AI Agent Setup and Support for Your Business

We set up named AI agents around an agreed business task. Your purchase covers the configuration, connected accounts, test cases and staff handoff in your chosen plan. Tell us what you want the agent to finish. Jeff will help you decide which work belongs in the first setup and which work needs a separate quote.

Phone, text, email, model usage, paid advertising and other provider charges are extra. We agree those accounts and spending limits before launch. Additional support costs 150 dollars an hour when you approve it. A role name doesn't grant access to an account or permission to spend money.

1. Choose the task and a likely plan. Start with the role cards below. A short description of the work is enough for our first conversation.
2. Call us about the setup. We confirm the accounts, approvals and operating budget, then agree a written scope with you.
3. Arrange payment by phone. Our existing payment process follows your approval. Onboarding begins once payment is confirmed.

Talk through your agent setup

You can discuss the first task with us at no cost. During business hours, our reply aim for your Trojan inquiry is five minutes. If calling doesn't suit you, leave a callback request and tell us which role caught your attention.

Caesar, Nova, Hailey, Penny, Atlas and Mercury describe different setup services. We use the names to make the choice easier. The practical question is what each role may read, prepare or change for your business. Two agents with the same name can have very different permissions because their clients need different work done.

You can also ask us about Jamie, our website assistant. A client setup can give a similar role an approved set of public business answers and an inquiry handoff. Jamie on this site helps explain Trojan's services and prepare a callback request. It doesn't buy a plan, collect payment or confirm an appointment.

We'd rather begin with an exact task than squeeze your business into every role. For example, an office may need Penny to prepare a draft from a job record. Adding Mercury won't improve that invoice. If the next problem is campaign spending, we can discuss that as a separate piece of work.

Read our AI and automation service page when you want to compare broader workflow options. This hub explains the named setups and purchase choices. The broader service page covers the kinds of systems and recurring work we can discuss with you.

A plan needs plain units so you can tell when a change adds work. We agree one business and its approved service areas before setup. Focus covers one area, Team covers up to two, and Operations covers up to three. An area unit covers one county or a list containing no more than ten ZIP codes.

The territory describes the information and operating rules we configure. It doesn't promise demand, complete weather coverage or search visibility in every location. A business that serves several disconnected areas may need different routing or notices. We'll discuss that layout before assigning its plan.

Hosting and the operating model are additional provider services. They don't become free because they're outside the connected business account count. We can use client-owned accounts billed directly, or agree documented provider charges passed through at cost. You'll know which arrangement applies before the work begins.

We include one training session within the plan's team size. Focus allows up to three people, Team allows five, and Operations allows ten. Choose people who will actually handle the output or change a setting. Training a larger group or several shifts belongs in the written agreement.

Operations can begin with 150 knowledge pages or files. Team's initial allowance is 75; Focus includes 25. Each needs to be text-ready and no larger than five megabytes. A scanned archive needing transcription or a large document collection needs its own estimate. We won't pretend that uploading a folder settles its accuracy.

The strongest starting point is a piece of work your staff can describe today. We ask who does it, what they need to know and where it gets stuck. A sentence such as "our office checks the same invoice fields each morning" gives us a better first discussion than a request to automate the whole company.

Jeff will ask what a correct output looks like. You might show us an approved invoice draft, a campaign brief or the wording staff uses to answer a common service question. Remove private customer details from a first example. Our initial call is for choosing the work and the route into a proper setup.

We also ask what should stop the task. Missing job details, an expired weather notice or a conflicting price can change the answer. Your staff may already know the rule but keep it in their heads. Getting that rule into the agreed setup is part of making the agent usable.

A staff owner matters after launch. Someone needs to correct a source when your business changes and decide whether a failed output is safe to retry. That person doesn't need to become a programmer. They need a clear route to pause the work and contact us when a change exceeds their authority.

