TheoryPractitioner30 min

Change Management for Claude

Why change management is the hardest part

By the time you have configured SSO, set up Projects, verified your DPA and built your onboarding materials, you have done approximately 30% of the work. The remaining 70% is the human work: getting people to actually change how they work.

This is not a technical problem. It is a change management problem, and it requires the same discipline as any other organisational change initiative — clear sponsorship, communication, training, early wins and sustained reinforcement.

The difference between a Claude deployment with 80% adoption and one with 25% adoption is almost never the technology. It is the change management.

Securing executive sponsorship

Executive sponsorship is the single biggest predictor of enterprise AI adoption success. An executive sponsor does three things:

  1. Signals priority — when the CEO or COO says "we are deploying Claude because [business reason], and I want to see all managers using it by Q3", people pay attention. When IT deploys a new tool without leadership sponsorship, it becomes optional.

  2. Removes blockers — when a department head resists adoption because they are concerned about job security impacts on their team, the IT admin cannot resolve that conversation. The executive sponsor can.

  3. Models the behaviour — an executive who visibly uses Claude (and shares examples of how they use it in meetings and communications) signals that this is a serious tool, not a novelty.

How to secure sponsorship:

  • Present the business case to the relevant executive: what is the productivity opportunity, what is the competitive context, what will success look like?
  • Ask specifically for three things: (1) a brief message from them in the company-wide communication, (2) their willingness to discuss their own Claude use in a company meeting, and (3) their support in addressing departmental resistance.
  • Keep them informed quarterly with the adoption metrics. They cannot champion something they do not have data on.

The department pilot strategy

The most effective deployment pattern:

Select 2–3 departments as pilots based on:

  • A department head who is enthusiastic (or at least not resistant)
  • A clear, high-volume use case where Claude will produce visible results
  • A team that communicates well internally (their peers will hear about success stories)

Run a genuine pilot, not just an access rollout. A pilot means:

  • Defined use cases
  • Active support (a named point of contact available to answer questions)
  • Feedback collection
  • A go/no-go review at 30 days

If the pilot is weak (access was given but no support or structure), you will not see meaningful results, and the pilot data will undermine the case for wider deployment.

Build the case from pilot results. Collect:

  • Time saved estimates (survey at 30 days)
  • Specific examples of value created (case studies from willing pilot users)
  • Quotes from users ("Claude helped me do X in Y minutes instead of Z hours")

Bring this to the all-company communication with the expansion announcement.

Identifying and empowering champions

In every organisation, there are a small number of people who adopt new technology enthusiastically, learn it deeply, and naturally teach others. These are your champions. Finding them early and giving them a role is the highest-leverage thing you can do for adoption.

Identifying champions:

  • Who sent the most messages in the first 30 days? (Usage Dashboard)
  • Who is sharing tips in #claude-help on Slack without being asked?
  • Who emailed you asking "can we do X with Claude?"

Empowering champions:

  • Formally recognise them as the "Claude champion" for their department
  • Give them early access to new features
  • Invite them to contribute to training materials (case studies, example prompts)
  • Create a champions Slack channel where they share insights and you share upcoming changes

Champion's role:

  • First point of contact for colleagues with questions
  • Provides feedback to the admin team on what is working and what is not
  • Presents at team meetings about their Claude use cases
  • Onboards new colleagues informally

Champions do not require a formal incentive beyond recognition and the satisfaction of being the team's expert. But they do require active engagement from you — a champion who is never contacted withers.

Handling resistance

Resistance to Claude typically takes one of three forms:

"I don't need it" — the employee genuinely cannot see how Claude is relevant to their work. Response: identify a specific, time-consuming task in their workflow and demonstrate Claude doing it. A 5-minute live demo of a relevant task is more persuasive than any amount of general communication.

"I'm worried about my job" — the employee fears that AI will eliminate their role. Response: this requires empathy first, then data. Acknowledge the concern seriously. Then share the evidence that augmentation (AI + human) outperforms both AI-only and human-only on most knowledge tasks. Be specific about your organisation's intent: "We are deploying Claude to give you tools to do more valuable work, not to reduce headcount." If this is a genuine commitment, make it explicitly.

"I don't trust it" — the employee has tried Claude and found it gave wrong information, or has read negative coverage about AI hallucinations. Response: validate the concern — Claude does make mistakes, and verification is part of the workflow. Provide clear guidance: "Claude is a first-draft tool, not a final authority. You always review its outputs." Training on prompt design and output validation addresses this directly.

Resistance from managers: sometimes the resistance comes from a department head who sees Claude as a threat to their team's perceived value ("if AI can do our work, why do we have 12 people?"). This is best addressed by the executive sponsor, not the IT admin. Flag it early.

Sustaining adoption

Initial adoption is easier than sustained adoption. Three strategies for keeping Claude usage alive after the novelty wears off:

Prompt of the week: share one new, relevant use case prompt every week in #claude-help or via email. Rotate by department. This keeps Claude visible and surfaces use cases people had not thought of.

Monthly spotlight: share one case study per month in the company newsletter or all-hands: "How [name/team] used Claude to [specific outcome]." Real examples from real colleagues are the most persuasive adoption driver.

Feature updates: when Claude releases new capabilities (a new model, new file support, a new interface feature), communicate it proactively. "Claude now has [feature] — here's how you can use it for [relevant task]." Keep the communication brief and specific.

Scenario: turning a sceptic

The Head of Operations at a shipping company is openly sceptical about the Claude deployment. "My team tracks containers. Claude can't track containers." He blocks the IT admin's attempt to include Operations in the pilot.

Six months later, his peer in Finance mentions that their team uses Claude to summarise shipping exception reports from 30 pages to a 3-paragraph brief. The Head of Operations asks to see a demo. The finance manager shows him in 10 minutes. He immediately sees the application for his own team's daily exception reporting.

Three weeks later, Operations has a higher Claude adoption rate than Finance.

The lesson: peer persuasion beats top-down mandates. The IT admin's most valuable action was ensuring Finance's success story was shared at an all-hands meeting.

Key takeaway

Deploy Claude like a product, not an IT rollout. Executive sponsorship, a structured pilot with measurable outcomes, identified champions, and a communication strategy for resistance are not optional extras. They are the deployment. Technology is ready in four weeks; adoption takes four months of deliberate human work.


📖 Official Documentation See this in practice in Anthropic’s live support docs: