Your team has Claude. The next 30 days decide whether it becomes a daily tool or a renewal nobody can explain. Confirm the plan and data settings in week one, build shared Projects and connect your files in week two, give every role named tasks in week three, and measure in week four.
SafeLab AI Engineering runs on-site Claude training for small business teams in Orange County, Los Angeles and San Diego. One 4-hour hands-on session costs $1,250 plus $125 per attendee (10 people is $2,500), and our team includes Claude Certified Associates and Developers. This is the plan we walk clients through, and you can run most of it yourself.
If you are still choosing between tools, start with ChatGPT vs Copilot vs Claude for a small business team and come back once you have picked.
Day 1: Know Which Claude Plan You Are On
Claude for business comes in two plans, and the one you bought shapes what the rollout looks like. These are Anthropic's published prices on claude.com/pricing as of November 2026.
| Plan | Price per seat per month | Seats | Worth knowing |
|---|---|---|---|
| Team, Standard seat | $20 billed annually, $25 billed monthly | 2 to 150 | All Claude features, more usage than Pro |
| Team, Premium seat | $100 billed annually, $125 billed monthly | Mix with Standard | Five times the usage of a Standard seat |
| Enterprise | $20 billed annually, plus usage billed separately | Minimum 20 | Adds role-based access, SCIM, audit logs, custom data retention |
For most teams of 5 to 50, Team with Standard seats is the right starting point. Hold Premium seats back until someone actually runs into limits, usually the person running long Cowork tasks or heavy document work.
What a Business Plan Gives You
Four features matter most in the first month. Each one is listed under Team on Anthropic's pricing page.
- Projects. A Project holds standing instructions and reference files, and Team plans can share Projects across the team. A "Client Emails" Project with your tone guide and three good examples means nobody starts from a blank chat.
- Connectors. The Google Workspace connectors cover Gmail, Google Calendar and Google Drive. The Microsoft 365 connector reaches SharePoint, OneDrive, Outlook and Teams using each person's own permissions. On Team and Enterprise, an Owner or Primary Owner has to add a connector for the organization before members can connect their accounts, and Microsoft 365 also needs a Microsoft Entra Global Administrator to grant consent.
- Claude Cowork. Cowork runs multi-step tasks on files and folders. Anthropic's Cowork guide for Team and Enterprise says it is on by default for Team organizations, and owners can turn it off in Organization settings.
- Claude in Chrome. The extension lets Claude read and work in web pages in the user's own browser, and Cowork can use it. This is how Claude reaches older web software that has no integration.
Data handling defaults
The pricing page states "No model training on your content by default" for Team, and Anthropic's privacy center says business chats and coding sessions are not used to train models. Two exceptions belong in your policy: conversations a user submits as thumbs up or down feedback may be used for training, and so may data from organizations that opt in to the Development Partner Program. Your contract is Anthropic's Commercial Terms.
If you handle health information, read the plan table closely. Anthropic lists its HIPAA offering under Enterprise only. On Team, keep protected health information out and design your workflows to be HIPAA-safe from the start.
The 30-Day Plan
Week 1: Set up and write the rules
The account owner does four things this week:
- Check the seat mix. Everyone on Standard to begin with, Premium only where you already know usage will be heavy.
- Turn on the connectors you need, and only those. If the team lives in Google Workspace, add Gmail, Calendar and Drive. If it lives in Microsoft 365, add that connector and check which write actions (sending mail, editing files) you want allowed.
- Publish a one-page policy. Name Claude as the approved tool, list the data that never goes in, and say when a person must review output before it leaves the building. Our AI usage policy guide has a template.
- Pick a Claude lead. One person who answers questions and keeps the shared Projects tidy.
Week 2: Build shared Projects and run a kickoff
Create one shared Project per team or recurring job: client emails, proposals, HR letters, weekly reports. Put the instructions and two or three good examples in each. Then run a kickoff where every person completes one real task in the room, on real (non-sensitive) work, using the Project built for it. The full agenda is in how to train your employees on AI.