1. Name the current job. We want to hear the task in your team's language, including the part that causes repeated checking.
2. Show a safe example. An approved, redacted input and output help us see the required fields without exposing a customer's record.
3. Choose its staff owner. Decide who receives the result, handles exceptions and approves any action beyond a draft.
4. Identify the accounts. List the services involved and who can grant access. Avoid sending passwords through a website conversation.
5. Agree the stop rule. Describe a missing fact or failed connection that should send the task back to a person.

Our first setup can stay narrow even when your long-term idea is larger. A team might start by preparing a daily list of unresolved requests. Connecting outbound messages comes later, after the team trusts the list and has agreed who may send them. We'll discuss that progression without promising an automatic expansion.

If your current process is inconsistent, we may need to settle its steps before configuring an agent. That can reveal a simpler fix. A clearer form, a staff checklist or a website edit may remove the problem on its own. We can discuss those options through our existing web and automation services.

Our setup and monthly fees pay for the agreed Trojan work. Connected services can charge for use. We keep those costs visible because a busy month, a repeated failure or a longer conversation can change the bill. You should understand the operating budget before staff relies on the workflow.

Phone work may involve a number rental, incoming and transfer minutes, transcription and speech services. Text messages may have segment and carrier charges. Email tools can have subscriptions or delivery limits. The provider and the enabled modules determine those costs, so we confirm them for your actual accounts.

A plan doesn't automatically include voice, SMS and email together. Optional channel ceilings apply only when the approved workflow uses that channel. Focus can have ceilings of 300 voice minutes, 500 text segments and 5,000 email deliveries. Team's ceilings are 900 minutes, 1,500 segments and 15,000 deliveries.

Operations can have ceilings of 2,000 voice minutes, 3,000 text segments and 25,000 email deliveries. These are operating limits to agree for the relevant modules. They aren't prepaid provider balances or promises that every setup sends that volume. A wider allowance needs your approval and an updated operating budget.

For an agreed website conversation workflow, one run covers up to eight visitor questions. Voice work has a ten-minute call limit and a human handoff; each call counts as one run. It uses one incoming number and one transfer destination unless we agree a wider arrangement.

We set the supported spending controls for your chosen services. An alert marks 80 percent of your approved budget. At the ceiling, optional outbound work pauses or routes to your named contact. Provider reporting can arrive late. We retain a safety buffer and explain the actual controls each provider supports.

Weather or source polling without a matching event sits outside the processed run count. It still belongs within the approved API budget. Source checks consume provider resources, and your account's pricing determines the charges. Our agreement should say how often a check is needed and what happens if the source is unavailable.

Paid advertising needs a separate media budget. Mercury or Hailey may prepare a campaign recommendation without permission to launch it. If you want approved account actions, we define them with the spend controls. Creative production and ongoing full campaign management remain separate work to quote.

We prefer client-owned provider accounts where the services allow them. Your business can see the billing and decide who has access. During setup, we identify the account owner and the permissions needed for each task. A general administrator login shouldn't be the default answer to every connection.

An account may let us inspect a campaign without letting us alter it. An invoice draft can also need separate authority before anyone issues it. We spell out those distinctions in the work you approve. If you only want drafts, the configuration should make that limit clear to staff and to the connected tools.

Keep customer records inside the approved project process. Our public website assistant uses public Trojan material. A client agent connected to private information needs its own access plan and source rules. Sending a confidential file to the public chat doesn't create a private client workspace.

Atlas can help you examine local deployment, but the word "local" needs a practical explanation. The device may run the agent software while a remote model handles an answer. We identify each connection and what information it receives. Your hosting choice alone cannot settle every data question.

We document the account and configuration handoff for the work we set up. Your agreement needs to address any provider restrictions, custom development and retained support access. When monthly support ends, managed scheduling and maintenance end with it. Your staff should know which tasks then need another owner.

Don't share a single password across the whole team just to make onboarding quick. Where supported, we use named access and the required permissions. If a tool needs a different approach, we discuss the tradeoff with the account owner. Changing an employee's role should include checking their agent access.

