Resources

Everything you need mid-call, without opening a PDF.

The three layers — lead with this

Anyone can sell a chatbot. The layers are why ours works, and why we are not competing on price with a $99/mo widget.

The agents

Voice and chat agents that talk to your customers

Custom-built for their business, not a template with their logo on it. They answer, qualify, book, follow up and hand off to a human on rules the client writes and approves.

Say it like this: “You get an employee who answers every call on the first ring and never has a day off.”

The console

Their CRM, ERP and dashboards in one place

A business console built for them: pipeline, jobs, customers, invoices and live numbers. Every conversation an agent has lands there as a searchable record.

Say it like this: “One screen that tells you the truth about your business, instead of four tools and a spreadsheet.”

The knowledge engine

Their own data, made answerable

Documents, price lists, service policies, past tickets and the old database nobody can query any more, embedded into a vector store. Agents retrieve from it before they answer.

Say it like this: “It answers with your prices and your policies, because it is reading your documents — not guessing.”

The five services, and exactly what they get

Same list as the public site — if you promise something that is not here, it is custom and it is quoted separately.

AI Voice Agents

Voice

Their pain: Calls arrive while the team is on a job, on another line, or asleep.

Outcome: Every call answered, qualified and recorded.

Vapi + Twilio + OpenAI Realtime + vector store + calendar

·24/7 inbound answering on their existing number
·Receptionist, intake and after-hours cover
·Approved answers only, retrieved from their knowledge base
·Caller intent, job type and urgency captured
·Live availability check and booking
·Transfer to a human on their escalation rules
·Outbound follow-up and reminder calls
·Written call summary into the console or their CRM

AI Chat Agents

Chat

Their pain: Their website answers questions at the speed of whoever opens the inbox tomorrow.

Outcome: Reply in seconds instead of tomorrow.

Next.js widget + WhatsApp Business API + Meta/Google Ads + RAG

·Website widget matched to their brand
·WhatsApp, Instagram and Facebook inboxes
·Paid-ads lead capture agent for Meta and Google traffic
·Answers grounded in their documents, with sources
·Qualification against criteria they define
·Quote requests and booking inside the conversation
·Human handover with full context
·Every conversation saved as a lead record

Business Console — CRM, ERP & Dashboards

Console

Their pain: Their operation runs on four tools, three spreadsheets and one person's memory.

Outcome: One screen that tells them the truth.

Next.js + Postgres + role-based access + their existing tools

·Custom CRM: pipeline, stages, owners, activity history
·Operations: jobs, scheduling, dispatch, status
·ERP modules where needed — quotes, invoices, stock
·Live dashboards on the numbers they actually run on
·Role-based access for owners, office staff and technicians
·Agent activity visible per customer and per conversation
·Works alongside tools they keep, replaces the ones they do not
·Data exportable at any time, no lock-in

Knowledge Engine (RAG)

RAG

Their pain: The answers exist — in a PDF, an old database and somebody's head.

Outcome: Agents that answer like their best employee.

Vector database + embeddings + legacy DB ingestion + access controls

·Ingestion of documents, PDFs, sheets and web content
·Migration of their legacy or old database into the knowledge layer
·Chunking and embedding tuned to their content
·Retrieval before generation — it quotes them, not the internet
·Source attribution on answers so they can audit them
·Scheduled refresh as prices and policies change
·Access rules so an agent never surfaces internal-only material
·Internal search for their own staff on the same index

Operations & Back-Office Automation

Back office

Their pain: Half their week is moving information between systems by hand.

Outcome: The admin runs itself.

Make.com + n8n + OpenAI + their CRM/accounting + Google Workspace

·Document and invoice data extraction
·Email classification and routing
·Two-way sync between systems they already pay for
·Trigger-based task creation and assignment
·Reactivation and follow-up campaigns on old records
·No-show and unpaid-invoice chasing
·Multi-location and multi-team workflows
·Scheduled reporting into the console

The agents, in their words

Nobody asks for “a conversational AI service layer”. They ask for a receptionist. Use these names.

Chat · Voice

Lead generation agent

Catches the enquiry the moment it arrives, qualifies it, and pushes a scored lead into the pipeline.

Voice

AI receptionist agent

Answers the line 24/7, handles approved questions, books the job or routes the caller.

Voice · Chat

Booking & operations agent

Checks real availability, books, reschedules, sends reminders, keeps the job board honest.

Chat

Paid-ads chatbot

Sits on ad landing pages and Meta lead forms and converts the click they already paid for.

SMS · Email

Follow-up & reactivation agent

Works the leads and past customers already in their database, through to an appointment.

Chat · Voice

Support & FAQ agent

Answers the same forty questions every week from their own documentation.

Internal

Back-office document agent

Reads invoices and forms, extracts the fields, files them, raises exceptions.

Internal

Internal knowledge agent

Their staff ask what the policy is, what a part costs, what happened on a job in 2023.