Typical duration
4–6 weeks
A contained evaluation window with a defined website, question set, and review cadence.
Pilot program
Now preparingThe Guftara pilot program is for businesses with useful information already published online and a clear customer-service workflow they want to evaluate.
Typical duration
A contained evaluation window with a defined website, question set, and review cadence.
Starting point
Begin with information the business already publishes and is authorized to use.
Primary goal
Measure answer quality, coverage, latency, language needs, and human handoff requirements.
Pilot path
Agree on the customer-service workflow, website boundary, expected questions, languages, and success measures.
Configure the website scan, review Indexed Pages, and establish a representative evaluation set.
Test grounded answers in the private Preview Assistant and assess voice requirements where relevant.
Review findings, risks, infrastructure needs, and the recommended next production milestone.
Deliverables
A reviewed set of successful Indexed Pages from the agreed public website scope.
A controlled space for the business team to test representative customer questions.
Results for grounding quality, answer coverage, response behavior, and observed failure patterns.
When voice is in scope: language, model, latency, capacity, and human escalation requirements for the next phase.
Measures
Targets depend on the website and use case; Guftara does not publish invented customer results.
How consistently answers stay supported by information on the submitted website.
How much of the agreed representative question set can be answered usefully.
Observed latency, clarity, source usefulness, and reviewer confidence.
Which questions should move to a person and what context the team needs at that point.
Include your website, intended workflow, expected languages, and the outcome you want to improve.