Public web content can contain instructions that conflict with your own rules. A source page should supply information, while your agreed permissions decide what the agent can do. We can test attempts to redirect a task, request extra access or hide an action inside a source document.

A person should remain able to pause the workflow. We agree where that control lives and who may use it. A failed connection or an unusual output needs a visible path back to staff. The setup should leave you able to decide whether the next run is safe.

We test representative tasks from the work you approve. An easy input is only part of that check. Your team also needs to see what happens when a price is missing, a service falls outside the area or a connected account is unavailable. These cases show whether the setup respects its boundaries.

For Penny, we might test a job record without a tax field or an invoice reference already used. The finance owner decides the required treatment. For Hailey, we might use an expired event or a location outside the agreed territory. The team should see why that notice doesn't trigger the normal response.

For Nova, a search report with no matching period should make the missing comparison clear. For Mercury, an instruction to increase spend beyond the allowance should reach the approval boundary. Those examples are planning cases. They don't describe measured results from a past client agent installation.

We check where the output appears. A technically correct draft can still land in an inbox nobody owns. Staff needs a recognizable label and enough source context to act. If a message requires a reply, the setup should show who receives that reply and what happens after their business hours.

The handoff also needs a way to report a bad answer without repeating private data in several places. We agree which record or task reference staff can send us. That lets us investigate the behavior while respecting the access arrangement. A screenshot of an error may need redaction before sharing.

We don't promise a universal launch date from a plan card. Account approval, source quality and the work you choose affect the sequence. Our call identifies what must happen before onboarding and which client decisions could hold it up. We'll confirm the actual work and timing in your agreement.

Once the approved workflows go live, monthly support covers operating checks and corrections within your allowance. A new output channel or a substantially different task can need another estimate. We explain that boundary so a small change request doesn't become an open-ended project without your approval.

Our existing project work gives you something concrete to discuss on the call. A visitor needs to understand an input, identify the action and know where the result goes. Those details matter for an agent handoff too. We can examine them without pretending a website screenshot is a client agent deployment.

The Jev website scanner asks for a site address before the scan begins. For an agent setup, we use the same plain question: what input lets the user start the intended job? A button shouldn't imply that an invoice was sent or a campaign launched if it only prepared a draft.

We also discuss labels for incomplete work. If the output is waiting for a manager, staff should be able to tell that apart from a completed action. A status can point to the person who needs to act. It should stay understandable even when a model's explanation is long or uncertain.

In the Texas Journeyman preparation project, the calculator presents labeled fields for a defined task. Your agent may need a job reference, a source date or an approved account. We identify those inputs from your process rather than asking users to type a long instruction each time.

People also need a correction route. A user may spot a wrong number after entering it, or a manager may return a draft with a question. We discuss how that correction travels through the workflow. Repeating the entire task can cause extra runs or repeated actions if the design doesn't distinguish an amendment.

Keyboard access, clear focus and readable status text belong in the interface discussion. We preserve the distinction between a navigation link and an action button. A client setup may use an existing business app rather than a custom website screen. We'll agree the interface that staff can actually operate.

Jamie is the assistant on Trojan's website. It can help you explore published services, discuss a suitable next step and prepare callback preferences. You can ask about a role or compare the plans on this page. The final decision about your client setup belongs in the conversation with us.

Our website conversation is separate from a configured agent connected to your business. Jamie doesn't have your office records or access to your advertising account. Describing a task here can help us understand your inquiry. It doesn't authorize a campaign action, transfer funds or start a paid setup.

If the live AI response is unavailable, the website can still offer a source-based reply. Our existing contact route remains available too. We want the next step to stay usable when one connection fails. You don't need the assistant to read this page, choose a plan or call Trojan.

