Peaklight

Use cases

What an AI employee actually does

Not a chatbot with a new name. A digital team member that holds a defined role, works within rules you set, and hands work to a human when the rules say it should.

The first role we staff

Customer support

Answering the people who write to you, in your voice, at any hour. The employee works the channels your customers already use, answers from your knowledge, and hands complex cases to your staff with the full conversation attached.

This fits if

  • Enquiries arrive at all hours and wait until morning
  • One inbox, one overwhelmed person
  • Seasonal spikes your team cannot absorb
A refund request that needs a humanIllustrative
  1. RequestCustomer asks for a refund outside policy
  2. KnowledgeRefund policy cited from your own documents
  3. PolicyRefunds above threshold require approval
  4. ActionEscalated to a named human with context
  5. AuditEvery step above recorded and reconstructable

Once support is earning its keep

Back office and follow-ups

The repetitive work that spreads across your best people: chasing outstanding items, preparing recurring reports, keeping records consistent. Duties are proposed and approved, never assumed.

This fits if

  • Repetitive work spread across your best people
  • Growth plans stuck behind hiring budgets
  • You want to scale service quality, not headcount
A routine follow-up the employee is cleared to sendIllustrative
  1. TriggerItem outstanding past the agreed window
  2. KnowledgeAccount history and prior correspondence
  3. PolicyFollow-up wording pre-approved by you
  4. ActionMessage sent, thread left open for reply
  5. AuditSent copy and reasoning stored

The part that is not optional

Governed by design

What the employee may say, what data may leave, and when a human steps in are enforced in software outside the model. That is the difference between an assistant you hope behaves and an employee you can hold accountable.

This fits if

  • Wary of handing customer data to a black box
  • Locked into a platform once before, never again
  • You need a real audit trail, not screenshots
A request that policy blocks outrightIllustrative
  1. RequestAsks the employee to share another customer record
  2. PolicyData egress rule denies the request
  3. ActionBlocked before the model is called
  4. ResponseCustomer told what can be shared instead
  5. AuditBlock recorded with the rule that caused it

By industry

Running a hotel?

We wrote the hotel version of this page in hotel terms: the guest questions answered from your own documents, and the guest who asks for a person reaching one.

How it works for hotels

Which role would you staff first?

Tell us what the work looks like and we will tell you honestly whether an AI employee is the right answer for it.

Talk to us