Trust & privacy

Patterns over people.

SightLift's default lens is the pattern: topics, trends, teams, with no one reading over shoulders. You decide how deep the view goes, and who sees what.

What leaders see

  • Topics and trendswhere AI is taking hold, team by team
  • Which skills are actually workingusage, acceptance, and lift per skill
  • Cost and valuespend joined to what it produced
  • Team-level rollupsthe way you run headcount and budget

What stays out of sight

  • Conversation textout of sight at the default data depth; you choose if that changes
  • Non-work topicsexcluded from semantic analysis; they touch cost totals and nothing else
  • Sensitive detailbehind a separate clearance, even for admins
  • Anyone else's dataevery customer's data is isolated to their own tenant

Your data, on your terms.

AI conversations can feel personal, even on company accounts. We respect that, so SightLift starts shallow and opens up only as you choose.

Default

Metadata only

Topic, complexity, tokens, skill usage. Enriched signal for the map and the score, with no conversation text read by a human.

When you choose

Deeper views

Expose more in the product over time, on your terms, for the people you pick, at the pace your rollout calls for.

Always gated

Sensitive detail

A separate clearance, distinct from admin rights, enforced where data leaves the system: redacted rows, never content.

Never trained on

No models of any kind are trained on customer data: no foundation models, no fine-tuning, no classifiers. Our AI subprocessors work under commercial terms that exclude training.

Never sold

Your data is yours, and it is never sold. Where it informs anything beyond your account, like cross-company benchmarks, it is anonymized and aggregated first.

Happy to talk it through.

Security review coming up, or just curious how the data depths work in practice: we welcome your conversation on it all.

Email [email protected]

Vulnerability reports go to the same address, and privacy requests to [email protected].