When a Product Is Enough and When You Need a Build
Last updated:
What a product does well
- Answering from your published content and documents
- Lead capture and ticket creation
- Handover to a person
- Multilingual answering
- Analytics and the unanswered-question log
For the large majority of businesses that is the entire requirement, and building it yourself would cost many times the subscription.
When you need more
- The bot must read live data — order status, stock, account balance
- It must act — book, amend, cancel something real
- It must authenticate the visitor before answering
- It has to work inside a product you sell to others
- Regulatory constraints require processing in a specific environment
The line is roughly: answering from documents is a product problem; answering from your database is an integration problem.
The hybrid answer
Use the product for the answering and add an integration for the live data. That is much cheaper than a full build and it covers most of the gap.
It is what we recommend most often when a client thinks they need something custom.
Cost comparison
| Approach | Typical cost |
|---|---|
| Product only | Free to a few tens per month |
| Product plus integration | £3,000–£9,000 once, then subscription |
| Full custom build | £15,000–£45,000 plus running costs |
Do not build a commodity
Building your own version of content-based answering is the most reliable way to spend a large sum reproducing something available for very little.
Build the part that is specific to you. Buy the part that is the same for everybody.
Frequently asked questions
Can a product read our order data?
When is a full build justified?
Can we start with a product and move later?
Do you build custom chatbots?
Been quoted for a custom chatbot?
Check whether a product plus an integration would do it. Happy to give you a straight comparison.