In-house live chat vs a managed AI chatbot: an honest cost and coverage comparison

Live chat is the channel customers expect an answer on straight away. The question is who should be behind it: people you employ, or an AI chatbot that someone else designs, runs and maintains for you. Both can work. Both can go badly. This comparison sets out the costs, the coverage arithmetic and the trade-offs as plainly as we can, including the cases where an in-house team is the better call.

What each option actually is

In-house live chat means your own staff answer the chat widget on your website, usually alongside phone and email, using a tool such as a help desk or a live chat platform. You hire, train, schedule and manage them.

A managed AI chatbot is a conversational assistant that answers from information you have approved, asks the follow-up questions your team would ask and hands warm enquiries to a person. A provider designs it around your business, tests it, launches it and keeps it accurate. You review and approve; you do not run it day to day. That is the model chatty leads offers, and we have tried to keep our own interest out of the numbers below.

The coverage maths most businesses skip

A week has 168 hours. One full-time agent covers 40 of them. Before anyone takes a day off, a 24/7 chat seat needs 4.2 people. Contact-centre shrinkage, the share of paid time lost to breaks, training, leave and sickness, averages 30 to 35 percent according to NICE, which pushes a true round-the-clock seat to about six hires. Most small businesses never try. They staff chat during office hours, which leaves evenings and weekends uncovered.

If you only need chat covered 9 to 5 on weekdays and your enquiry volume is low, one or two people can do it well, often while doing other work. The coverage problem only bites when customers expect evenings and weekends, or when volume spikes and the person on chat is also the person on the phone.

What an in-house team costs

US Bureau of Labor Statistics data puts the median customer service representative at $21.53 an hour, or $44,770 a year. Benefits make up roughly 31 percent of total pay for office and administrative roles, so a fully loaded agent is close to $65,000 a year before software seats, a supervisor, recruiting or the desk they sit at. Six of them for genuine 24/7 coverage is a little under $400,000 a year. Salaries differ by country, but the ratios travel: every hour of coverage is a payroll hour.

Then there is churn. Contact centres are known for high staff turnover, and every replacement means paying for hiring, onboarding and the weeks before the new person is fully productive. In a team of two, losing one person means losing half your coverage overnight.

What a managed AI chatbot costs

Managed chatbot services are usually priced as a one-off set-up fee plus a flat monthly rate, scoped to what the chatbot has to do rather than how many hours it works. The monthly figure does not change when a conversation arrives at 2am on a Sunday, and it does not need a backfill when someone resigns. Set-up typically takes a couple of weeks, during which the provider learns your services, designs the conversation and loads your approved information.

The honest caveat is that a chatbot’s cost is only low if it is actually resolving conversations. A poorly built one that pushes everyone to a form has simply added a monthly fee in front of the same problem. Ask any provider what share of conversations are resolved without a person and how handoffs reach your team.

Where in-house live chat wins

Judgement calls. A human reads tone, handles an upset customer and makes exceptions a chatbot should never make on its own. If a large share of your chats involve complaints, refunds, negotiation or regulated advice, people need to be close to the conversation.

High-value, consultative sales. When a single deal is worth tens of thousands and the buyer wants to talk to a specialist, the chat is the start of a relationship. A skilled salesperson on chat can be the best-converting channel you own.

Deep product knowledge that changes daily. If answers depend on live stock, bespoke pricing or things that are decided in the moment, your own team will always be closer to the truth than any knowledge base, however well maintained.

Where a managed AI chatbot wins

Hours. It answers at 10pm, on public holidays and during the lunch rush, and it answers the fifth visitor at the same speed as the first. For businesses whose enquiries arrive outside office hours, this alone is usually the deciding factor.

Repetitive questions. Opening hours, service areas, prices, what is included, how long it takes. Most website chats are variations of the same twenty questions, and paying salaried hours to answer them is expensive. A chatbot answers them consistently from approved information and frees your people for the conversations that need them.

Predictable cost and no staffing risk. The fee is flat, there is no recruitment, no shrinkage and no churn. For a small team, not having to find and train a replacement at short notice is worth more than the spreadsheet suggests.

The hybrid most businesses end up with

In practice the choice is rarely either or. The pattern that works best for most small and mid-sized businesses is a chatbot as the first responder on every conversation, with clear rules for when a person takes over: urgent issues, existing customers, anything the chatbot has not been approved to answer. During office hours the handover is immediate. Outside them, the chatbot captures everything your team needs and the follow-up happens first thing in the morning.

That keeps the strengths of both. People spend their time on the conversations that need judgement, and nobody who arrives at the wrong hour is met with silence.

Questions to ask before you decide

How many of last month’s chats needed a human decision, and how many were the same question in different words? What hours do your enquiries actually arrive, not the hours you are open? If your best chat person left tomorrow, what would happen to the channel? And for any chatbot provider: who designs the conversation, what happens when it does not know, and how quickly does a warm lead reach a person?

Answer those four honestly and the right option is usually obvious. If it is a managed chatbot, we are happy to show you what yours would sound like. If it is an in-house team, we will tell you that too.