AI & Automation
Chatbots that answer from your content, not from imagination
A chatbot is only useful if it resolves the question. The ones that damage a brand are the ones that answer confidently from nowhere, cannot admit uncertainty, and trap the customer in a loop with no route to a person.
Indicative
From £6,000
Fixed price agreed in writing before any build starts.
Get a quote+44 7488 265083Background photo by Jep Gambardella on Pexels
The problem this solves
Generic chatbot widgets are trained on nothing in particular and answer accordingly. Customers quickly learn to type "agent" immediately, which means you have paid for a tool that adds a step. The fix is grounding the bot in your actual content and designing the escalation path first.
What you get
Knowledge base ingestion
Your documentation, product data, policies and past tickets indexed into a vector store, with a refresh pipeline so answers do not go stale when content changes.
Retrieval-grounded answers with citations
Every response is generated from retrieved passages and shows its sources, so customers can verify and your team can trace a wrong answer to the document that caused it.
Escalation and handover
Clear routes to a human with full conversation context passed across, plus automatic escalation on low confidence, repeated failure or detected frustration.
Channel deployment
Website widget, WhatsApp Business, Facebook Messenger or in-app — sharing one knowledge base so answers stay consistent everywhere.
Analytics that matter
Containment rate, escalation reasons and unanswered questions. The unanswered log is the most valuable output: it tells you what your documentation is missing.
Safety and tone controls
Topic boundaries, refusal behaviour for anything requiring regulated advice, and a brand voice specification.
How we work
Question audit
We analyse your real enquiry history to find the questions worth automating, which is usually a surprisingly short list covering most of the volume.
Content readiness
Gaps in your documentation are identified and flagged first — a bot cannot answer from content that does not exist.
Build and ground
Indexing pipeline, retrieval tuning and prompt design, evaluated against a held-out set of real questions.
Pilot behind a flag
Released to a traffic slice with human oversight, measuring resolution against the baseline before wider rollout.
Launch and tune
Full rollout with weekly review of the unanswered-question log for the first month.
What you should expect
- Measured containment rate against your pre-launch baseline
- Every answer traceable to a source document
- Customers never trapped without a route to a human
- A ranked list of documentation gaps, refreshed weekly
Built with
- Mistral
- Claude
- pgvector
- Qdrant
- Next.js
- Node.js
- Python
- WhatsApp Business API
- Redis
- PostgreSQL
Mainstream, well-supported technology — chosen so you can hire for it and so another team could take the project over.
AI Chatbots — your questions
Including the ones about cost, which most agencies leave off the page.
A grounded chatbot on a single channel typically runs £6,000 to £14,000 depending on how much content needs ingesting and how many systems it must query. Monthly running cost is usually £40 to £300 in model and hosting fees at typical SME volumes — we model it against your actual enquiry numbers before you commit.
Yes, by calling your systems rather than guessing. We connect it to your order or CRM API with authentication, so it can answer "where is my order" from live data. This requires identity verification in the conversation, which we design as part of the build.
It says so and offers a human. That behaviour is deliberate and tested — a bot that guesses is worse than no bot, because it generates support tickets and erodes trust. Low-confidence cases are logged so the gap can be closed in your documentation.
Yes. We deploy to WhatsApp Business API alongside the website widget, sharing the same knowledge base. Note that WhatsApp has its own template and opt-in rules for business-initiated messages, which we handle during setup.
Related services
Most projects touch more than one of these.
AI Development
Custom AI systems scoped around one measurable task, with accuracy, latency and cost per request agreed before build.
Read moreAI Agent Development
An agent with access to your tools, scoped permissions, a full audit trail, and a hard stop before anything irreversible.
Read moreRAG & Knowledge Bases
Retrieval-augmented systems that turn scattered documents into answers with citations, for staff or customers.
Read moreLLM Integration
AI features added to an existing product, with cost caps, streaming, fallbacks and an evaluation set — and no lock-in to one provider.
Read moreAI Process Automation
The repetitive tasks in your week identified, costed, and automated where the numbers justify it — with a human in the loop where they do not.
Read moreWeb Development
Custom websites and web applications engineered for speed, search visibility and long-term maintainability.
Read moreTalk to someone who has built this before
A short call is usually enough to tell you whether this is the right service for your situation — including when it is not.