# How We Think - AI Garage > The practical principles AI Garage uses to frame problems, build simple working systems, and focus effort where it creates the most value. **URL:** https://aigarage.in/how-we-think - [01. Problem Clarity](https://aigarage.in/how-we-think/problem-clarity) (markdown: https://aigarage.in/how-we-think/problem-clarity.md): A clearly stated problem is already partly solved. - [02. Gall's Law](https://aigarage.in/how-we-think/galls-law) (markdown: https://aigarage.in/how-we-think/galls-law.md): Working complexity grows from working simplicity. - [03. Pareto Principle](https://aigarage.in/how-we-think/pareto-principle) (markdown: https://aigarage.in/how-we-think/pareto-principle.md): A small number of causes often create most of the outcome. --- ## 01. Problem Clarity > A clearly stated problem is already partly solved. *Often described as Kidlin's Law.* Writing a problem clearly forces assumptions, constraints, people, and outcomes into view before a solution takes over the conversation. **URL:** https://aigarage.in/how-we-think/problem-clarity ### What it means Most difficult problems arrive as symptoms: too many manual steps, inconsistent output, slow decisions, or a vague sense that AI should help. A useful problem statement converts that discomfort into something observable. Clarity does not produce the complete answer. It removes several wrong answers. Once the affected person, current behaviour, constraint, and desired change are explicit, the solution space becomes smaller and more useful. ### Why it matters - **AI amplifies the brief:** A vague prompt or product brief produces polished ambiguity. Clear inputs give people and models a shared target. - **Tools stop leading:** Teams can evaluate automation, agents, or a simple process change against the problem instead of choosing technology first. - **Success becomes testable:** A specific problem statement makes it possible to define what better looks like before anything is built. ### Questions to ask - Who experiences the problem, and during which exact moment or workflow? - What happens today, and what evidence shows that it is a problem? - What should change if the problem is solved - and what must remain unchanged? ### In practice: Replace the solution sentence Instead of “we need an AI support bot,” write: “Support leads spend two hours each day answering repeat order-status questions because customers cannot see reliable updates.” The second sentence gives a team several solutions to test - and a way to judge them. ### Use it when - Starting discovery - Writing a product brief - Choosing what to automate ### Related - [AI Builder Fellowship](https://aigarage.in/fellowship/ai-builder-fellowship): Research a real problem before shaping and shipping an MVP. - [AI Use-Case Discovery](https://aigarage.in/#quick-wins): Turn an unclear opportunity into a practical build direction. ### Attribution The wording is commonly circulated as “Kidlin's Law,” although its precise origin is uncertain. We use it as a practical problem-framing heuristic, not a scientific law. --- ## 02. Gall's Law > Working complexity grows from working simplicity. A reliable complex system is usually evolved from a simpler system that already worked - not designed as one giant finished system from day one. **URL:** https://aigarage.in/how-we-think/galls-law ### What it means Complex systems contain many interactions, assumptions, and failure modes. Building all of them at once makes it difficult to know which part created the result - or the failure. A small working system creates evidence. It reveals actual usage, exceptions, and operational constraints. Complexity can then be added in response to what is learned rather than what was imagined. ### Why it matters - **Models are only one component:** Production AI also depends on data, permissions, interfaces, fallbacks, evaluation, and human review. Each layer needs evidence. - **Small systems expose reality:** A narrow workflow in real use teaches more than an elaborate architecture built around assumed behaviour. - **Reliability can evolve:** Teams can add monitoring, automation, and scale after the central loop works and its failure modes are visible. ### Questions to ask - What is the smallest end-to-end loop that creates a useful outcome? - Which steps must work now, and which can remain manual while we learn? - What evidence would justify adding the next layer of complexity? ### In practice: Keep one complete loop For an AI lead-qualification system, begin with one source, one qualification path, one human handoff, and one measurable outcome. Prove that loop before adding channels, personas, dashboards, and autonomous follow-ups. ### Use it when - Scoping an MVP - Designing automation - Moving a prototype to production ### Related - [Automate by AI Garage](https://aigarage.in/automate): Start with one owned workflow and grow operational capability from it. - [Shushi case study](https://aigarage.in/case-study/shushi): See how a focused workflow became a reliable operating system. ### Attribution John Gall described this systems principle in Systemantics: complex systems that work are found to have evolved from simpler systems that worked. --- ## 03. Pareto Principle > A small number of causes often create most of the outcome. *The 80/20 rule.* The useful lesson is not that every split is exactly 80/20. It is that effort and impact are often uneven - and prioritisation should reflect that. **URL:** https://aigarage.in/how-we-think/pareto-principle ### What it means Workflows rarely distribute value evenly. A few request types may create most support volume; a few handoffs may cause most delays; a few features may serve most users. Pareto thinking asks teams to find the vital few before optimising everything. The ratio is a prompt for investigation, not a promise or a substitute for measurement. ### Why it matters - **Automation has an opportunity cost:** Time spent automating a low-volume edge case cannot be spent improving the workflow that dominates the team's day. - **Data beats loud requests:** Frequency, effort, risk, and value help separate high-impact problems from merely visible ones. - **Scope becomes defensible:** A team can explain why it is solving one part first and what evidence would move another item up the list. ### Questions to ask - Which inputs, cases, or steps account for most volume, delay, cost, or risk? - What is the smallest intervention that would materially change that concentration? - What will we deliberately leave alone until the first intervention is measured? ### In practice: Rank before you automate List recurring work by frequency and time per occurrence. Multiply the two, then add business risk. The first automation candidate should come from the concentrated impact - not from whichever task is most annoying today. ### Use it when - Prioritising a backlog - Finding automation candidates - Reducing project scope ### Related - [Quick Wins](https://aigarage.in/#quick-wins): Find the focused intervention that can unblock a larger workflow. - [AI Automation Agency](https://aigarage.in/ai-automation-agency): Map and prioritise production workflows before scaling automation. ### Attribution The principle is associated with Vilfredo Pareto's observations about uneven distributions and was later popularised in quality management as the idea of the “vital few.”