Trapped with no way out
No clear route to a person, or a hand-over that drops the conversation so the customer has to explain everything again. The frustration lands on your brand, not the bot.
AI chatbots and voice assistants
AI Chatbots and Voice Assistants for Web, WhatsApp and Phone
We build assistants that answer customers and staff from your own content, look up orders and bookings in your systems, and pass the conversation to a person — with the full context — the moment it needs one. On web chat, WhatsApp or a phone line, with analytics that show what people ask and what each conversation costs.
- Answers grounded in your content, not the open web
- Hand-over to a person with the full conversation
- Cost per conversation tracked from day one
Why customers give up on chatbots
Most chatbots let customers down in the same few ways.
Scripted bots taught people to type “speak to a human” straight away. Language models fix the stiffness, but bring new ways to go wrong if they are not grounded, limited and watched.
Where an assistant helps customers and staff
The strongest cases are frequent, answerable questions and simple transactions that currently wait in a queue:
- Order, delivery and booking status
- Product, policy and eligibility questions
- Appointment booking and changes
- Qualifying sales enquiries out of hours
- Staff questions to IT, HR and operations teams
Invented policies and prices
An assistant answering from a model’s general knowledge will happily make up a returns window or a discount — and customers may hold you to what it said.
Costs that grow with chatter
Long conversations, large prompts and premium voice models on every call add up quickly. Without per-conversation tracking, the first warning is the invoice.
Transcripts nobody reads
The questions that reveal a confusing policy or a broken checkout step sit unread in chat logs, and never reach the people who could fix them.
What we build
Assistants for the channels your customers already use.
One assistant, several front doors: the same grounded knowledge, tools and hand-over rules behind your website chat, your WhatsApp number and your phone line.
Website and in-app chat
A chat widget or in-app assistant styled to your brand, answering from your content and keeping the conversation when a customer changes page or signs in.
Website development
WhatsApp Business assistants
Assistants on the WhatsApp Business Platform that answer questions, send order updates and booking reminders within WhatsApp’s messaging rules, and pass chats to your team’s inbox.
Discuss this
Voice assistants for phone lines
Inbound call handling with speech recognition and natural-sounding voices: routine questions answered, details taken, and callers transferred to the right team with a summary on screen.
Discuss this
Hand-over to your team
Transfers into Zendesk, Intercom, Freshdesk, HubSpot or your contact-centre platform with the transcript, the customer’s details and the reason for escalation.
Helpdesk and CRM integration
Grounding in your own content
Answers drawn from your help centre, policies and product data through a retrieval layer, with the assistant declining questions its sources do not cover.
RAG and knowledge search
Conversation analytics and cost tracking
Dashboards for resolution, hand-over reasons, unanswered questions, satisfaction and cost per conversation, with transcripts your team can search.
Analytics and dashboards
How we work
From real transcripts to a live assistant.
Our seven stages, shaped for conversational AI: real transcripts before design, hand-over rules before build, and a soft launch before the assistant meets every customer.
- 01
Discovery
We read a sample of real chats, emails and call notes to find the questions people ask most, which ones need a system lookup, and which must always reach a person.
Top question types and hand-over rules
- 02
Strategy
We choose the channels to start with, the systems the assistant may read from or act on, the models for text and voice, and a target cost per conversation.
Channel plan, integrations and cost model
- 03
UX & architecture
We write the assistant’s tone, greeting and limits, design hand-over and fallback messages, and for voice, the prompts and confirmations that work when callers cannot see a screen.
Conversation design and a tone guide
- 04
Development
Two-week sprints with a harness of scripted test conversations, so every change to prompts, content or models is checked against the same cases.
A working assistant on a test channel
- 05
Testing
Scripted and adversarial conversations: off-topic requests, attempts to make it say what it should not, angry customers, and on voice, accents, interruptions and background noise.
Conversation test results
- 06
Launch
A soft launch on part of the traffic, or out of hours only, with your team watching transcripts and hand-overs before the assistant takes on more.
A live assistant on real traffic
- 07
Optimisation & support
We review unanswered questions and hand-over reasons, update content and flows, tune costs, and add channels or actions as confidence grows.
Regular conversation reviews
Technology
The channels and tools behind our assistants.
We build on established messaging and telephony platforms, and keep the conversational logic in code you own.
Channels
Website and in-app chatWhatsApp Business PlatformMicrosoft Teams and SlackSMS
Voice and telephony
TwilioSpeech-to-textText-to-speechRealtime voice models
Models
GPT, Claude and GeminiSmaller models for routingOpen models in your own cloudModeration and PII redaction
Helpdesk and CRM
ZendeskIntercom and FreshdeskHubSpot and SalesforceContact-centre platforms
Knowledge
Help centre and FAQ contentProduct and policy dataRetrieval with citationsContent sync on update
Analytics
Resolution and hand-over ratesUnanswered question reportsCost per conversationTranscript search
If your helpdesk’s built-in AI assistant already covers what you need, we will tell you — and can help configure it rather than build from scratch.
Use cases
Assistants UK businesses put in front of people.
Customer-facing and staff-facing, each with a defined job and a clear line where a person takes over.
Where-is-my-order assistant
Looks up the order and courier status on web chat or WhatsApp, explains the delay in plain terms, and opens a ticket when a parcel really is lost.
