I met individually with team members to understand their responsibilities, repetitive work, reference material, and desired support.
AI implementation · Marketing operations
Turning individual needs into role-based AI assistants.
I led a discovery-driven program to understand how different team members work, define useful AI opportunities, and build personalized Claude assistants supported by testing, documentation, and shared governance.
- My contribution
- Program design, discovery, assistant development, and rollout
- Adoption
- Seven role-based AI assistants actively in use
- Implementation model
- Discover, define, build, test, refine, and govern
- Core platform
- Claude with Smartsheet and shared documentation
The assistant portfolio
One implementation model. Personalized support for each role.
7role-based assistants
actively in use
ContentResearchOperationsactively in use
BUILT AROUND REAL WORKBrand grounded. Human reviewed.
Conceptual system illustration✓ Source documents✓ Shared standards
7role-based AI assistants actively in use
1:1discovery and feedback with team members
6repeatable stages in the implementation model
1shared governance structure
01
The challenge
Access to AI did not give the team a clear path to adoption.
Different roles relied on different information, recurring tasks, and decision-making processes. A generic assistant could not reflect those differences or give people enough context to make AI part of their regular work.
Without a structured rollout, team members would be left to experiment independently, shared knowledge could remain tied to individual accounts, and leadership would have limited visibility into how assistants were created and maintained.
The needMove the team from a blank prompt window to practical assistants built around real responsibilities, approved information, and repeatable workflows.
02
The approach
Start with the person and the work—not the platform.
I translated each conversation into proposed assistant functions and sent them back for confirmation before building.
I created role-specific projects with instructions, business context, working standards, and supporting materials.
Team members reviewed and used their assistants, then shared feedback that shaped the next version of each workflow.
03
The system
Personalized assistants supported by one repeatable implementation model.
Customize the workflow, not just the name
Every assistant reflects a defined set of responsibilities, source materials, and recurring tasks.
Build with team members
Users confirm the need, review instructions, test the assistant, and participate in refinement.
Keep shared tools with the company
A company-managed structure protects shared assistants and institutional knowledge when roles change.
Design for review and maintenance
Security, permissions, documentation, feedback, and manager review are part of implementation—not an afterthought.
Use cases in the portfolioContent and campaigns · Workflow and reporting · Customer and product insights · Technical communications · Tradeshow planning · Budget processing
04
The results
Results & Business Impact
Active adoption across marketing
7 assistants in use
Implemented seven role-based AI assistants, all actively used across the marketing team, establishing a documented approach to AI adoption.
Human review retained
Brand-aligned drafts
Improved content consistency and brand alignment by grounding AI assistants in foundational brand documentation while retaining human review and refinement.
Reduction in research time
Estimated 50–60%
Reduced research time by an estimated 50–60%, allowing team members to move into content development and production sooner.
Testing before implementation
Faster initial builds
Accelerated automation development using Claude and GitHub, enabling initial working solutions to be developed in minutes rather than waiting hours or days for specialist support. GitHub-based testing supports validation before implementation.
The research time reduction is based on direct experience and observations, rather than formally tracked performance data.
The strongest AI assistant does not begin with the technology. It begins with understanding the person, the work, and the problem worth solving.
Explore another system
Giving a global rollout one source of truth↗