Private & Secure AI
How much does an AI customer service assistant cost?
The short answer
A business-grade AI customer service assistant typically costs £2,950 to £8,000 to set up in the UK for a single website or inbox, and £8,000 to £25,000 where it needs to read from booking, CRM or ticketing systems and act on them. Running costs are usually £150 to £600 a month depending on volume. Most builds go live in three to eight weeks.
Indicative price ranges
Website and inbox assistant
£2,950 – £8,000
Trained on your approved content, answers FAQs, captures qualified enquiries, hands off cleanly to a human with the conversation attached.
Connected assistant
£8,000 – £25,000
Reads live data — bookings, orders, availability, account status — with permission-aware answers, escalation rules and reporting.
Multi-channel operation
£25,000+
Web, WhatsApp, email and voice with shared context, write access to systems, human-in-the-loop review and full audit logging.
These are indicative UK ranges based on projects we have actually delivered. Scope drives the number, not the sales conversation — we quote a fixed price after a scoping session, and we will tell you if an off-the-shelf tool is the cheaper answer.
Typical timescales
- 3–5 days
Content and intent audit
We read a real sample of your enquiries and separate the questions worth automating from the ones that need a person.
- 1–2 weeks
Knowledge build
Approved answers, tone, boundaries and the explicit list of things the assistant must never attempt.
- 1–3 weeks
Integration and testing
Systems connected, escalation paths built, then adversarial testing against real historic enquiries.
- 1–2 weeks
Supervised launch
Live with every conversation reviewed, tuned daily until resolution quality is steady.
Who this suits
- Teams answering the same fifty questions every week — opening times, availability, order status, access, policies.
- Businesses with seasonal or event-driven enquiry spikes that swamp a small team.
- Organisations losing out-of-hours enquiries because nobody is in the inbox at 9pm.
- Anywhere a member of staff currently reads a PDF aloud down the phone.
Who it does not suit
- Low enquiry volumes — under roughly 200 a month the payback is slow.
- Complex advisory or regulated conversations where every answer needs professional judgement.
- Businesses with no accurate, current source of truth for the answers.
- Anyone hoping to remove the human option entirely. That damages trust faster than it saves money.
What actually drives the cost
Number of systems it must read
Static content is cheap. Live availability, order and account lookups each need an integration and its own failure handling.
Write access
Answering is one risk profile. Booking, amending or refunding on a customer's behalf needs far more guardrail and approval work.
Channels
Each channel — web, WhatsApp, email, voice — has its own build, tone and testing. Voice is the most demanding.
Content quality
Three versions of the same policy in circulation is the biggest hidden cost. The assistant will surface all three.
Escalation design
Knowing precisely when to stop and fetch a human is design work, and it is the part that protects your reputation.
The savings maths, honestly
The case for an assistant is time, not headcount. Work it out with your own numbers rather than a vendor's:
- Count enquiries per month.
- Estimate the share that are repeat, factual questions — for most businesses this is 60–80%.
- Multiply by the average handling time, including the interruption cost of stopping other work.
- Multiply by a loaded hourly rate.
A venue handling 3,000 seasonal enquiries a month at four minutes each is spending roughly 200 hours a month on answering. Recovering even half of that is a substantial number against a five-figure build — and the out-of-hours enquiries that used to go unanswered are pure upside.
We publish the detail of our own deployment on the AI concierge page, including resolution rate and hours recovered, so you can sanity-check the model against something real rather than a projection.
What separates a good build from a chatbot
- It says "I don't know." Then it hands over, with the conversation attached, so the customer never repeats themselves.
- It only uses approved content. No open-web guessing about your prices or policies.
- It respects permissions. A customer sees their own data and nothing else.
- It is measured. Resolution rate, escalation rate and enquiry quality reviewed weekly, not assumed.
Where the risk actually sits
The technology is rarely the problem. The problem is a confident wrong answer about a price, a policy or an availability. That is why we start with a content audit rather than a model choice, and why the first fortnight of live running is supervised.
If you need the assistant to work on your internal documents rather than customer-facing content, read building an internal AI knowledge system instead.
Proof
Where we have done this
How we worked these numbers out
The price bands and timescales on this page are first-party figures taken from projects The Digital Hub has quoted and delivered for UK clients since 2020, not from third-party survey data. Ranges are rounded and exclude VAT. They are reviewed at least every six months and updated when our own delivered costs move. Where a figure depends on a third-party licence or hosting cost, that cost is named in the text so you can check it yourself.
Sources
- Guidance on AI and data protection — Information Commissioner's Office (ICO)
- International data transfer agreement and addendum — Information Commissioner's Office (ICO)
- Data, privacy, and security for Microsoft 365 Copilot — Microsoft Learn
- Enterprise privacy commitments — OpenAI
Written by
Tom Beachell
Founder & Technical Director, The Digital Hub
Tom leads scoping and delivery at The Digital Hub, quoting and shipping bespoke websites, CRM platforms and private AI systems for UK businesses. The price bands and timescales on this page come from projects he has personally scoped.
Published 2026-08-26 · Last reviewed August 2026 · About The Digital Hub
Common questions
Will it replace our customer service team?
No, and we would not sell it that way. It absorbs the repetitive factual questions so your team handles the conversations that need judgement. Teams typically report fewer interruptions rather than fewer people.
What happens when it does not know the answer?
It stops, tells the customer plainly, and escalates to a human with the full conversation attached. Escalation is designed first, not bolted on.
Where is the data processed?
We deploy on UK or EU-hosted infrastructure with zero-retention agreements where the data warrants it. See our guide on UK-hosted AI and GDPR for the compliance detail.
How quickly does it pay for itself?
For businesses above roughly 500 enquiries a month, typically within the first year. Below that, be sceptical of anyone promising a fast payback.
Can it handle phone calls?
Yes, but voice is a separate build with telephony, latency and a different testing regime. Start with web and email unless the phone is genuinely where your volume sits.
Next step
Get a fixed price for your own version of this
A 30-minute scoping call, then a written fixed price and delivery date. If an off-the-shelf tool is the cheaper answer for you, we will say so on the call.
- 1You tell us the process and the tools you pay for today.
- 2We map the scope and flag anything that will move the number.
- 3You get a fixed price, a schedule and an honest recommendation in writing.
Reply within 24 hours · No obligation · Full code ownership on every build