PROFESSIONAL SUMMARY | EXAMPLES | WORK EXPERIENCE | CREATIVE

AI TOOLING EXPERIMENT

The Summarizinator | Simple Meeting Notes Summary Tool

Project Rationale

As there are already AI notation summary tools out there, I entered this project merely wanting to experiment and gain some experience with AI tool building.

• Problem Statement
• Meeting notes are often inconsistent and unclear
• Specific decisions made
• Tasks, assignees, and timeframes
• Meeting notes lack a standardized format and communication channel
• Notes take time to assemble and sometimes lag significantly after the meeting closes

• Proposed Solution
• Create a standardized summary format with professional voice using AI tooling to allow for immediate processing of multiple notes from attendees, regardless of state of cleanliness
• Consistent format and voice
• Specific callouts for decisions and action list
• Immediate output to integrated communication channel of choice
• User Story
• As a primary scribe for project huddles and decision meetings, I need a tool that I can immediately input notes from any/all attendees in their various formats and states of completeness and output a concise record of the meeting, inclusive of decisions made and actions assigned, within minutes of completing the meeting and ready to send out to all project stakeholders using the chosen (AI integrated) communication channel (email, report slide, slack, Confluence doc, etc.)

• Actions
• Spun up a GPT instance (OpenAI's ChatGPT 4o) to outline the project and ideate on solutions
• Prompted some specific parameters to begin with and guide process
• No coding on my part
• Utilize only free tools or current level of GPT subscription
• Take meeting notes/bullets/chicken scratch and turn that into usable notes/summary to share with the project team in a standardized format
• More detailed adjustments to the prompt added during iteration process
• Generated several dummy notes to serve as practice data, emulating the same meeting from multiple perspectives and notation styles
• Work through many iterations and permutations with GPT until the process met basic criteria and utility

• Result
• It works...
• It isn't actually plugged into a comms channel or PDF/file generator, so output is text with formatting suggestions for slides or docs
• It likely has more quirks to iterate through, I have made several tweaks during testing
• It isn't polished or useful enough as-is to share, though is potentially worth improving given the right circumstances
• The basic digestion of notes and the refined output has generally been good, I've made very few edits to the notes during testing

The Summarizinator | Meeting Summary GPT

Business Impact

• As a crude experiment, this has successfully proven to be a potential time saver
• For a company not already using AI support, a refined version of this could become a significant time saver when multiplied across several TPMs/notetakers

Approach

• Selected OpenAI's ChatGPT as my starting point
• Decided to utilize a natural language GPT model as this was my first AI toolbuilding experiment
• Already had a subscription
• Utilized my everyday generalist GPT to talk through the ideation and initial formative iterations of this tool
• Primary use cases and audiences
• Input and output considerations
• Language, voice, and format decisions
• Ran through several Alpha versions before formalizing a prompt to create the specialized GPT
• Created the GPT tool and ran several more tests, tweaking and training it along the way

Challenges

• This was a private experiment
• No communication channels to actually integrate into
• No actual data to work with, needed to generate dummy data
• First time building a tool like this

Key Learnings

• Trust but Validate
• ChatGPT is extremely easy to use and knowledgeable enough to trust as a guide for a simple tool like this
• That said, there were several points where I needed to ask it to rethink or restate things
• There is room for AI support within the TPM discipline (I was previously skeptical)
• I am emboldened to look for more complex processes to automate/support the TPM discipline

Next Steps

• When an employer allows me to integrate this, test, tweak, and refine
• Consider more complex problems to solve and prototype
• Real-Time Program Status Reports (Executive/Stakeholder on-demand reporting tool with RAG status, gantts and timelines, risks and mitigations, and dependency callouts)
• PMO Program Intake Prioritize Assign Schedule (PIPAS)
• Jira-integrated resource/expense budget tracker; report budget implication by ticket/epic/sprint for past, current, and projected work with insights and suggestions