30-day customer-agent outcome pilot

Test one customer job. Measure what changes.

We set up one agent around knowledge you approve, run it for 30 days with a named person owning follow-ups, and count what actually happened next to your baseline.

Answers that come only from knowledge you approved — with the source shown.
Customer requests saved with context, and contact details when the visitor chooses to share them.
Every handoff sends a person an email with the reason — and the visitor's details when they chose to share them. Sales pilots add a named owner, a response deadline, and overdue flags.
A workspace matched to the job: simple request lists by default, a sales pipeline when the job is sales.
A count of everything that happened — questions, answers, requests, handoffs — kept separate from the results a person confirmed.

Tell us about the customer job

We review fit before proposing scope. Sending this form does not start billing and does not promise acceptance.

Your details

Pilot fit

Check us first

Security controls

See it live

Two demo sites, two live agents — pick one to talk to.

PortfolioAvery StoneFreelance brand & web designer. Ask about her work, pricing, and availability.B2B wholesaleEmber & Oak WholesaleA Seoul supplier. Ask its buyer agent about fit, wholesale terms, onboarding, delivery, and samples.
How we measure

How the pilot works

A controlled company outcome test, not a vanity chatbot launch.

We write down the agent's job, who it talks to, which channel, which questions, who answers, and what counts as a result — so the numbers at the end mean something.

  1. 01

    Scope

    Pick one agent job, one customer group, one main channel, and one set of questions that keeps coming back.

  2. 02

    Ground

    Approve what the agent may use — products, services, processes, policies, links, documents — and where it has to stop.

  3. 03

    Launch

    Publish or embed the agent with attribution, request capture, handoff, and a named person on responses already set up.

  4. 04

    Operate

    Check unanswered questions, request quality, response time, handoffs, and any errors from connected systems.

  5. 05

    Read out

    Put the baseline counts next to the pilot counts, note anything else that changed, and keep early signals apart from confirmed business outcomes.

What we measure

Eligible visits, customer conversations, captured requests, human handoffs, unanswered questions, and reviewed answer quality. Sales pilots also track first-human-response time. Sales, service, and booking outcomes are counted only when a person or a connected system confirms them.

Best initial fit

Any company — selling to consumers, to businesses, or both — that gets the same questions again and again, has enough conversations to watch, can approve the source material, and has a person who owns the next step. The best place to start is one narrow, repeated customer job — not a particular industry.

Is the pilot paid?

Yes — the guided pilot is a paid engagement. We propose scope and price after reviewing fit. Prefer to start on your own? Every plan includes the same measurement.

See plans and pricingStart a self-serve trial

We don't promise an uplift number — no one honestly can before seeing your real traffic. We promise the full count instead: every question, grounded answer, captured request, handoff, and response time, next to your baseline. You judge the lift.