AI Workspace Setup

    Claude Projects Setup, ChatGPT Plugins and Gemini Skills, Built for Your Team

    For teams with AI seats and blank chat boxes. We build, test and hand over the Claude Projects, ChatGPT plugins and Gemini skills your team needs, in your own accounts, in two to four weeks.

    Scope my workspace Take the AI readiness assessment

    Why teams pay for AI workspace setup

    Buying seats is the easy part. Section's AI Proficiency Report 2026 found that 85% of knowledge workers have no value-driving AI use case, and a blank chat box doesn't give them one. A well-built instance does: it already knows your products, rules and formats, so the person using it starts from a solid draft instead of an empty page.

    Most of the market for this work is gig sellers offering a quick template, or enterprise vendors with large minimums. Operator Intelligence (OI) sits between them: a fixed-scope build for lean teams, done in your accounts, tested against real tasks and documented so your people can maintain it.

    Claude Projects setup, ChatGPT plugins and Gemini skills: three setup tiers

    Tiers are defined by scope, not seat count. We confirm scope and price in a written proposal.

    TierScopeGood fit forTypical timeline
    StarterAbout 3 configured instances for one team, with instructions, curated knowledge and a short user guideA single team proving the valueAbout 2 weeks
    Team5–8 instances across a department, plus Skills and a shared team prompt libraryDepartments ready to standardize how they use AI2–3 weeks
    OpsTeam scope plus connectors and MCP, governance, admin settings and a maintenance planCompanies connecting AI to live systems and data3–4 weeks

    How AI workspace setup works

    Seven steps take you from discovery to handover, with a short working session with your owner each week.

    1. Discovery

      We interview the team, review your plans and tools, and pick the workflows each instance will serve, with a definition of good output for each.

    2. Knowledge curation

      We gather the documents each instance needs, remove what is outdated or sensitive and structure the rest so the model can use it.

    3. System instructions

      We write and version the instructions: role, rules, formats, tone, what to decline and when to ask a clarifying question.

    4. Skills and connectors

      Where your plan supports them, we package repeatable procedures as Skills and connect approved tools with least-privilege permissions.

    5. Testing on real tasks

      Each instance runs a test set of real tasks from your team and must pass a review rubric before handover. We fix what fails.

    6. Handover training

      We train users and the owner on how to use each instance, check its output and update it.

    7. Optional maintenance

      A monthly review keeps instances current as your documents, plans and tools change.

    Claude Projects vs ChatGPT plugins vs Gemini skills

    All three hold instructions and reference knowledge for repeat work. The right one depends on the platform your team already uses. We write those instructions and that knowledge as portable, SKILL.md-style assets, so they move between ChatGPT, Claude and Gemini with little rework. Plan features and limits change often, so we confirm details against your current plan during discovery.

    Claude ProjectsChatGPT plugins and skillsGemini skills
    What it isA shared Claude workspace with project instructions and a knowledge baseA bundle of one or more skills (reusable instructions plus resources) and, optionally, an MCP server for tools; the successor to custom GPTsA reusable Gemini prompt in the open SKILL.md format, invoked with "/"; the replacement for Gems
    Strongest atWriting and analysis grounded in rich reference materialWide distribution across a ChatGPT workspace and calling external toolsWork inside Google Workspace: Gmail, Docs, Sheets and Drive
    Pairs withSkills, connectors and MCP, CoworkChatGPT Projects, connectors and MCP serversGemini in Workspace apps and NotebookLM
    Good fit forTeams that work from long documents and strict rulesCompanies standardized on ChatGPTCompanies that run on Google Workspace
    Watch forKeeping knowledge files current as documents changeSprawl: overlapping plugins with no owner, and old GPTs left unmigratedSharing settings that depend on your Workspace admin

    Which platform should you choose?

    Start with the platform your team already pays for and opens every day, because adoption beats feature lists. Many clients run two: Claude for writing, analysis and connected work, plus ChatGPT or Gemini where another team has already standardized. If you're choosing from scratch, our page on Claude training for teams explains what Claude adds with Skills, connectors and Cowork.

    Claude is a trademark of Anthropic, ChatGPT of OpenAI and Gemini of Google. Operator Intelligence is an independent consultancy and is not affiliated with any of them.

