The short answer
Published UK pricing puts an SME chatbot at £3,000 to £25,000, and an LLM assistant with retrieval over your own documents at $15,000 to $40,000. The costs people miss are the running ones: API usage, hosting, and a vector database, every month, forever.
What the market charges
| What you are building | Published range | As published |
|---|---|---|
| UK SME chatbot | £3,000 to £25,000 | GBP |
| Rule-based chatbot, no LLM | $5,000 to $15,000 | USD |
| LLM-powered with retrieval (RAG) | $15,000 to $40,000 | USD |
| Advanced multi-modal | $40,000 to $100,000 | USD |
| Enterprise, multi-agent orchestration | $100,000 to $300,000+ | USD |
| RAG knowledge base added to an existing build | £3,000 to £6,000 | GBP |
Regulated sectors, banking and healthcare especially, add 25 to 35% to the baseline.
The running costs
This is where AI differs from ordinary software. A normal feature is built once and costs hosting. An AI feature has a per-use cost forever.
Monthly, ongoing
- LLM API usageRoughly £50 to £300 a month depending on conversation volume. This scales with success.
- Cloud hostingCommonly $200 to $1,000 a month for an AI workload.
- Vector database, if you use retrieval$50 to $500 a month.
- Maintenance and model changes15 to 20% of build cost annually. Models get deprecated; your prompts and evaluations need revisiting.
Integration costs, itemised
Published add-ons for connecting an AI feature to systems you already run:
| Integration | Published add-on |
|---|---|
| CRM (HubSpot, Salesforce) | £1,000 to £2,000 |
| E-commerce (Shopify, WooCommerce) | £500 to £1,500 |
| Booking system | £500 to £1,000 |
| Payment processing | £1,000 to £2,000 |
| Custom API or database | £1,000 to £3,000 |
The question to ask before spending any of it
We build AI features when they genuinely do something a query cannot. What we do not do is add a chatbot to a product that needed better navigation.
Common questions
How much does it cost to add AI to an app?
Published UK pricing puts an SME chatbot at £3,000 to £25,000. An LLM-powered assistant with retrieval over your own documents is commonly quoted at $15,000 to $40,000. A RAG knowledge base added to an existing build runs £3,000 to £6,000.
What are the ongoing costs of an AI feature?
LLM API usage of roughly £50 to £300 a month depending on volume, plus cloud hosting and, if you use retrieval, a vector database at $50 to $500 a month. Budget 15 to 20% of the build annually for maintenance and model changes.
Does my app actually need AI?
Usually not. Most requests we get for AI features describe a problem a database query or a well-designed form solves more cheaply and more reliably. AI earns its place when the input is genuinely unstructured language.