Most people budget for an AI assistant the way they budget for a phone contract: find the monthly price, multiply by twelve, done. Then month two arrives and they realise the price was never the expensive part.
The real year-one cost is time. Somebody has to teach the assistant, write the answers your website never had, and read the conversations each week to catch what it got wrong. None of that appears on an invoice, which is exactly why it gets left out of the AI assistant budget and then quietly resented later.
Below is every line worth including, plus a worked example you can overwrite with your own figures.
The line everyone remembers
The subscription is easy to find and easy to compare. On SpideyChat it is a flat monthly fee: Free at $0 for one assistant and 100 messages, Starter at $20 a month ($17 billed annually) for 3 assistants and 2,000 messages, and Pro at $59 a month ($49 billed annually) for 10 assistants, 9,000 messages and the support ticket inbox. The full breakdown is on the pricing page.
The trap is not the price itself. It is choosing the plan for your average month instead of your busiest. If your enquiries double in spring, budget for spring. And if the provider you are weighing charges per conversation or per resolution, the annual figure is a range, not a number. We cover that in more detail in AI chatbot pricing models explained.
The four costs that never reach an invoice
These are the lines that decide whether year one feels cheap or expensive.
- Setup hours. Describing the business, crawling the site, uploading documents, choosing a name and welcome message, switching on the actions you want, pasting the script tag. A focused owner can do this in an afternoon. A distracted one takes a fortnight of stolen half-hours.
- Content writing. This is the big one. The assistant answers from your content, so if your site never states your call-out fee, your cancellation terms or the postcodes you cover, somebody has to write those down. Expect to write 20 to 40 question and answer pairs in the first month.
- Weekly review. Reading a sample of conversations and the unanswered questions list, then fixing the knowledge. Twenty to thirty minutes a week is typical once things settle. Skip it and quality drifts.
- Follow-up capacity. Captured leads need a reply from a person. If the assistant surfaces ten extra enquiries a month, that is ten extra callbacks. Budget the time, or the leads go cold in the inbox.
There is also a fifth, smaller line: a content fund. The unanswered questions list will show you pages your website is missing. Some businesses write them in-house; others pay a copywriter. Either way, put a figure against it.
A worked year-one budget
Here is an example for a small service business. Every number below is hypothetical. Replace the hourly value with what an hour of your time is genuinely worth, which for an owner is rarely less than what you would pay someone to cover you.
Assume an hourly value of $40, the Starter plan paid monthly for three months and then switched to annual billing.
| Line | Hours | Cost |
|---|---|---|
| Subscription, months 1 to 3 at $20 | - | $60 |
| Subscription, months 4 to 12 at $17 | - | $153 |
| Setup (teach, configure, install, test) | 5 | $200 |
| Content writing, month one | 6 | $240 |
| Content writing, rest of year | 6 | $240 |
| Weekly review, 25 minutes for 50 weeks | 21 | $840 |
| Content fund (two new website pages) | - | $150 |
| Total | 38 | $1,883 |
Look at where the money actually sits. The software is $213 of that. Your time is $1,520. The weekly review alone costs about four times the subscription.
That is not an argument against doing it. It is an argument for being honest about it, and for treating the review as part of the job rather than an optional extra. The review is also where most of the value comes from, because every fix you make stays fixed.
Spreading the cost across the year
A year-one budget is lumpy. Plan for that instead of dividing by twelve.
- Month one: heavy. Setup, most of the content writing and daily checks of conversations in the first week. Perhaps 12 to 15 hours.
- Months two and three: moderate. Weekly review, a handful of new answers each week as gaps appear. Around 2 to 3 hours a month.
- Months four to nine: light. The review settles to 20 or 30 minutes. New answers only when you change a price, add a service or notice a pattern.
- Months ten to twelve: a refresh. Re-crawl the site, retire outdated documents, update seasonal answers. Budget a half day.
If you have a busy season, move the refresh to the month before it starts.
Where budgets usually go wrong
The most common mistake is costing time at zero because it is the owner's time. The second is assuming the content already exists. Most small business websites describe services beautifully and say almost nothing about price ranges, lead times or what happens if you cancel, which are precisely the questions visitors ask.
The third mistake runs the other way: over-budgeting for customisation nobody needed. A brand colour, a name and a sensible welcome message are enough. Spending a week on persona tuning before a single visitor has used the assistant is time better spent on answers. Onboarding an AI employee goes through what genuinely needs attention first.
A fourth, quieter mistake is forgetting who does the follow-up. If the owner reviews conversations but a part-time administrator returns the calls, both people's hours belong in the budget. Split the lines by person, even roughly, so nobody discovers in month three that the assistant added two hours a week to a job description without anyone agreeing to it.
What to do before you commit a figure
Open a spreadsheet and copy the table above. Put your real hourly value in, then do three things before you trust the total.
- Count the questions your team answered by email or phone last month that your website does not answer. That number tells you how much content writing is really ahead of you.
- Check your busiest month's enquiry volume against the message allowance for each plan, and pick the plan for that month.
- Book the weekly review into the calendar now, as a recurring 30-minute slot, before you sign up for anything.
If you want to test the setup time estimate before spending money, the Free plan will tell you within an afternoon whether your figures are optimistic. Then run the return side of the equation through the chatbot ROI calculator, so the budget has something to be weighed against.