Who are the main characters?
A 2-person graphic design studio handling projects for small and medium enterprises such as flyers, menus, and social media graphics. They generate 8–10 quotation sheets per month. Previously, they opened old Word files and manually changed client names, items, and amounts before saving and sending.
What problems did they face?
The quotes were manually edited in Word, leading to three common issues: missing items (realizing halfway through the project that some items weren’t quoted), incorrect unit prices (occasionally one zero too few or too many), and inconsistent payment terms on each quote (making it hard to explain when clients compared). Each quotation took on average 45 minutes, with much of the time spent searching for how similar past projects were quoted.
How did AI get involved?
Without purchasing a new system, they used existing AI chat tools and split the process into three stages:
Stage 1: Template First—They fed the AI one year’s worth of quotes (de-identified) and asked it to summarize. It generated 3 templates: graphic design (flyers/posters/menus), social media management (subscription-based), and event visuals (project-based). New projects start by selecting a template rather than creating a quote from a blank Word file.
Stage 2: Checklist—Before sending each quote, the AI runs a checklist: ensuring all items are included, quantity × unit price is correct, tax (5%) is included, payment terms (deposit ratio, final payment timing) are consistent, and quote validity period is noted. This step caught 2 real errors such as “quantity 5000 typed as 500” and 2 cases where tax was omitted.
Stage 3: Human Approval—Amounts, discounts, and special terms (e.g., rush fees, revision limits) are always confirmed by the studio lead, with no AI involvement. The checklist focuses only on completeness and calculation accuracy, not on whether the price itself is appropriate.
Key numbers
Time per quote dropped from 45 minutes to 15 minutes. After two months with the checklist, 2 pricing errors and 3 missing items were intercepted. The template library allowed turnaround from “I’ll get back to you tomorrow” to “I’ll send it before the end of today” for new quote requests.
What is still decided by humans?
How much to charge, whether to offer discounts, and how to price rush jobs—these business judgments remain with humans; AI does not participate. AI only distills historical experience into templates and acts as a second set of eyes before quotes go out.
