Already Paying for Claude? A 30-Day Fix When Nobody Uses It

1. The Short Answer: Start From Your Own Usage Data

If your firm already pays for Claude Team or Enterprise and few people use it, the fix takes 30 days and starts with the usage analytics you already have. Find out who is active and why the rest stopped, write down three workflows and one page of data rules, train each team on its own documents, and measure again on day 30.

Low usage looks like a training problem, so the usual response is another webinar. The real gap tends to be smaller: nobody has said which tasks Claude is for, or which documents may go into it.

For a firm with Claude but no standard processes, that is the implementation strategy. Write the standard first, train people to it, then measure it.

The month, one step per week:

Everything in the plan runs on what you already pay for. The scarce input is an owner with a few hours a week.

2. Why Paid Seats Sit Idle

Seats go idle for ordinary reasons, and the model is rarely one of them. Check these five first:

• No defined workflows. Telling people to use Claude more gives nobody a task for Monday.
• Fear about data. An associate unsure whether a CIM may go into Claude will keep it out. When nobody has written the rule down, silence reads as no.
• Generic training. A tool-neutral webinar summarizes a press release. Your team needed to watch its own teaser screened against its own criteria.
• No time. Learning a new tool competes with a live process, and the live process wins.
• The September change. Anthropic's announcement on September 16, 2026 that chat and Cowork are merging can leave people unsure which mode the firm expects.

The seat bill is the small number. Assume, for illustration, 30 Team standard seats billed annually: 30 x $20 x 12 = $7,200 a year. The bigger cost is the work your team still does the slow way.

Each reason on that list is a decision nobody has made yet. The next four weeks make them, one at a time.

3. Week 1: Audit What You Have

On Team or Enterprise, open claude.ai, click your initials in the lower left, then Analytics. Anthropic's Help Center says: "Usage analytics are available to Team plan Owners and Primary Owners, and Enterprise plan Owners, Primary Owners, and Admins." Enterprise Admins see everything except spend.

The Overview shows weekly active members, active members against assigned seats by group and by member, adoption level, product stickiness, skills with their uses and cost per use, connectors, estimated time saved and spend. A Chat tab counts chats, projects and artifacts. A Cowork tab shows sessions per day and daily, weekly and monthly active users.

Pull four answers out of it:

• Who is active, and how often. Compare active members with assigned seats, group by group. A busy group next to an empty one tells you where to start.
• What they use. Chat only, or projects, artifacts and Cowork sessions too.
• Which skills and connectors get used. A connector nobody touches is either unknown or unwanted.
• What it costs. The spend report exports as a CSV per user and per model, up to 90 days back. Enterprise adds an Analytics API, which the Primary Owner turns on, and analytics chat.

Then note the setup on one page: plan, seats, connectors and their tool permissions, organization-wide skills and the Cowork approval settings.

Numbers show who stopped. Only people can tell you why. Book five 15-minute interviews: two heavy users, two people who stopped and one partner. Ask each:

1. What would you use it for if you knew it was allowed?
2. What made you stop, or never start?
3. Do you use any other AI tool for work, including a personal account?
4. Which weekly task would you hand over tomorrow if it worked?

A yes to question three may mean some work has moved to personal accounts, which carry none of the commercial protections. The shadow AI guide covers that fix.

By Friday you know where usage lives, where it died and why. That is enough to choose the workflows.

4. Pick Three Workflows and Write Them Down

Pick three workflows, one each for the three teams with the clearest weekly use. Good candidates happen every week, involve a lot of reading and end in something a person can check.

Write each one on a single page under four headings:

• Input. The documents that start it, where they live and what state they must be in.
• Steps. What the person does in Claude, in order, including the project to open and the prompt that worked.
• Check. What a person verifies before the output goes anywhere: every number against the source, every quote against the document.
• Owner. One named person who keeps the page current.

