Plan Mode - The Secret Weapon

Planning, specs, prototyping and context are the single biggest force multiplier in our AI-enhanced web workflow. A structured plan that tells AI exactly what to build, in what order, with what constraints, will save you hundreds of hours of debugging and reflowing business logic. The difference between a useful AI output and a spaghetti mess is almost always the quality of the input.

Planning is more than a few sessions with the latest model's chat interface. It is a deliberate phase where we define the architecture, map the problem space and build rapid prototypes. We identify which parts of the build have known solutions and which require genuine engineering. Writing plans that AI assistants can consume directly; structured, specific, and unambiguous. Tooling has caught up to that habit: Cursor's Plan Mode formalises research-then-build, and their Plan Mode launch note makes the same bet we do, better plans, cleaner output.

• Domain knowledge is best with a local indexed directory of files, nothing beats this currently. Scrape what you need, store it locally where possible and build a wiki.
• Domain specific skills combined with our growing set of standard patterns, referencing local indexes, builds an efficient initial project input.
• Be very specific: data models, relationships, the component tree, API contracts, error handling strategy, limitations, constraints.
• Define acceptance criteria and what that looks like before you lean on AI - without that essential definition, you get plausible output instead of shippable work.

The difference in output quality with the right structure is not incremental, it is categorical! Spend the time planning.

This pattern is not unique to AI of course, those who have shipped production systems know that the hard work is not in the development, it is in the review and refinement. When AI tooling matured enough to be useful, we built our harnesses on top of our existing workflows to deliver what we've always done, but better.

Before AI-assisted development, a typical Pendulum project split roughly into three phases: planning and discovery, development and implementation, and review and refinement. Each phase took a noticeable amount of time to produce the necessary data and detail ready for the next step. With AI-assisted development, the distribution shifted dramatically.

• Planning is faster and more efficient, assisting the team in research and building the necessary context and prototypes, the key to success is a close team that moves at the same pace. Research and writing specs are 100 times faster at a minimum.
• The payoff - at build time development time dropped initially by around 60% and this keeps climbing as we refine our workflows. AI produces cleaner, more architecturally sound code because it has clear instructions, established patterns, and well-defined boundaries.
• Review and refinement - Automation has always been a factor, and reviewing its process and outputs is core to its usefulness. AI is no different here, we must be confident in the output. Understanding the implementations and ensuring thedata flows match the intention of the plans is a must. 2 times faster at least, without any drop in quality.

The net effect is that total project time reduced by about a third. Not by cutting corners, accepting lower quality or blind faith in AI's output, but by front-loading more of the thinking that prevents rework.

We pass this efficiency directly on to our clients. When build time drops by a third, budgets stretch further. Features that would have been cut from scope to stay within budget become achievable, and "nice to haves" that normally get triaged out of the first release actually make the cut.

For clients, this means more product for the same investment. A dashboard that includes the export functionality. A booking system that launches with the referral programme built in. A content platform that ships with the multi-language support rather than deferring it to phase two.

It also means less risk. Shorter build cycles mean faster feedback loops so issues surface earlier. Architectural mistakes, the expensive kind, get caught in the plan and prototyping far more often.

The temptation with AI is to treat it as a shortcut past planning. It is the opposite. AI makes planning more valuable, not less. The better the plan, the better the output. The worse the plan, the faster you'll generate technical debt.

Every project we run starts here, with briefs written to be consumed by both humans and AI. Every architectural decision is documented before generation begins. This is not overhead. It is the reason we ship a third faster and our clients get more for their money.