As part of our new AI In Depth class, there’s an entire module on customizing AI to improve results, by giving it context about you and your organization, and tuning various personalization settings that the AI services provide. Several people asked me (Ken) about my specific settings for Claude, so I’m going to list everything here that I’ve configured. Note that I’m describing what I do as of May 14th, 2026…this is an area that’s changing quickly, so I expect in a few weeks or months things might be very different.
As an example, Claude now supports “Organization Settings”, which I now use to provide instructions that apply to everyone at AI for Community. Also, Claude is phasing out Custom Styles, so I asked Claude to create a description of my “Ken” style, and I put that into my personal instructions for Claude. And finally, Connectors and Skills are now part of the Customize functionality, accessed (confusingly, to me) not via Settings but as a left-side panel item, similar to Projects and Artifacts.
Also, I’m not going to talk about Claude skills, which would be an entire module on its own. And I won’t cover the Claude Projects that I’ve created. I’ll be specifically discussing the customizations that can be done as part of Claude’s settings.
Organization Settings
Organization instructions
Prioritize accuracy: verify facts, cite sources when available, and challenge questionable claims or assumptions – mine or yours. Push back against biased prompts that assume a conclusion. When you cannot actually perform a task reliably, say so directly rather than produce a plausible-sounding guess dressed up as analysis. End responses with a confidence score and brief justification.
Claude Settings (Only for Ken)
General
What best describes your work?
Instructions for Claude
Product management
Writing Style
Write with a conversational, practical tone that balances enthusiasm with candid assessment. Use informal language and direct address. Acknowledge both benefits and limitations — don’t oversell. Include specific examples and honest observations about challenges. Avoid em-dashes and en-dashes. Use double hyphens ( — ) for pauses or asides instead.
Tone markers to hit:
- Sounds like a knowledgeable colleague talking, not a marketer
- Willing to say “here’s the trade-off” or “this didn’t work as expected”
- Direct but warm — not formal, not breezy
- Uses “you” and “I” naturally
- Specific over vague (“three workshops completed” not “solid progress”)
What to avoid:
- Bullet-point-heavy responses when prose flows better
- Corporate hedging (“it’s worth noting that…”)
- Overselling or burying limitations
- Em-dashes, en-dashes, or excessive bold/headers in conversational responses
Notifications
Response completions: on
Capabilities
Memory
Search and reference chats: on
Generate memory from chat history: on
General
Tool access mode: Load tools when needed
Connector discovery: on
Location metadata: on
Customize (accessed via left panel, via the More… popup)
Skills
I’ve created a “pdf-certificate” skill, which anyone in the organization can use to create a PDF certificate of completion for someone who needs/wants that after training is over.
I also have loaded the “skill-creator” skill, which I used to create the “pdf-certificate” skill. I’ll need this whenever I create a custom skill.
Connectors
I have the following connectors active and configured. Note these all had to first be enabled by our AI for Community administrator, which (fortunately) is me.
Gmail, Google Calendar, Google Drive – standard productivity tools. All are configured to always allow read access, and always require confirmation for write/delete actions.
Netlify – a hosting service I use to deploy custom webapps that I build using Claude. This is also configured to always allow read access, and always require confirmation for write/delete actions.
Slack – our main messaging tool. This is configured similarly to to the other tools, other than it has an extra “Interactive Tools” configuration area, where I have “Create a draft message” as always allowed.