Examples to adapt:

• Deal team. Screen each inbound CIM or teaser against the firm's written criteria, producing a one-page fit summary and first questions for management.
• Credit team. Pull covenant definitions and reported figures from the credit agreement and latest compliance certificate into the team's tracker, checked line by line.
• Investor relations. Draft answers to a new LP due diligence questionnaire from the firm's last approved answers, flagging any question with no approved answer.
• Finance. Draft variance commentary for the quarterly report from the month-end numbers, for the controller to edit.

Writing a workflow down feels slow. It is also the only way a second person can repeat what the first one figured out.

If someone has asked how to build a team operating system on Claude, these pages are version one, and the operating system guide covers the rest.

Three pages people use are worth more than a playbook nobody opens.

5. Settle the Data Question in Writing

People will not put a CIM into a tool they are unsure about. Write the rule on one page, have whoever owns compliance sign it, and cover five parts:

• What goes in. Name the document types: CIMs and teasers where the NDA allows it, the firm's own models and memos, portfolio company reporting, LP questionnaires.
• What never does. Name these too: credentials, for example, and personal data the task does not need. Your CCO and counsel decide the list.
• The plan terms. Anthropic's Commercial Terms say "Anthropic may not train models on Customer Content from Services." Chats are kept under your organization's retention settings, deleted chats leave Anthropic's back end within 30 days, and data is stored in the United States. Personal Free, Pro and Max accounts run on consumer terms, where chats can train models unless the user opts out, so firm documents stay off them.
• Ratings. Per Anthropic's Privacy Center, a thumbs up or down stores the whole conversation for up to 5 years. Owners can switch chat ratings off in Organization settings, Data and Privacy.
• Cowork. Automatically approve mode is on by default and Always allow for connector tools is off by default (Help Center). Set both on purpose. Local Cowork history stays on the laptop, outside Anthropic's retention settings, where admins cannot delete it centrally.

Anthropic's own Cowork safety page says: "Avoid granting access to local files with sensitive information, like financial documents." At an investment firm that describes most files, so name the folders Cowork may open.

This is operational guidance: your CCO and counsel own the interpretation, and the deal data guide covers what they will ask.

A rule people can read in two minutes gets followed. A policy nobody can find gets guessed.

6. Weeks 2 and 3: Train on Your Own Documents

Book the sessions in week 2, once the pages exist, and run them in week 3: one per team, on its own redacted documents, against its written workflow.

Two hours covers one workflow. Keep it roughly 70% hands-on, with everyone working on a real document.

A deal team that wants more depth can split the work into three sessions over two weeks:

1. Screening. A folder of redacted CIMs, screened against the firm's criteria with the workflow page open.
2. Diligence. Questions put to a working folder, draft diligence Q&A, and connectors to licensed data such as PitchBook, FactSet, S&P Capital IQ or Daloopa where the firm has them.
3. IC materials. A memo first draft in the house format, and a decision on which workflows deserve a saved skill.

A credit team runs the same shape on spreading, covenants, memos and portfolio risk, as the credit training guide describes.

Everyone leaves with one task to run on live work the next week. Anthropic describes its Claude Academy as "Free courses, tutorials, and use cases from Anthropic's education team." Use it for the basics, so session time goes on your documents. If you run Enterprise, have your owner take the course Deploying Claude Enterprise with Confidence.

Two hours on your own documents teach more than a day on someone else's.

7. Turn Good Prompts Into Skills and Projects

A prompt that works but lives in one analyst's notes helps one analyst. Move it somewhere the whole team inherits it.

Projects hold the work. Set up one project per deal or per fund, with the documents, criteria and instructions for that work, so a new team member starts where everyone else is.

Skills hold the method. In Anthropic's words, "Skills are folders of instructions, scripts, and resources that Claude loads dynamically to improve performance on specialized tasks" (What are skills?). Each one is a folder with a SKILL.md file.

Anthropic's provisioning guide says "Organization-wide skill management is available to Team and Enterprise plans." An owner uploads a .zip containing the SKILL.md file in Organization settings, Plugins and skills, and "The skill is immediately provisioned to all users in your organization." Giving a skill only to certain groups, bundled in a plugin, is Enterprise only.

