A setup includes a written plan, connecting only the accounts the job needs, approval rules, instructions, testing on your real work, and a handoff with plain-language notes.
A serious AI agent setup is not an install file and a login. It is a short diagnosis, a written plan, a limited first workflow, the connections and permissions to run that workflow, a test you can see, and a handoff so you know how to supervise it. We start with a free 30-minute discovery call and quote paid work after that call, before anything starts. This page explains what that engagement should include, and what it should not. We do not publish prices here.
Why start with the job, not the tool?
People often ask us to “set up AI” the way they would ask someone to set up a new laptop. An agent is closer to hiring a junior teammate for one job. If you cannot name the job, you are not ready to connect it to email or a calendar.
A good first job is narrow and visible. Examples we hear often: draft replies to common inbound email, propose times for a meeting, send you a morning summary of what landed overnight, or capture after-hours questions and flag the ones that need a person. A poor first job is “handle the business” or “automate operations.” That is a wish, not a brief.
On the discovery call we ask what you use today for email, calendar, and to-dos, what you have already tried, and what you want help with first. The call is for diagnosis. We do not build a full setup on the call, and we do not hand you a finished agent at the end of 30 minutes.
What is in the written plan?
After the call you should get a plan you can accept or decline. At minimum it should say:
- The first workflow, in plain language, including what “done” looks like.
- The tools and accounts that will be involved, and which of them you already have.
- The agent tool we recommend, and why — or a recommendation to stay with a tool you already use. We work with leading agent tools, including ones we have set up such as OpenClaw, Hermes Agent, and Grok Bot. We do not push a house platform.
- What the agent may do on its own, and what it must pause for.
- What we will deliver, what you will need to provide (access, time, a decision-maker), and what is out of scope.
- A quote for that scope, before work starts.
If a plan skips the approval rules or treats “we will figure out security later” as a phase, it is not complete. Connecting an agent to a live inbox without those rules is how drafts become accidental sends.
What does the setup work include?
Once you approve the plan, the setup work usually looks like this.
Access and inventory. We confirm the accounts, devices, and chat apps in scope. We use the least access that can do the job. We do not need, and should not be given, a dump of unrelated passwords or customer files.
Install or provision. That may mean software on a machine you control, or a hosted teammate, depending on the recommendation. Official install steps belong to the vendor; our job is to get a stable, supervised install that matches the plan.
Connections. We connect only the tools required for the first workflow. A calendar connection is not a reason to also connect a bank, a CRM, and a social account “while we are in there.”
Permissions and voice. We set who can talk to the agent, what it may read, what it may write, and when it must ask. We also set the tone of drafts so they sound like your business, not like a generic assistant.
Testing you can see. We run the workflow with you. You should see a real example: a draft sitting in the right place, a proposed time, a summary you can check. If the only test is “it installed,” the job is not done.
Handoff. You should leave with a short explanation of how to talk to the agent, how to approve or deny a step, how to pause it, and who to contact if it behaves badly. You should also know what we will not do unless you hire us for more work — for example, adding a second workflow, or a new channel.
What should not be included unless you ask?
A first setup should not quietly expand into a rebuild of the business. It should not require new hardware as a default. It should not depend on a monthly “care plan” you did not agree to. It should not include made-up benchmarks, a guaranteed number of hours saved, or a promise that the agent will work without supervision.
It also should not include a live send to customers on day one unless that was explicit in the plan and the approval path is in place. We would rather start with drafts and internal alerts.
Pricing stays off this page. Scope changes the quote. If you want a second workflow, a second location, or on-site time in Los Angeles, that belongs in the plan as a clear addition, not as a surprise.
Is setup remote or on-site?
Most of this work can be done remotely, anywhere in the United States. In Greater Los Angeles we can also work on-site when being in the room helps — for example, when we need to sit with you at the actual desk, see the messy real workflow, and walk through the handoff in person. On-site can be at your office or your home, for setup and walkthrough. If you are in Los Angeles and weighing the two, read Remote vs on-site AI setup in Los Angeles on the LA site.
What happens after the setup?
An agent is not finished the way a logo is finished. Models change, chat apps update, and your workflow will drift. You should expect to look at what it is doing, especially in the first weeks, and to correct the brief when it is wrong. If you later want a second job automated, that is a new, quoted piece of work, not an unspoken extra.
If you are still earlier in the process and want the everyday version of this, start with AI agents for small business inbox and scheduling and Securing an AI agent: basics. When you are ready to talk through your own stack, book a free 30-minute discovery call.
Questions
Is the discovery call part of the paid setup?
No. The discovery call is free and 30 minutes. We use it to understand the problem. Setup after the call is paid, and we quote it before work starts.
Do you publish a price list?
No. Cost depends on the scope. The plan you get after the call includes a quote for that scope. This page is about what the work should include, not what it costs.
Will I get a finished agent on the first call?
No. The call is for diagnosis. We do not build or hand over a full setup on the call.
What do I need to provide?
A decision-maker, a clear first job, and access to the tools in scope — typically the email, calendar, or chat app the workflow uses. You do not need to be technical. You do need to be available to review drafts and approval rules.
What if I already have ChatGPT, Claude, or Copilot?
Tell us. We start from what you already use. Sometimes the right setup is to get a current tool working on a real workflow. Sometimes it is to add an agent such as OpenClaw, Hermes Agent, or Grok Bot. We recommend from the job, not from a product we sell.