When you choose the callback path, Jamie can help carry a service choice and planning note to the contact form. You inspect those details and enter your own contact information before sending. Staff confirms availability and the preferred callback time. The conversation itself doesn't reserve a place on a calendar.

For a similar setup on your business website, we would agree the facts it may answer and the next action it may offer. A service list, prices and business hours need an update owner. A conversation should be clear when a request needs human confirmation or falls outside the approved information.

We'd also decide which details belong in a callback note. Collecting a customer's preferred time can be helpful. Collecting private financial or medical information in a general service chat can create a different problem. Your source rules and inquiry fields should match the actual business need.

Jamie may guide a buyer toward Focus, Team or a wider conversation, but the written agreement sets what we sell. If your question involves private deployment, hardware or several approvals, we need to talk it through. The public answer shouldn't hide that extra work behind a simple monthly price.

You can build your own business agent if you have the time and the right person to maintain it. Our business AI agent guide walks through a small first task and the choices behind a working setup. It can help you decide what you want to handle yourself.

Buying our setup gives you an agreed piece of configuration and support work. It doesn't remove your responsibility for approved business facts or staff decisions. We need your account owner and task owner involved. An agent can only follow a reliable process if the business has decided what that process is.

DIY can be sensible for a draft-only task using public information. A workflow that issues invoices, sends customer messages or changes account spending needs a closer permissions check. Consider what happens when the person who built it is away. Someone else needs to know how to pause and maintain the work.

For a weather-driven contractor idea, read our weather response agent guide. It explains the path from a source notice to a staff action. Pair it with Hailey's service page when you want us to configure that path using your approved territories and accounts.

Atlas is the relevant role when your main question is local or private deployment. Nova fits the search task itself. Choosing the runtime and choosing the work are separate parts of the same purchase. We can talk through both without suggesting that one provider or machine suits every business.

Our human approval guide is useful when your team needs to decide where a person should intervene. That existing guide deals with approval in recurring work. The new DIY guide focuses on building the agent, while this hub helps you choose and buy our setup services.

Bring any existing automation to the first call. We may be able to work with it, or it may impose limits on the new task. We need to know which service owns its schedule and what a second system would change. Adding another tool without checking the first can create duplicate notices or conflicting actions.

If none of the role descriptions fits, tell us about the work rather than choosing the nearest name. Custom development, difficult account connections or model adaptation need a separate quote. Our web application service explains the broader development conversation when an agent setup alone won't cover it.

Ongoing support should have an owner and a boundary. We include operating checks, corrections and small changes within your monthly hours. New workflows, account migrations, substantial data cleanup and special model work need another estimate. We'll ask before treating those changes as chargeable extra work.

Your business facts will change. A new price, service area or staff contact can affect several answers. Decide who tells us about those changes and which records they update. If one old file remains available, the agent may continue to find it. We need a clear rule for retiring outdated information.

The support fee starts once your agreed workflows are live. Renewal then follows a monthly schedule. The written agreement confirms payment and cancellation terms before you commit. A cancellation should be raised before the next renewal. We'll identify the handoff of your accounts and configuration and the managed work that ends with support.

If an agent runs on a local device, your team also needs a plan for that device. Power, updates, backup and staff access can affect availability. Hardware and special device support are separate costs to agree. We won't assume a computer left in an office has an owner just because it can run the software.

Keep the original account ownership visible. If a staff member leaves, remove their access through the provider's supported controls and check the workflow destination. An agent still sending notices to an unused inbox can look active while nobody receives the work. We include that sort of operating question in the agreed checks.

You can choose a smaller ongoing arrangement than your eventual vision. We can discuss which approved workflows are actually used before expanding. More agents or runs are worthwhile only when their tasks help your staff. The purpose of the plan is to make the work and its costs understandable.

Call us with the task you want to improve, the likely role and any account limits you already know. We'll help you choose the next step, confirm the work and arrange payment through our existing process. You can read every role and guide without opening the website assistant.

Discuss the work you want finished