Jardine Studio
SERVICE / WEB DEVELOPMENT / AI AGENTS

Voice and chat AI agents wired into the systems your team already uses.

An AI agent that actually works is wired into the systems the team already runs, not bolted onto the corner of the page. Voice agents on OpenAI, Anthropic, ElevenLabs, Vapi, or Retell that answer calls, book appointments, and write back to the CRM. Chat agents that handle inquiry and intake. Knowledge bases the team can ask in plain language.

A FEW BUSINESSES THE STUDIO HAS WORKED WITH
Palisades Lodge of Big Pine logo
The Black Salt Room logo
BLIZZARDFIRE PROTECTION
Bodie Foundation logo
Night Rose Deathcare logo
WHERE IT BREAKS

The team is doing agent work by hand because the agent does not exist yet.

Most agent engagements start the same way. The team has the repeat-work pattern, has the script in their head, has the integration list, and has been pricing the agent off and on for six months. The work stays manual.

  • After-hours calls go to voicemail or walk to a competitor.

    Hospitality, wellness, and service businesses lose a measurable share of inquiries to the team not being on at 9 PM or on weekends.

  • Inbox triage is the team's actual job most days.

    The marketing lead spends two hours a day classifying inquiries that an agent could route in seconds.

  • The intake form takes a week to update.

    Conditional questions, file uploads, lead qualification logic. Each one is a feature put off because there is no clear owner.

  • The booking flow is half automated and half email.

    The widget takes the date. The team still has to reply to confirm special requests by email.

  • The team has tried a chatbot widget and uninstalled it.

    Off-the-shelf chat widgets do not know the business. They route to FAQ pages the team did not write.

WHAT IT COSTS

Manual triage is paying someone for repeat work.

The cost is rarely a single line. It is the missed inquiries, the team hours, and the customer experience that drifts on the wrong side of expectations.

  • Hours of repetitive work the agent should be running.

    Triage, qualification, scheduling, follow-up. The team is doing the work the agent was built to do.

  • Inquiries the team cannot answer fast enough to close.

    Response-time is the conversion signal in most service categories.

  • Calls missed outside business hours that walk to a competitor.

    Hospitality and wellness in particular see a measurable share of evening and weekend inquiries the agent would have captured.

  • A team bottlenecked on inbox triage instead of customer work.

    The expensive members of the team are spending hours on work the agent is built for.

HOW WE DO IT

Five moves that ship an agent the team actually uses.

The studio scopes the use case before picking the model, picks the integrations before designing the UX, and ships monitoring before the agent goes live to customers.

  • Use case scoping written before code.

    Voice or chat. Inbound or outbound. The exact decisions the agent makes, the data it reads from, the systems it writes to.

  • Model and provider picked deliberately.

    OpenAI for general reasoning. Anthropic for long-context. ElevenLabs for voice synthesis. Vapi or Retell for orchestrating voice agents.

  • Integrations to the systems the team already runs.

    HubSpot, Salesforce, Pipedrive for CRM. Cal.com, Acuity, Mews for scheduling. Airtable, Notion, Google Sheets.

  • UX designed for the agent, not adapted from a chatbot widget.

    Voice agents have a natural-conversation script and a fallback path. Chat agents have a UX that explains scope.

  • Monitoring and guardrails before the agent talks to customers.

    Logging, fallback paths, escalation rules, and a human review surface for the first month after launch.

What would the agent handle that the team is doing by hand?

Bring the agent use case and the systems the team already runs. The first call shapes the rest.

WHAT YOU GET

Outcomes every agent engagement ships with.

Specific deliverables that hold across voice and chat, hospitality and wellness, inbound and outbound.

  • An agent that knows the business, not a chatbot widget on a FAQ.

    Trained on the team's actual content, scoped to the team's use case, integrated with the team's systems.

  • Integrated with the CRM, calendar, and email platform.

    No new SaaS to babysit. The agent reads and writes where the team works.

  • Monitored, with a human handoff path that works.

    Logging, escalation rules, and a review surface so the team can audit what the agent did.

THE ENGAGEMENT

How the work moves.

  1. Phase 1: Scope

    Use case, voice or chat, inbound or outbound, integration list, success criteria.

  2. Phase 2: Build

    Agent logic, system prompts, integrations to CRM, calendar, and email.

  3. Phase 3: Integration testing

    End-to-end runs through the agent path with real data and edge cases.

  4. Phase 4: Soft launch with monitoring

    Agent goes live to a subset of inquiries with logging and human review.

  5. Phase 5: Full launch and tune-up

    Agent goes live to all inquiries. First-month monitoring included.

QUESTIONS

Things worth knowing.

What is the difference between an AI agent and a chatbot widget?
An agent is scoped to a use case and integrated into the team's systems (CRM, scheduling, email). It reads from real data, writes to real systems, and makes decisions inside a defined boundary. A chatbot widget is a UI bolted onto the site that answers questions from a FAQ page. Agents are the work the studio builds.
Which models and providers does the studio build on?
OpenAI for general reasoning. Anthropic for long-context and tone-sensitive work. ElevenLabs for voice synthesis. Vapi or Retell for orchestrating voice agents. OpenPipe for fine-tuning where the use case warrants. Choice is driven by what the use case needs, not by what is fashionable.
Can the agent integrate with my existing CRM and scheduling tool?
Usually yes. The studio builds on HubSpot, Salesforce, Pipedrive, Cal.com, Acuity, Mews, Airtable, Notion, Google Sheets, Resend, Twilio, and Slack. The scope call confirms which systems the agent reads from and writes to.
What happens when the agent does not know the answer?
The fallback is built into the agent before launch. Voice agents transfer the call with context. Chat agents hand off to a tagged email with the conversation summary. Escalation rules are defined in the scope, not figured out at runtime.
Does the studio have a case study for AI agent work?
Not yet in-studio. The studio is shipping agents as part of current engagements and the first published case will land in 2026. The proof today is the named providers, the named integrations, and the agency-side AI and automation work the team did at Copper CRM, Standard Bots, and prior teams.
How much does an AI agent cost?
Scope drives the price. A single-channel chat agent on an existing site and a multi-channel voice plus chat agent integrated into the full CRM stack are different engagement shapes. The first call sizes the scope before the studio quotes.
ALSO HERE

An agent that knows the business, not a chatbot on a FAQ.

Bring the use case and the integrations. The first call settles whether voice, chat, or workflow is the right shape.