Cut AI coding costs by 42% with smarter context management

If you're paying someone (or something) to do research for your business, you want it spending time on the actual work, not rifling through filing cabinets looking for the right document. That's the problem a new open-source tool called Graft has just started solving for businesses using AI coding assistants.

Less Wasted Time Means Lower Bills

Here's the context: AI assistants that help build and maintain websites or software charge based on how much they process. Think of it like a taxi meter. Every time the assistant has to search through your entire codebase to find a relevant file, that meter is running. Graft is a small piece of software that sits between your instructions and the AI assistant, pointing it directly to the right files instead of letting it rummage around blindly.

The headline figure doing the rounds is a 42% reduction in the amount of searching the AI has to do. That translates directly into less cost per task, and faster results. If you're a small business owner who has brought in a developer using AI tools to build or maintain your website, this kind of efficiency gain shows up in your invoices.

Faster Turnaround on Your Digital Projects

Speed matters beyond cost. When a developer is building a new booking system for your clinic, or adding a product filter to your e-commerce shop, the back-and-forth between giving instructions and seeing results has traditionally involved a lot of waiting. Part of that wait is the AI assistant doing what amounts to unnecessary legwork.

Tools like Graft are part of a broader shift happening right now in how AI coding assistants are being used professionally. Anthropic (the company behind the Claude AI assistant that many developers use) has been publishing guidance this week on how to get more value from these sessions, which signals that this whole area of efficiency is being taken seriously at an industry level. When the companies building these tools start publishing optimisation guides, it usually means businesses are starting to notice the costs.

The Hidden Risk: Paying for Inefficiency You Don't Know About

Most small business owners commissioning web development work have no visibility into how efficiently the tools are being used. You see a day rate or a project quote. You don't see whether the developer's AI assistant spent 40% of its time doing pointless searches through irrelevant files.

This isn't about distrusting your developer. Good developers are already looking at tools like Graft precisely because they want to deliver better work faster. But it's worth knowing this conversation exists, because it gives you an informed question to ask. Developers who are paying attention to efficiency tooling are, generally speaking, more likely to deliver predictable timelines and costs.

There's also a separate but related development worth flagging: AI tools are increasingly adding what amounts to invisible watermarking to the content they generate. If your developer is using AI to write copy, product descriptions, or blog posts as part of a project, you may want a separate conversation about where AI-generated text is appearing on your site and how that affects your brand's voice.

What To Do About It

1. 1.Ask your developer what AI tools they use and how they manage costs. A competent developer should be able to answer this clearly. If they look blank, that's useful information.

1. 1.Request a brief efficiency summary on your next project. It doesn't need to be technical. Something like "here's roughly how many AI-assisted hours went into this" gives you a baseline for future comparisons.

1. 1.If you're commissioning content as well as development, ask specifically whether AI is being used to draft copy. Not to ban it, but so you can decide whether it matches your brand voice before it goes live.

1. 1.Bookmark this area. AI development tooling is moving quickly right now. Checking in on it every couple of months means you won't be the last to know when your competitors' developers are working twice as fast for the same price.