Week 3: Give every role named tasks
Vague encouragement rarely changes habits. This week, each person moves one or two recurring tasks into Claude for the whole week and notes what worked.
| Role | Named task for week 3 | Feature to use |
|---|---|---|
| Owner or manager | Turn the week's email threads into a Friday status summary | Gmail or Outlook connector |
| Operations | Build a checklist or SOP from a voice memo or messy notes | Shared Project with the SOP format |
| Sales | Draft follow-up emails after each call, then edit and send | Project with tone guide and examples |
| Office or HR | Draft offer letters and reminders from approved templates | Project plus Drive or SharePoint files |
| Finance or admin | Reconcile two exports and list the mismatches for review | Cowork on a folder of files |
| Customer service | Draft replies to common questions using your policy documents | Project with the policy files attached |
Every output in this table is a draft. A person reviews and sends it.
Week 4: Try Cowork on one real workflow, then review
Pick one workflow that eats hours every week and involves several files, and build a Cowork task for it with the person who does the job today. Keep the rules tight: what Claude may read, what it must never change, and that nothing sends itself. Then hold a 30-minute review: collect wins, fix confusion, and retire any Project nobody opened.
A Real Example: An Orange County Home Care Agency
We trained the office team of an Orange County home care agency and set them up on Claude Cowork. The schedule lives in decades-old software with no integration options, so staff export three files each week and Claude reads the calendar through Claude in Chrome, strictly read-only. Open-shift scheduling dropped from about 5 minutes to 39 seconds per household, 87% less time. HR email drafting takes 75% less time, and every draft still waits in Gmail for a person to review and send.
The staff timed both numbers themselves on real work, and the rules (read only, nothing sends itself, no client health details in drafts) were written down before anything was switched on. The full write-up is in the home care Claude Cowork case study.
The lesson for your 30 days: the result came from one carefully chosen workflow, built with the person who owns it.
Numbers to Check at Day 30
You need a short list of honest counts. The usage figures are in your admin console, where Team plans include usage analytics.
| Signal | Healthy at day 30 | If it is low |
|---|---|---|
| People using Claude in a typical week | More than half the team | Revisit the named tasks per role |
| Named tasks moved into Claude | At least one per person | Assign them in the next team meeting |
| Shared Projects opened in the last week | Most of the ones you built | Merge or delete the unused ones |
| One timed task, before and after | A measurable drop in minutes | Rebuild the instructions with the person who does the work |
| Restricted data that went in | Zero | Retrain on the policy that week |
Time at least one task before and after. A single measured result, like the 39 seconds above, does more for adoption than any survey.
When to Bring In a Trainer
You can run this plan yourself. A trainer earns the cost when one of these is true:
- Week 3 stalls. People cannot name a task, or every task they try is a one-off question.
- Nobody has time to own it. The plan needs a lead, and the lead has a full-time job already.
- You want Cowork on a real workflow and do not want to learn its permissions and limits by trial and error on live data.
- Your data has rules. Client records, health information or HR files need a clear line on what stays out.
SafeLab AI Engineering trains teams on Claude on-site at your office across Orange County (Irvine, Newport Beach, Costa Mesa, Santa Ana, Anaheim and the rest of the county), Los Angeles and San Diego. The session is 4 hands-on hours, staff work on their own real tasks, and staff interviews before the session are included. We leave behind prompts, Project instructions and short guides, plus safe data habits for client information. The price is $1,250 per session plus $125 per attendee for attendees 1 to 10, $100 each for 11 to 25, and $75 each from attendee 26, with a 5-attendee minimum: 5 people is $1,875, 10 is $2,500, and 20 is $3,500. Sessions held on or before December 31, 2026 are 20% off.
Our team includes Claude Certified Associates and Developers. See Claude training for teams, or compare all three tools on the AI training page. California employers may also be able to get training costs reimbursed through the state's Employment Training Panel, which requires at least 8 hours of training per trainee under its published rules, so a funded program is planned as two or more sessions.
The Bottom Line
Claude Team gives a small business shared Projects, connectors to the files and email you already use, and Cowork for multi-step work. None of it shows up in daily work without a plan. Set the rules in week one, build Projects in week two, name tasks in week three, and measure in week four. If the team uses ChatGPT too, the same structure works there: see you've purchased ChatGPT for your team, now what.

