Help CentreQ-Hub AI — how it works and how AI usage will be charged
Summary
Q-Hub AI is available to customers now and we are not charging for it. Every account gets an allowance of AI tokens, and we are topping accounts up for free on request while we tune efficiency and understand our own cost base.
This page sets out what the AI system is, how usage is measured, what the current position is, and what the projected charging model looks like when we do move to a paid model.
What the Q-Hub AI system is
Q-Hub AI sits across the platform rather than being a single feature. In broad terms it does three things:
Ingests and indexes customer content — documents, forms, processes, registers and other records held in the customer's hubs are processed and turned into a searchable knowledge index for that company.
Retrieves the right context for a question — when a user asks something, the system finds the relevant records for that user, within their access permissions, rather than sending everything to a model.
Reasons over that context — a large language model produces the answer, summary, draft or suggestion, grounded in the retrieved records.
This is a retrieval-augmented generation (RAG) design. It matters commercially because the cost is driven by how much content is ingested and how much reasoning is done, not by seat count.
Where users meet it
Surface | What it does |
|---|---|
AI search / deep search | Natural-language search across the customer's Q-Hub data |
Q-Hub Assistant | General chat assistant available across the app |
Chat with… | Contextual chat against a specific document, form, process or record |
AI builders | Assisted creation of forms, processes and requirements |
AI actions | AI-assisted steps inside workflows |
Compliance gap analysis | In development — analysis of a customer's system against a standard |
Platform and data handling
Hosted on AWS (eu-west-2) alongside the rest of Q-Hub, with MongoDB Atlas as the vector store.
Reasoning runs on Google Vertex AI in the London region.
Retrieval respects existing Q-Hub access controls — a user cannot retrieve content they could not otherwise open.
Company admins control whether AI is on, which users have it, and which content is excluded.
How usage is measured
Usage is measured in tokens. A token is the unit we deduct for AI work carried out on the customer's behalf. Tokens are consumed by activity, not by the number of licensed users, so a small team running heavy analysis can use more than a large team doing occasional searches.
What tokens are actually used for
Indicative costs for common tasks. Ranges are real — the cost of a task depends on how much content the AI has to read to do it.
Task | Typical cost | What drives the range |
|---|---|---|
Build a simple form from a PDF | 7–20 tokens | Length and complexity of the source PDF |
Build an audit template | 10–20 tokens | Number of questions and depth of the standard |
Analyse a process for customer trends — initial request | 20–80 tokens | Volume of data being analysed |
Analyse a process for customer trends — follow-up questions | 5–10 tokens | Context is already loaded, so follow-ups are cheap |
Assess audit requirement compliance | 10–50 tokens per requirement | Volume of connected evidence and content |
Two things worth understanding from these figures:
Follow-up questions are cheap. Once the AI has loaded the context for an analysis, further questions on the same topic cost a fraction of the initial request. Users get better value by working through a topic in one session than by starting fresh each time.
Per-requirement work scales. Compliance assessment is charged per requirement, so a full standard is a large job.
Current position — no charge
Item | Position today |
|---|---|
Charge to customer | None. AI usage is not billed. |
Standard allowance | 300 tokens per account |
Additional tokens | Granted free on request, subject to review |
Review basis | Usage pattern, account size, and our own cost exposure |
Why | We are still improving the efficiency of the AI and balancing our underlying cost. We would rather learn from real usage than cap it early. |
Projected charging model
The following is the model we expect to move to but is not yet confirmed.
Monthly refreshing allowance by plan
Each customer receives a token allowance that refreshes monthly.
Plan | Refreshing tokens per month | Equivalent list value (excl VAT) |
|---|---|---|
Pro | 1,000 | £50 |
Essentials | 500 | £25 |
Basic | 300 | £15 |
Additional token purchase
Item | Value |
|---|---|
Price per token | £0.05 (5p) |
Minimum order | 1,000 tokens |
Cost of minimum order | £50 |
Additional tokens are bought in blocks of 1,000. Smaller top-ups are not offered under this model.