AI Tools for PMMs: What Actually Works and Why
Eight tools, real use cases, and the workflows that stuck. Co-facilitated by Lance Spence and Eve Horne at PMM Jam 2026 Beta.
Specific tools, real workflows, and what each one is actually good for
Why AI produces generic output, the prompting framework, the think step, and platform risk. Read Part 1
Lance opened with the framework and tool demos; Eve led the multi-LLM testing, NotebookLM, and Canva AI walkthroughs
From principles to practice.
Part 1 covered the why: AI is lazy without context, the prompting framework that fixes that, and the platform ownership risks most PMMs do not think about until something goes wrong. This post covers the what: specific tools, real use cases, and what the group was actually building and testing in real time.
Not every tool will fit every workflow. The goal here is not a definitive ranking but an honest view of what each one is genuinely good for and where it falls short, drawn from practitioners who were using them under real deadline pressure.
Eight tools, one honest take on each.
The go-to for message development and strategic reasoning. The extended thinking mode works as a scratchpad, it shows reasoning before producing a final answer, which makes it easier to catch hallucinations and maintain a paper trail on data-heavy work. Lance's primary tool for copy and strategic thinking; he develops messaging here before porting structured output into other tools.
Lance built a working nonprofit landing page prototype in real time: hero text, dual CTA, stat blocks, email capture. He referenced Notion's visual style for design direction. The result was rough but communicable in a fraction of the time a Figma mockup would take. Not production-ready, useful for rapid prototyping and early-stage stakeholder alignment before design commences.
Works only on documents you upload, no web scraping, which dramatically reduces hallucinations. Eve fed it materials for a nonprofit partner and generated a full narrative audio podcast summary, complete with statistics, a program framework, and a story arc. Lance uses it to fast-track product knowledge on new consulting clients. You can interrupt the generated podcast mid-playback to ask questions. Others in the group added use cases: creating structured slide decks from dense strategy documents and academic research.
Eve's go-to for SEO research, Gemini draws from Google's data directly, which gives it an edge for consumer search pattern questions. Also her strongest image generation tool when style and quality matter. Running the same prompt across multiple LLMs, Gemini consistently produced the sharpest graphics.
Eve's preference for brand-quality image generation. She described running the same prompt through ChatGPT and Gemini, then taking the ChatGPT output into Gemini for further layering and refinement. Each model has a distinct visual personality, knowing which one you are looking for matters before you start.
Eve demonstrated live: uploading a logo, a Gemini-generated poster, and a brand color palette, then generating an animated video from the combined inputs. Key lesson from the session: feeding everything at once produces hard-to-control output. Feed one element at a time and add layers incrementally to maintain control over the result.
A PMM with no coding background built a complete portfolio website over two to three weeks using Claude Code alone. Separately, another used it to build a full PMM portfolio inside a custom UI as a demonstration of AI-assisted creativity. The consistent feedback: genuinely usable without any prior coding knowledge.
Mentioned for automating repeatable Excel reporting workflows. The unexpected benefit: a shared language with engineering teams. Developers could see the generated code and immediately understand what the PM was trying to build, bridging a communication gap that prose never quite manages.
"The best people using AI are not just prompting and using the spit-out. They are layering it, running through multiple tools and adding their own judgment at every step."
Eve Horne, PMM Jam 2026 AI Peer SessionSix workflow principles.
Run identical prompts through Gemini, ChatGPT, and Claude and compare the outputs. Each model has a distinct personality and a different data source. Once you know which handles which task better, you stop fighting the wrong tool and start routing work intentionally.
A NotebookLM audio summary can become Canva animation input. A Claude messaging document can become a Lovable landing page brief. The most effective AI-assisted work is a series of deliberate handoffs between tools, with your editorial judgment connecting them.
Upload technical documentation or research materials and have NotebookLM generate a custom audio explanation. Listen on the go, interrupt to ask questions, and arrive at your first stakeholder conversation already knowing the vocabulary. Works for competitive analysis, internal strategy docs, and dense client briefs.
Feeding Canva AI, Lovable, or any generative tool everything at once produces generic, hard-to-fix output. Feed one element at a time: one reference, one constraint, one example. Incremental inputs give you more control and make iteration faster.
If you need to show a stakeholder or creative team what a landing page concept looks like, Lovable produces a working rough in minutes using the same prompting framework from Part 1. It removes the blank-canvas problem from early-stage alignment conversations.
Claude thinks. Gemini searches. ChatGPT generates. NotebookLM stays in its lane. The most effective PMM AI workflow is not picking one tool and staying loyal. It is understanding the personality behind each model and routing specific task types accordingly.
You are still the one with the judgment.
Every tool on this list still required a PMM's eye to decide whether the output was good enough, whether the language matched the audience, and whether the story was worth telling. AI does not know that the same word means something completely different to your sales team and your engineering team. It does not know that a competitor's talking point slipped into your positioning draft. It does not know what a specific executive actually cares about on a Monday morning after a bad quarter.
Use AI to get there faster. You are still the one deciding where "there" is.
The prompting structure behind these results.
Every tool explored here produced better results when fed through Lance Spence's AI Onboarding Prompting Framework: Goal, Who, What, Why, Structure, Examples, and the Think step. It is available as a free one-page PDF.
Read Part 1 first.
Part 1 covers the principles behind these results: why AI produces generic output without context, how the prompting framework was built, the think step that reduces hallucinations, and the platform ownership risks most PMMs learn the hard way.
Lance operates product marketing like a newsroom. He built the AI Onboarding Prompting Framework shared in both sessions and won the PMM Jam 2026 Impact Award with Team EnCompass.
Eve is the founder of Plankowner Marketing and creator of PMM Jam. She led the multi-LLM testing, NotebookLM, and Canva AI walkthroughs in this session, drawing from active client work and PMM Jam's own content infrastructure.