Forty Customer Questions a Week Cost About a Dollar per Thousand Answers. A Support Bot Built From Your Own Policy Pages

Why this pays

A one-person shop gets the same forty questions a week: where is my order, do you ship to Ireland, can I return this, does it run small. Each one costs a few minutes and a little goodwill when it waits until evening. A bot that answers them from your own policy pages costs about a dollar per thousand answers, and the two numbers that decide whether it is still answering at the end of a busy month are printed on each provider's page.

The value is not the dollar saved; it is the forty replies that go out in your shop's words while you pack orders. The play is a desk assistant first, an afternoon in a browser with no code, and a widget on the site only if someone can host a small server.

The play, in three lines

1. Write the shipping, returns, sizing and payment pages as short plain files, one per topic, before you open any console.
2. Paste them into Google AI Studio's System Instructions with one line above, answer only from these pages, link a billing account, and play the customer with your ten commonest questions until eight come back right.
3. When it is right, turn the same prompt into one API call with the files attached, size one exchange, write the 429 fallback message, and read the ceiling before the price.

The path above walks every step. What follows is the reasoning behind each and the numbers you quote from.

Where the money comes from

• Nowhere, directly. This is a cost line, not an income line: about a dollar per thousand answers on the cheapest route.
• The hours it returns. Forty questions a week answered from the desk assistant in seconds each, in the shop's own words, instead of forty evening emails.
• The sale it keeps. A customer who asks whether you ship to Ireland at 2pm and hears yes at 2:01pm is still on the page.

The pain you are selling the cure for

The customer wants one answer, now, in plain words, from the shop itself, not a chat window that invents a returns policy. The pain on your side is the repetition and the invention: a bot that reads your policy pages and says "that is not in our policy, email us" when it does not know is the whole product. Everything in the path is there to make it say only what your pages say.

Step one: the policies as files

• One file per topic, in plain words, covering the questions you actually get: shipping countries and times, returns and who pays postage, sizing, payment methods, order changes.
• OpenAI's file search accepts .pdf, .docx, .md, .html, .txt and .json among others, with the first 1 GB of storage free and $0.10 per GB per day after. Gemini's File Search takes documents up to 100 MB each, 1 GB total on the Free tier and 10 GB on Tier 1, stored free. For a shop's policies, storage on either is zero.

Path A: in a browser, an afternoon

• Google AI Studio opens the Playground with a new chat prompt. Run settings, top right, holds the System Instructions field; paste the policy files with one line above them: answer only from the pages below, in the shop's voice, and if the answer is not in them say so and offer the shop's email.
• Be the customer. Type the ten questions you get most, the way customers type them, and read the answers against the pages. Where it invents, tighten the line; where it is right but cold, add a sentence about tone. The quickstart's own example is exactly this shape, a customer service chatbot that only talks about a company's product.
• The result is a desk assistant: paste the customer's message, copy the answer, send it from your own inbox. For forty questions a week that is most of the value of a bot, with no code and no server. Get code turns the same prompt into the API call for Path B.
• Before any real customer message goes through it: Free tier content is used to improve Google's products, Paid tier content is not, and linking a billing account moves the project to Tier 1 typically instantly. Link it first.

Path B: one call, with the files attached

• OpenAI, the file search tool in the Responses API: upload each file, create a vector store, add the files, poll until each file's status is completed, then send the customer's message with the file_search tool and the vector store id in the tools list. The response holds a file_search_call item and a message with file citations. Beyond tokens: $2.50 per thousand tool calls, and the storage above.
• Anthropic, prompt caching: put the policy text in the system prompt and mark it with a cache_control block of type ephemeral. The cache lives 5 minutes, refreshed at no cost each time it is used; a 1-hour cache costs extra. On Claude Haiku 4.5, base input $1 per million tokens, writing the cache $1.25, reading it $0.10; Sonnet 5 $2, $2.50, $0.20; Opus 5 $5, $6.25, $0.50. Your 5,000 tokens of policy are read at a tenth of the price on every exchange after the first.
• Gemini, File Search: create a store, upload the files, pass the store as a tool on each request. Storage and query-time embeddings are free; you pay for embeddings once at indexing and for retrieved document tokens as regular context. It cannot be combined with Google Search grounding in the same request.

