Skip to content
The Giga platform

Build the agent. Understand the outcome. Improve what happens next.

Configure the work your agents do. Follow the conversations they handle. Work with Scout to find and test improvements.

Scout works alongside your team. Build an agent, investigate behavior, propose a change, or run a test.

01

Build

Describe the agent. Make it yours.

Policies, knowledge, and connected tools define what your agent can do.

Scout builds it with you. Ask it to draft or update instructions and connect the tools a workflow needs.

Policies

Instructions you can read. Behavior you can change.

Define scenarios, brand voice, and rules. Review the draft before publishing a new agent version.

PoliciesScenarioDraft
Synced across 2 agents. Edits apply to both.

Service Appointment

Use when a customer wants to book or move an appointment.

  1. 1.

    Check availability with find_available_appointments.

    • If a time is open: offer an available appointment time.

    • If nothing fits: follow the scenario “Waitlist”.

  2. 2.

    If a person is needed, transfer with transfer_conversation.

Tool access

Give the agent the tools to finish the job.

Attach APIs to an agent and control how they can be called. An available integration is not automatically an available agent tool.

Data sourcesAPIs overview
Attached APIs
Default Actions
In your team, not attached
add_servicePOST
Shared resources

Connect your systems. Reuse your knowledge.

Give agents access to selected knowledge, shared policies, APIs, and code. Keep reusable resources separate from each agent’s configuration.

Knowledge baseService policies
Attached toService agentBilling agent
TitleTags
Viewing 4 of 214 documents
02

Observe

Every outcome has a story.

Move from performance measures to the conversations and actions behind them.

Scout finds the story behind a number. Ask which conversations moved a metric, and why the agent acted as it did.

Conversations

Go from a metric to the conversation.

Open the cases behind a number. Follow their interactions, customer messages, agent responses, and actions.

Booking rate 64%Booked · 755#4821

I need to move my appointment to Friday.

Customer · 2:14 pm

Happy to. What time works for you?

Service agent v12 · 2:14 pm

Afternoon, if possible.

Customer · 2:14 pm

Friday has 2:30, 3:15 or 4:00.

Service agent v12 · 2:15 pm

2:30 works.

Customer · 2:15 pm
create_bookingBooked · Friday, 2:30
Evaluation details

See the evidence behind the assessment.

Inspect resolution, sentiment, and other extracted fields alongside the conversation. Evaluation criteria can be configured for your agent.

Tickets#4821
Resolved

Customer moved to Friday at 2:30. Booking confirmed.

1m 12sSep 16, 2:14 pm
HallucinationLow
SentimentPositive
IntentReschedule appointment
booking_confirmedtrue

Actions taken

Ran find_available_appointments and offered three times. Booked 2:30 with create_booking.

Quality review

Find patterns across your conversations.

Review quality measures over time, spot what’s working, and route cases into human review.

ReviewTickets
Match allTimes offered is 3Resolution is Resolved
DescriptionCreatedOutcome
212 tickets since Sep 12
03

Improve

Make the next version better.

Follow one change from conversation evidence to a measured comparison.

Scout proposes the next version. Give it a KPI. It grounds each change in conversation evidence and tests it before you publish.

Objectives

Define what success means.

Specify the outcome you want to measure and which conversations qualify. Track objectives and guardrails for your agents.

KPIsAppointment booking rate
Current64%
Target75%11 pts to go
Counts as success
A booking is confirmed
Eligible conversations
Customers requesting an appointment · 1,180 in 7 days
Guardrails
Transfer rate ≤ 15%11%
Median latency ≤ 1.5 s1.2 s
Opportunities

Turn a recurring pattern into a proposed change.

Review improvement items and trace them back to the conversations that prompted them.

KPIsAppointment booking rate
Triage1

New ideas Scout found.

Ready to experiment1

Built and tested. Waiting for you.

Collecting data0

Experiment running on live traffic.

Drag a card here to start its experiment.

Track every change

See what Scout would change.

Compare the draft with the live version, then approve it before publishing.

Agent changesLive v12 → Draft v13
Policies 1 changedTools 0Code 0
Scenario / Service Appointment
1.

Check availability with find_available_appointments.

−

If a time is open: offer an available appointment time.

+

If times are open: offer up to three, so the customer can pick.

If nothing fits: follow the scenario “Waitlist”.

Scout’s change · acceptedAccepting doesn’t publish it.
Experiments

Test the change. Measure the outcome.

Test each new version on a slice of traffic to verify the improvements.

ExperimentsOffer a choice of timesComplete
Control · 50%Live v1264.1%
Variant · 50%Draft v1370.8%
Measure
Appointment booking rate · target 75% · 2,360 conversations
+6.7 pts · guardrails held
Version history

Learn from each new version.

See how your agents improve.

KPIsAppointment booking rateTarget 75%
Live 64.1%
Since +12.8 pts
Experiments3 2 adopted · 1 ready
AI agents for customer support

See what Giga would
resolve at your company.

AI agents for email, chat, and voice.
Enter your work email to see what they’d resolve.