Everyone uses AI. Few get value from it.
The adoption numbers stopped being interesting a while ago: McKinsey’s latest global survey found that agentic AI — systems that take actions, not just write text — is already in use at 40% of large organizations, and Deloitte’s enterprise research puts a third of surveyed companies on the path to using AI to rebuild core processes.
The interesting number is a different one. Ask those same companies what measurable value they have captured, and the room goes quiet. Using AI is easy. Getting money out of it is not.
Where AI genuinely earns its keep
From the work we have shipped with clients this year, the wins cluster in unglamorous places:
- Drafting and summarizing. First drafts of emails, reports, meeting notes and support replies. The AI does the blank-page work; humans edit.
- Triage. Routing incoming support tickets and emails to the right person, with a suggested reply already attached.
- Search over your own documents. Contracts, policies, wikis — answering “what does our policy actually say?” in seconds instead of an afternoon.
- Developer productivity. Code completion and test scaffolding. Real time savings, but only when somebody reviews the output like a junior colleague’s work.
Where it still disappoints
- Fully autonomous decisions. Systems that “just handle it” sound great until they confidently mishandle the one edge case that matters.
- Anything requiring exact facts. If a wrong number is expensive, a human verifies it. Every time.
- Solving a problem you haven’t defined. “Let’s add AI” is not a goal. “Cut first-response time on support emails from 4 hours to 30 minutes” is.
How to start without burning money
- Pick one boring, repetitive workflow. Not a moonshot — something your team does weekly that they hate.
- Measure the baseline first. If you don’t know how long it takes today, you will never prove the AI helped.
- Keep a human in the loop. Every automation ships with an easy undo and a clear owner.
- Give it three months. Kill what doesn’t show numbers. Scale what does.
The companies winning with AI in 2026 aren’t the ones with the boldest vision. They’re the ones that automated their fifth-boringest task and measured the result. Start there.