One exchange, then the month

• The table above is one exchange multiplied by a thousand. The AI job cost calculator has the token prices for every model on this page.
• The line that trips people: Anthropic's minimum cacheable prompt length is 4,096 tokens for Haiku 4.5, 1,024 for Sonnet 5, 512 for Opus 5. A 3,000-token policy set on Haiku 4.5 is processed without caching and no error is returned; the only sign is cache_creation_input_tokens and cache_read_input_tokens both reading zero. Pad short policies past the floor with the FAQ, or pick the model whose floor you clear.

The ceiling comes before the price

• The chart above is OpenAI's usage tiers: Free, for users in an allowed geography, $100 of usage a month; Tier 1 at $5 paid, also $100; Tier 2 at $50 paid, $500; Tier 3 at $100 paid, $1,000; Tier 4 at $250 paid, $5,000; Tier 5 at $1,000 paid, $200,000. Underneath sit rate limits in requests and tokens per minute and per day, hit on whichever comes first; file search has its own, 100 requests a minute on Tier 1.
• Anthropic's spend caps: $500 a month on Start, $1,000 on Build, $200,000 on Scale, with a token bucket that may enforce 60 requests a minute as one a second, per model.
• Gemini's Tier 1 needs a linked billing account, caps the account at $250 and spend at $10 per rolling 10 minutes; Tier 2 needs $100 paid plus 3 days and caps at $2,000.
• None of this touches a thousand exchanges. A product that goes viral on a Saturday can touch all of it.

Write the fallback before the cap

• Anthropic: once the spend cap is reached, API usage pauses until 00:00 UTC on the first day of the next month unless a higher limit is requested; every request returns HTTP 429 with enforced_spend_limit_reached and no retry-after header, so retries fail until access resumes.
• Gemini: 429 RESOURCE_EXHAUSTED, clearing on its rolling 10-minute window.
• The bot's code, or your desk routine, treats a 429 as a message, not a crash: one plain sentence, the shop's email, and a note to you.
• Anything that is not a live reply, the week's ticket summary, tagging, translating the FAQ, goes through OpenAI's Batch API: 50% cheaper, a separate pool of higher limits, done within 24 hours, never competing with the customer at the counter.

Where the bot lives

• A bot on the website needs a server between the page and the provider, because the key cannot sit in the page.
• OpenAI's ChatKit is the embeddable chat for that; your server creates a ChatKit session with a unique user identifier for each end user and hands the client secret to the widget. Its no-code companion, Agent Builder, is being deprecated and is scheduled to shut down on November 30, 2026.
• The honest options are two: a developer who hosts a small server, or the desk assistant from Path A. For forty questions a week, the second is often the right size.

What ends it

• Anthropic may suspend all access over a suspected Usage Policy breach and disclaims liability for the loss that causes. OpenAI may limit or suspend access for a breach, without prior notice where necessary.
• The defence is the folder from step one: the policy files and the system prompt are yours, rebuildable on another provider in an hour. The data page reads what each provider keeps of the conversations.
• If the shop lives on Shopify, the plan is $25 a month paid monthly or 19 US$ a month paid yearly, and the bot's dollar sits under it.

What to do today

1. Write the shipping and returns pages as two short files.
2. Paste them into Google AI Studio's System Instructions with the answer-only-from-these line, link a billing account, and ask it your ten questions.
3. Eight right: you have a desk assistant before dinner. The drafting page turns the same system prompt into one API call.
4. Write down the caching floor and the 429 fallback before it ever talks to a customer unsupervised.