← Blog

The PM's AI Toolkit

What actually works across 7 workflow stages — and where the gaps still are.

The takeaway

Over 200 AI tools claim to help PMs. Discovery and communication tooling works. Prototyping had the biggest leap. But knowledge management — the layer that makes every other tool better — is still unsolved. The biggest risk isn't missing out on AI. It's drowning in tools that don't talk to each other.

Most PM AI tools are wrappers around the same LLM APIs with a PM-flavored prompt. Some are useful. A few are transformative. Many are solutions looking for problems.

Here's what's actually working across the seven stages of PM work.

The 7 stages of PM work — and what's actually helping
1
Discovery & Research
Dovetail, Kraftful, Sprig, Tavily
Mature
2
Strategy & Thinking
ChatGPT, Claude, Squad AI
Early
3
Specification & Docs
ChatPRD, Delibr, Notion AI
Growing
4
Prototyping & Validation
Lovable, Bolt, v0, Uizard
Strong
5
Execution & Delivery
Jira AI, monday dev, ClickUp AI
Adequate
6
Stakeholder Communication
Granola, Gamma, Fathom
Mature
7
Knowledge Management
The gap. No widely-adopted solution.
Unsolved

Strategy is the hardest stage to automate because it requires the deepest context. Nobody has cracked persistent strategic context.

The honest takes
Discovery: mature
Dovetail saves real time on research synthesis. But insights don't persist — you can't ask it six months later what users said about notifications in Q1.
Prototyping: the clear winner
Lovable and Bolt let a PM with zero coding ability produce testable prototypes in under an hour. The risk: building the wrong thing faster isn't progress.
Specs: fast first drafts
ChatPRD generates a spec in 10 minutes instead of 2 hours. But editing it to match your product still takes just as long without persistent context.
Communication: easiest to justify
If you spend 15+ hrs/week in meetings, Granola alone saves 3–4 hours. No product context needed. No persistent state required.
8–12
tools per PM
zero shared context

Seven workflow stages, 3–5 tools per stage. Research lives in Dovetail. Specs in Notion. Prototypes in Lovable. Sprint plans in Jira. The PM is the integration layer — carrying context between tools and remembering which decisions live where.

The real problem: tool fatigue
Tool sprawl (typical PM)
Dovetail → research
ChatGPT → strategy
Notion AI → specs
Lovable → prototypes
Jira AI → execution
Granola → meetings
Gamma → decks
Claude → thinking
8+ tools. Zero shared context. PM is the integration layer.
Consolidated (the goal)
Research tool → discovery
Product partner → strategy + spec + proto + exec
Meeting tool → communication
3 tools. Shared context in the middle. Knowledge compounds.

Every AI tool in stages 1–6 would be dramatically better with persistent product context. None of them have it.

Choosing your stack
Start with your bottleneck — 20 hrs/week in meetings? Start with communication tools. 5 specs/month? Start there. Don't adopt tools for stages that aren't your constraint.
Evaluate the context ceiling — "Will this be better on day 100 than day 1?" If it treats every interaction as independent, the time savings are real but static.
Minimize integration tax — Every new tool costs onboarding, subscriptions, context-switching, and cognitive load. Most don't save more than they cost.
Look for knowledge accumulation — Tools that learn your product, decisions, and constraints will be dramatically more useful in a year. Tools that don't will be exactly the same.

Context is the multiplier. Every AI tool without it is running at a fraction of its potential.

This is what we're building at DISKO.

Stages 2–5 plus knowledge management. One agent. Multiple skills. Shared context that compounds over time.

Join the beta →