Custom development of your contact center platform for your processes

Agents copy data from one system to another, check order status across three windows, keep track of things in separate spreadsheets – and every inquiry takes an extra minute. A contact center platform covers the standard needs: telephony, one workspace for all inquiries, reporting, and integrations with Bitrix24 and Kommo (amoCRM). But every company has processes that fall outside the standard. Here is when you need custom development, how to tell it apart from configuration, how to estimate whether it pays off, and how to run the work so you get results instead of an endless project.

When you need custom development

• You need an integration with an internal system that has no prebuilt connector.
• Your process needs a new module or report.
• You need an AI workflow built for your task.
• Your agents or customers need their own interface.

Typical examples: an AI voice agent tells a customer their order status using data from your ERP; an agent sees details from several internal systems right in the customer record; data from external sources has to be collected and processed on a schedule; customers need a customer portal with their inquiry history.

The telltale sign is that agents keep doing the same thing by hand: copying data between systems, looking things up in several windows, tracking work in spreadsheets. Each of those steps takes time and creates errors. This is where custom development pays off fastest.

What manual work costs

Let's run the numbers from the AI agent savings calculator: 20,000 inquiries a month, with agents spending 60% of their working time on inquiries – the rest goes to breaks, training, and idle time. Say an agent spends one extra minute on each inquiry moving data from the ERP into the customer record. That is 333 hours of inquiry time, or 556 paid hours a month – three and a half full-time agents paid just to copy data, every month of the year.

With an AI workflow, the numbers get bigger. If an AI voice agent resolves 30% of inquiries without a live agent at an average handle time of 10 minutes, 6,000 inquiries come off your agents' plates – 1,667 paid hours, about 10 full-time positions. Subtract the AI cost per inquiry, and what is left is your monthly saving.

Custom development pays off when the monthly saving multiplied by the period is greater than the cost of development and maintenance. So before any work starts, measure how many minutes the manual task takes and how often it happens: without those two numbers you can neither make the decision nor check the result later.

First, check whether configuration is enough

Administrators can do a lot themselves, without developers:

• call forms in the form builder – completed forms are saved to the form log along with the call;
• the IVR (phone menu) and inquiry routing;
• real-time reports and statistics, plus data export to BI tools;
• a contact directory with employee skills, schedules, and statuses for transfers;
• automatic callbacks for missed calls.

For example, agents need to record the reason for each inquiry and the outcome of the call. An administrator builds that form in the form builder. You only need development if the form data must flow automatically into an internal system that has no prebuilt integration. Not sure? Start with discovery: it answers exactly that – what can be configured and what has to be built.

Why custom features survive updates

The biggest risk with a homegrown feature is that it stops working after the next update, and agents are back to juggling two windows. To prevent that, new features connect to the platform through its official entry points. The API lets your system request data and send commands. Webhooks let the platform notify your system about an event, such as "call ended." MCP connectors give AI agents access to your systems. No custom code bypasses the platform – no direct database queries, no copies of tables. This approach has three advantages:

• custom features survive platform updates because they rely on its interfaces, not its internals;
• access permissions apply to them: new screens and data are visible only to the people who should see them;
• data from new modules lands in the same reports and exports as everything else instead of living in a separate spreadsheet.

For example, an AI workflow uses the same knowledge base as your agents, and its results go into the same reports – so you can compare the AI agent and your agents on the same metrics. Custom development works in the cloud, on your own servers, or in a hybrid deployment: the platform API is the same everywhere.

How the work goes

1. Discovery: we map out the process, the users, the data, and the success criteria – what should change after launch.
2. Work plan: what is done through platform configuration, what gets developed, and which systems are involved.
3. Development in short stages – iterations: each one delivers a working result, and you review the interim versions.
4. Testing before release: changes are tested before they reach agents on live calls.
5. Launch and user training.
6. Support from the same team that built the feature.

Short iterations let agents and managers see results early and adjust requirements while changes are still cheap. That is why you should bring the agents who will use the result into discovery: they know the exceptions that are not in any written procedure.

How to estimate timeline and scope

You cannot accurately estimate a large project before discovery: no one knows yet how the data is structured or what exceptions the process has. So the estimate gets refined step by step:

• after discovery, the task is split into iterations, each with a clear result;
• near-term iterations get a firm estimate, later ones a rough one;
• requirements change along the way – that is normal, but any new estimate is discussed before the next iteration starts;
• you can stop at any step with a result that already works.

You do not pay for a large scope blindly, and the team does not promise what it has not studied yet. If the first iterations show the task can be solved more simply, the project gets smaller.

Outsourcing or outstaffing

• Outsourcing: we own the whole project, from discovery to support. A good fit when you have no development team or yours is busy with other work.
• Outstaffing: our developers join your team and follow your processes. A good fit when you have developers and need to add experience in telephony and contact centers.

You can switch formats: start with outsourcing, then, once the feature is live and your team knows it well, move to outstaffing or hand further development to your own developers.

Common mistakes

• Starting development without discovery – and redoing it after the first meeting with agents.
• Paying for development when configuration can handle the task.
• Building features that bypass the platform – with direct database queries or data copies that drift away from the source.
• Not setting success criteria – and not knowing when the job is done.
• Not showing interim versions to agents and managers.
• Not deciding up front who will support the feature after launch.

Almost all of these mistakes come from one thing: treating custom development as a one-time delivery. In reality it becomes part of a live system and is maintained together with the rest of the platform.

How to measure the result

Success criteria are set during discovery and checked after launch over comparable periods: average handle time and after-call work time, share of inquiries resolved without an agent, number of manual steps and window switches, data errors, and use of the new feature. Turn the minutes saved into money with the same formula: minutes × inquiries per month ÷ 60 ÷ share of working time × hourly cost. If the result falls short, the next iteration starts with finding out why, not with new features.

Where to start

1. For one week, log what agents do by hand: which task, how many minutes, how many times a day.
2. Turn that into money with the formula above or in the savings calculator.
3. Mark what configuration can handle: forms, routing, reports.
4. Bring the rest to discovery, along with the agents who do that work.
5. Decide up front who will maintain the feature after launch.