Build one skill per written workflow from the prompt that survived week 3. Analytics then shows each skill's uses and cost per use. Owners can also switch off User-created skills if compliance wants one approved source (plugin settings).

Anthropic also publishes a private equity plugin built from skills, and the skills guide covers how to make it yours.

A prompt belongs to the person who wrote it. A skill belongs to the firm.

8. Chat, Cowork and Tasks After the Merge

On September 16, 2026, Anthropic announced: "Starting today, Claude Cowork and chat are merging into one Claude." The merged version is "rolling out on Pro and Max plans over the next few weeks, with more plans to follow."

On Team and Enterprise, less has changed than the headline suggests. Organizations that have not turned the merged version on keep chat and Cowork as they are, and Enterprise owners get 30 days' notice. Ask your owner which version your firm runs, and tell everyone.

The difference between a chat and a task:

• A chat is a conversation. You ask, Claude answers, and you carry what you need into your own documents.
• A task is work you hand over. Claude works through the files and connectors you allow and, by default, asks before acting. Hand one over when the job has several steps and ends in a document someone will check.

Screening a folder of teasers against your criteria is a task. Asking what a covenant definition means is a chat. The Cowork vs chat guide has more examples.

On approvals, keep each connector's write and delete tools at Needs approval or Blocked while a workflow is new (connector controls).

On usage, a premium Team seat carries 5x a standard seat's usage. On Enterprise, Anthropic's billing page says "All usage across your organization is billed together at API rates, regardless of who consumed it." Analytics shows chat and Cowork on separate tabs.

A confusing screen is a poor reason for an idle seat. Tell people which version they have and which tasks to hand over.

9. Champions and Office Hours

Name one champion per team in week 2. The champion owns the team's workflow page and its skill, and fields the questions that follow.

Pick from the week 1 analytics: someone already active, respected by the team and willing to give it an hour or two a week. Patience matters more than seniority.

Then hold office hours: 30 minutes at the same time each week, where anyone brings a real task and the champion works through it with them on screen. Log the questions. The ones that repeat become the next workflow page or skill.

Put the champion's hours in their calendar, and see the champion guide for the rest.

People learn tools from the person at the next desk. A champion puts that person next to every team.

10. Week 4: Measure and Decide

On day 30, open Analytics again, put the numbers next to your week 1 page, and decide in one meeting with the partners.

What good looks like is your call, written down in week 1 before anyone sees a result. Set your own targets for:

• Weekly active members by group, against assigned seats.
• Each written workflow run on live work every week, as its champion confirms.
• Time on one measured task, timed in week 1 and again in week 4. Treat the Overview's estimated time saved as a pointer; your own timing is better evidence.
• Skill uses and spend per user, against what you expected.

What to cut. A workflow nobody ran after its session gets rewritten or dropped. Switch off connectors nobody touched, and retire skills with no uses.

When to add seats. Add seats to a team once it has a written workflow, a champion and a session behind it; before then, a new seat is one more idle login. On Team, a power user who keeps hitting a standard seat's limits may need a premium seat, at $125 a month or $100 billed annually (Anthropic pricing). On Enterprise, read the spend report before adding anyone, and set spend limits per user if a few people drive most of the bill.

Send the partners a one-page day-30 readout: targets, numbers, decisions. If budget is part of it, the rollout cost guide lays out the lines.

Then start the next three workflows the same way. The second month is easier, because now there are pages to copy.

11. Where to Start

Do five things this week, before anyone books training:

1. Open Analytics and write down weekly active members against assigned seats, by group.
2. Book the five interviews for next week.
3. Name an owner for the month and a likely champion for each team.
4. Ask your compliance owner for the one-page data rule, with a date.
5. Time one task each team does every week, so day 30 has a baseline.

If the plan itself still needs setting up, start with the rollout playbook. Weighing outside help? The partner guide has an RFP.