Launch and deployment
Cloud or your own servers for a contact center: how to choose
The cloud goes live in a day and takes server maintenance off your plate. Your own servers keep recordings and customer data inside your network. It is the same platform either way, so you can start in the cloud and move later.
Where your contact center runs – in the cloud or on your own servers – determines how soon you take the first call, how much you spend up front, and who picks up at night when something breaks. In CONTACT.WORK the features are the same in both cases: the same telephony, AI agents, and analytics. Four things differ – where data lives, who maintains the servers, how long launch takes, and what drives your costs.
How cloud and on-premises differ
- Where data lives. In the cloud, call recordings, chat history, and customer records are stored in our data centers in the US and the UK. On-premises, they stay inside your network.
- Who maintains the servers. In the cloud, we do: we monitor load and update the platform. On-premises, your IT team does, or we do under a support contract.
- How long launch takes. The cloud goes live in 1 day. Installing on your own servers takes 2-7 days: one server is faster, several take longer. If you do not have servers yet, add time for purchasing and security review.
- What drives costs. The platform itself is free in both cases: no licenses and no platform fees, you pay for technical support under the support plan you choose. In the cloud, hosting is added on top, priced by load and recording volume. On-premises, instead of hosting you pay for installation work and your own hardware.
- What happens when the internet goes down. In the cloud, agents in the office lose their connection to the platform. On-premises, the platform keeps running inside your network.
What each day of waiting costs
Let's put a dollar figure on the difference in launch time. Take the numbers from the missed calls calculator: a company gets 300 calls a day, and an overloaded line misses 15% – 45 calls. Half of them were potential purchases, with a 10% conversion rate. Every business day, that is 2-3 lost sales, and twice as much revenue once you count the repeat purchase the customer also takes to a competitor.
The cloud starts winning that money back the next day; your own servers, 2-7 days after purchasing and approvals. In this example, a week of waiting (5 business days) costs as much revenue as about 22 average orders. If your data can be stored in the cloud, launch there right away and move in-house later.
When to choose the cloud
The cloud is the fastest start: sign today, and tomorrow your agents take calls in the browser. No servers to buy. The cloud fits if:
- you are launching a new contact center or trying the platform with part of your team;
- your headcount changes with the seasons – new seats are ready the same day;
- you have no in-house server or telephony specialists;
- your agents work from home or from different cities;
- your policies allow customer data to be stored in our US and UK data centers.
The weak spot of the cloud is the office internet connection. One hour offline for a line of 11 agents taking 120 calls an hour means 11 paid agent hours with no work and 120 unanswered calls – at the same conversion rate, up to 6 lost sales. That is why an office running on the cloud needs a backup connection from a second internet provider.
When to choose your own servers
Companies choose their own servers when data must stay inside their network or they already have the infrastructure to host it. Typical reasons:
- your security team requires recordings and personal data to be kept in-house – common in banking, healthcare, and government;
- the contact center must connect to systems that are closed to the internet: your ERP or accounting system, an internal CRM, a contract database;
- you have your own data center, phone system, and IT staff – you put your own hardware to work instead of paying for hosting;
- AI agents must run without external services: speech recognition, speech synthesis, and the language model can run on your own GPU server.
How many servers to buy
Your hardware budget depends on how many agents are on the line at the same time. We size the exact configuration for each project; here are rough guidelines:
- up to 10 agents – 1 server;
- 11-30 agents – 2 servers;
- 31-60 agents – 3 servers;
- more than 60 agents – 4 servers;
- more than 200 agents – we size the configuration for the project.
Each server gets a standby server with the same configuration, so one failed machine does not take the line down. Call recordings go on a separate disk or network storage: about 90 GB per 1,000 hours of conversations. For example, 50 agents on the line talk for roughly 1,000 hours a month. They need 3 servers plus 3 standby servers, and a year of recordings takes about 1,080 GB – a 2 TB disk. For AI agents on your own servers, add a GPU. Plug your numbers into the server calculator.
Why you need more servers as your team grows. Voice travels in real time: a report page can wait a second, a conversation cannot. If one machine connects calls and builds a monthly report at the same time, agents hear the audio break up. So at 11 agents telephony moves to its own server, at 31 the applications and database get their own so heavy reports do not slow down the interface, and above 60 each part of the platform gets its own machine. The base server is 4 cores, 16 GB of RAM, and a 240 GB SSD; virtual machines work too.
Hybrid: part on your servers, part in the cloud
Most often, telephony and call recordings – which hold the most personal data – stay on your servers, while AI agents and analytics run in our cloud, if your company's policies allow it. Agents do not notice the split: it is one interface. We work out what goes where with your security team.
How to move from the cloud to your own servers
It is the same platform, so call flows, settings, and integrations move with it, and the agent interface does not change. Many companies start in the cloud and move in-house once their data requirements are clear:
- Together we define which data must stay with you.
- You prepare the servers, and we install the platform on them in 2-7 days.
- We move call flows, settings, and integrations.
- We switch your numbers over at a time that works for you.
The reverse move – from your own servers to the cloud – is also possible.
How to make the choice
- Ask your security team which data must stay in-house by law or company policy. If none, start in the cloud.
- Estimate how many agents will be on the line now and in two years, and plug those numbers into the server calculator.
- Decide who will maintain the servers at night and on weekends: your IT team or us under a support contract.
- List the internal systems the platform must work with, and check whether they are reachable from the internet.
- Compare costs over 3-5 years, not the first invoice: on-premises, that is hardware including standby servers, installation, and your IT team's time; in the cloud, hosting for every month.
- Calculate what a week of waiting costs your line: it may pay to start in the cloud while your servers are being prepared.