Real AI ResultsDigest
4 listings · every one sourced

AI results at real companies

Outcomes companies report from AI and agents, with the fine print: what each number shows, what it does not, and where it came from.

EverythingEarnings callsPeople movesJobs
Result of the week · Capital OneNYSECOF
10 daysData agents analyzed ~350 cloud resource types. By hand, the author estimates 6–9 months.
Company claimOpen the breakdown
Get the weekly digestThe rows that matter, and what each one proves. Free.
Result · Capital OneNYSECOF

Data agents analyzed ~350 cloud resource types in 10 days

Company claimCompany tech blog, Jun 9, 2026Added

Capital One engineer Ram Manohar Bheemana reports using agents to analyze dormancy across roughly 350 cloud resource types in 10 days. Agents wrote the Spark SQL detection logic from a standardized inventory; people validated it.

~350cloud resource types analyzedReported
10 dayswith agentsReported
6–9 moby handEstimate only

What it shows

  • A large bank used agents on a real inventory problem, not a demo.
  • Agents wrote the logic; people kept the review.

What it doesn't show

  • Dollars saved.
  • Time spent on human review.
  • Error rates in production.

Questions we're still asking

  1. 01How was the 6–9 month baseline estimated?
  2. 02How many hours did people spend validating the output?
  3. 03What was the error rate in production?
  4. 04What did it actually save?

Company claim: The company published or said it about its own work. Nobody has audited the numbers. How we check

More results

Read the sourceReport a correction