprise analysts at Gartner or Forrester formalize process-throughput metrics as a procurement standard, and whether any major vendor gets ahead of it by reframing their benchmarks before buyers force the issue.

Source: I don't think AI will make your processes go faster