Appointment booking by phone
Handles booking, rescheduling and cancellation calls, checks availability in your booking system and confirms by SMS — passing anything clinical straight to your staff.
Out-of-hours enquiry capture
Answers property questions from your listings, books viewings and gathers enquiry details overnight, so the team starts the day with complete, prioritised leads.
Account and billing questions
Explains bills, payment dates and plan options from the customer’s own account after they verify, and hands anything about arrears or vulnerability to a trained person.
IT and HR help in Teams or Slack
Answers staff questions from your policies and runbooks, completes the simple requests it is allowed to, and raises a ticket for everything else.
Website sales qualifier
Answers product questions from approved content, asks the qualifying questions your sales team would ask, and books a call into the right salesperson’s calendar.
Why Techsleight
What working with us is actually like.
No inflated numbers — just how we run projects, and what you can hold us to.
Product engineering, not ticket-taking
We ask what the software is for before we estimate it — and we will tell you when something should not be built, or should be bought instead.
AI where it earns its place
LLM features, retrieval and automation built with evaluation, guardrails and cost controls, and plain software where that is the better answer.
Full-stack under one roof
Design, frontend, backend, mobile, cloud and QA in one team, so nothing falls between suppliers.
UK-focused delivery
UK business hours, estimates in pounds, and a contract with a UK company. Our engineers are based in the UK and India.
Flexible engagement
A fixed-scope project, dedicated developers or a monthly retainer — and you can move between them as the work changes.
You own everything
Code, IP, cloud accounts and documentation are yours from day one. We sign an NDA before discovery if you need one.
Support after launch
We stay on for fixes, upgrades and new features, or hand over cleanly to your in-house team with the documentation to match.
FAQs
Questions we get asked.
Straight answers on scope, cost, timelines and how we work. If yours is not here, ask us directly.
Can the chatbot hand a conversation over to a person?
Yes, and we design that first. Hand-over rules cover what the assistant must never handle, when a customer asks for a person, and when it is going round in circles. The agent receives the transcript, the customer’s details and a short summary, so nobody repeats themselves. Out of hours, it takes the details and says honestly when someone will reply.
What stops the assistant promising something we do not offer?
It answers only from your approved content — policies, help articles, product data — retrieved for each question, and declines when those sources do not cover it. Prices, refunds and eligibility come from your systems, not from the model’s memory. We test with questions designed to tempt it into inventing things, and review transcripts after launch.
Can one assistant run on WhatsApp and our website?
Yes. The knowledge, tools and hand-over rules are shared, and each channel gets its own front end. WhatsApp has its own rules on when a business may message first and its own charges set by Meta, which we check for your use case during design; voice needs shorter answers and spoken confirmations.
Will callers and customers know they are talking to an AI?
We recommend they always do. The assistant introduces itself as automated, offers a route to a person, and never pretends to be a named member of staff. It is the honest approach and it sets expectations. The exact wording, and any disclosures your sector needs, are for your compliance team to confirm.
What does an AI chatbot cost to run per conversation?
It depends on the channel, the model and how long conversations run: voice costs more than text, and WhatsApp adds Meta’s charges. We estimate cost per conversation during discovery, then keep it down with smaller models for simple turns, shorter prompts and caching. Model and channel fees are paid directly to those providers. Discovery starts from £2,000.
Which helpdesk and CRM systems can the assistant connect to?
Most with a usable API, including Zendesk, Intercom, Freshdesk, HubSpot and Salesforce, plus common contact-centre platforms for voice transfers. The assistant can read customer and order records, create tickets and log conversations against the right contact, so your team sees chatbot conversations alongside everything else.
How do you measure whether the assistant is working?
Not by how many chats it closes on its own — a customer who gives up looks the same as one who got an answer. We track resolution confirmed by the customer, repeat contacts, hand-over reasons, unanswered questions, satisfaction and cost, and review a sample of transcripts with your team.
How is customer data handled in chats and calls?
With UK GDPR in mind: the assistant collects only what the task needs, personal data is redacted from logs where it is not required, retention periods are set per data type, and callers are told if calls are recorded. We document the data flows for your DPIA; your DPO makes the decisions on lawful basis and retention.
Can the assistant change a booking or issue a refund?
It can take simple actions — rebooking an appointment, updating an address — after the customer verifies and confirms. Anything involving money, such as a refund, is prepared by the assistant and approved by a person. For longer, multi-step tasks across several systems, see our AI agent development work.
Can we start with one channel and add more later?
Yes, and we usually recommend it. Starting with web chat or one phone queue lets you check answers, hand-overs and costs on real traffic before adding more. Because the knowledge and rules are shared, adding WhatsApp or voice later is mostly front-end and conversation design work rather than a rebuild.
Start a project
What do customers ask you most?
Tell us the channels you use, the questions that fill your queue and the systems that hold the answers. We will tell you what an assistant could handle, where it should hand over and roughly what it would cost to run.
What happens next
- 1A senior engineer reads your brief within one working day, and replies with questions or a first view.
- 2A 30-minute call to understand the goal, constraints and what good looks like — no sales script.
- 3A written proposal with scope, milestones, team and a GBP estimate you can take to your board.
Techsleight Labs is a trading name of Krapton IT Consultancy.
Explore