    What we need from you

    • A named owner who will maintain the instances after handover
    • Admin access, or an admin available during setup sessions
    • Three to five priority workflows, with examples of good output
    • Source documents for each workflow, with sensitive data flagged
    • A reviewer per instance to score test results
    • About two hours a week from the owner during the build
    • For the Ops tier: an IT or security contact to approve connectors

    How we measure a workspace setup

    A setup is finished when it works, not when it is delivered. Before handover, each instance must pass its test set of real tasks. After handover we track the same signals at 30, 60 and 90 days: active users per instance, how often each one is used, which workflows moved to it and what the team changed or asked for. Instances nobody uses get fixed or retired. Companies that want those habits in every team can extend the setup into an AI champions program.

    That readout shapes the next step, whether that's a Team AI Bootcamp for broader skills or custom AI agents for workflows that should run on their own.

    From configured instances to connected systems

    The Ops tier is where setup meets your systems. Connectors and MCP let an instance read from and act in your tools, so permissions and review gates matter as much as the prompt. Operator Intelligence runs this pattern in production: for ROAM Luggage, requested site changes are executed with Claude via MCP inside a Shopify development theme, and a team member reviews and publishes. Critical site updates went from days to minutes. See the ROAM Luggage case study.

    Frequently asked questions

    Who owns the instances you build?

    You do. Every Project, plugin and skill is built in your accounts, and the instructions, knowledge files and documentation belong to your company. If you stop working with us, everything keeps working.

    Do you need access to our data?

    Only to the documents each instance needs, and only inside your own accounts. We flag sensitive material during knowledge curation and follow your AI acceptable use policy, or help you adopt one.

    Can you set up Claude Team, Cowork and MCP connectors?

    Yes, as part of the Team and Ops tiers, where your plan supports them. We configure admin settings, approved connectors and permissions with your admin and document every decision. Plan features change; we configure to your current plan. Teams that want to build with Claude Code itself can add Claude Code training.

    How is this different from hiring a freelancer to build a ChatGPT plugin?

    A gig build usually delivers one template. We start from your workflows, test each instance on real tasks, train your team and measure use for 90 days. The result is a maintained system, not a one-off file.

    What does AI workspace setup cost?

    Cost depends on the tier, the number of instances, the platforms involved and whether connectors and governance are in scope. We don't publish fixed prices; see how we price and request a scoped proposal.

    What happens when the tools change?

    They will, and two just did. OpenAI is retiring custom GPTs in favor of ChatGPT plugins (retirement is scheduled for December 11, 2026, and dates vary by plan), and Google is replacing Gems with skills; on business accounts, Gems can no longer be created, edited or used from no sooner than March 1, 2027. We migrate existing GPTs and Gems to the new formats, and optional monthly maintenance keeps instructions, knowledge and skills current. Without maintenance, your owner has the documentation to do it.

    Scope your AI workspace

    Tell us which team, which tools and which workflows. We'll recommend a tier and send a scoped proposal.

    Scope my workspace

    Sources

    1. OpenAI Help Center, Custom GPT retirement and migration FAQ (2026): OpenAI is retiring custom GPTs and replacing them with ChatGPT plugins, which bundle skills (reusable instructions plus resources) and/or an MCP server for tools; retirement is scheduled for December 11, 2026, dates may vary by plan, and some Enterprise deferrals run to February 2027
    2. Google Workspace Updates blog (2026): Google is replacing Gemini Gems with skills, reusable prompts in the open SKILL.md format invoked with "/"; on business accounts, Gems can no longer be created, edited or used from no sooner than March 1, 2027
    3. Google Workspace, Gems migration guide (2026): On business accounts, Gems can no longer be created, edited or used from no sooner than March 1, 2027

    Related

    Custom GPT & Gems migration guide

    Move custom GPTs to plugins and Gems to skills before the deadlines.

    Claude Training for Business Teams

    Projects, Skills, connectors, Cowork and Claude Code.

    Team AI Bootcamp

    Hands-on skills for the people using your instances.

    AI Training & Fluency

    All programs, from executive workshops to bootcamps.

    Custom AI Agent Development

    When a workflow should run on its own.

    AI Acceptable Use Policy Template

    Data rules to set before you connect anything.

    ChatGPT & Gemini training

    ChatGPT plugins, Gemini skills and prompt libraries.