How a global VPS provider runs operations and grows sales with Coryntas AI.

Growth was creating work everywhere

For a VPS provider, growth does not only mean more support tickets.

More customers create more product questions, technical incidents, billing cases, provisioning requests, migrations, account changes, and infrastructure work. At the same time, a larger fleet means more resources to monitor and more decisions about where workloads should run.

Much of this work is highly dependent on context.

The correct response to a customer request can depend on the product, datacenter, account, network, service status, and current policy. A connectivity issue may originate inside the VPS, at the hypervisor, somewhere in the network path, or outside the provider’s infrastructure entirely.

Infrastructure management has a similar problem. A node that looks healthy at one moment may become constrained during the next traffic peak. Allocating every available resource may improve utilization on paper while leaving too little headroom for workload spikes.

The objective was therefore broader than automating customer support.

The provider wanted AI that could participate in the actual operation of the business.

AI carries work across customer and infrastructure operations

Coryntas connects the provider’s customer conversations, business rules and operating systems so the AI system can carry work through support, administration, infrastructure and sales.

The work includes:

Sales and product selection — understanding customer requirements, recommending suitable services, continuing email conversations, notifying customers when capacity returns, and organizing sales promotions.

Customer service — handling account requests, product questions, policies, service changes, and common operational requests.

Technical support — investigating problems involving VPS instances, operating systems, network connectivity, services, and configuration.

Billing operations — working with invoices, payments, service changes, and billing-related customer requests.

Infrastructure operations — monitoring and operating the systems behind the VPS platform.

Resource and network operations — investigating network conditions and managing how infrastructure capacity is used.

Administrative operations — handling provisioning, migrations, configuration changes, and large-scale operational tasks.

Each agent operates within a defined scope. Access to infrastructure and business systems is controlled according to the work the agent is expected to perform.

What the agents actually do

The difference is easiest to see in the workflows themselves.

Diagnosing a VPS that suddenly becomes unreachable

A customer reports that they can no longer connect to their VPS.

The technical agent does not start by generating a generic troubleshooting checklist.

It identifies the affected VPS and begins checking the actual environment.

The agent can verify whether the instance is running, inspect its network interfaces, check routing and connectivity, examine the relevant service state, and review logs inside the guest where access is available.

It can then compare that evidence with information from the infrastructure around the VPS.

The system may find that SSH is running and connections are reaching the guest, but the client is resetting them during authentication. In that case, restarting the VPS or changing infrastructure would not solve the underlying problem.

In another case, the agent may identify an interface problem, abnormal network conditions, a failed service, or an issue affecting the host.

The workflow becomes:

1. Customer report
2. identify VPS
3. check guest state
4. test connectivity
5. inspect services and logs
6. inspect infrastructure context
7. isolate the likely cause
8. resolve or escalate

When the cause can be determined and the required action is within the agent’s authority, the agent can continue with the resolution.

When it cannot, the staff member receiving the escalation gets the investigation and evidence rather than starting from the beginning.

Using infrastructure more efficiently without sacrificing performance

Resource allocation is a different problem.

The provider has infrastructure serving approximately 100,000 VPS, and simply filling every node as much as possible is not a useful optimization strategy.

CPU, memory, storage, and network demand change over time. Different workloads peak at different hours. Some nodes may appear underutilized while others are approaching conditions that can affect performance.

The infrastructure agents continuously evaluate resource usage and workload distribution across the environment.

Historical utilization and workload behavior can also be used to estimate periods of higher demand.

Instead of waiting for a node to become overloaded, the system can identify increasing pressure beforehand and incorporate that information into placement and migration decisions.

A typical workflow can include:

1. Collect resource state
2. analyze utilization trends
3. estimate upcoming demand
4. identify inefficient or risky allocation
5. determine available capacity
6. rebalance workloads or adjust placement
7. verify conditions after the change

The objective is to keep more of the infrastructure productively utilized while maintaining enough capacity for workloads to operate smoothly during periods of higher demand.

That helps reduce two expensive outcomes at the same time: infrastructure sitting unnecessarily idle and infrastructure being packed so aggressively that customer performance suffers.

The same information can also influence where new VPS instances are provisioned, which workloads should be moved, and which nodes are likely to require attention before the next peak period.

Making available capacity easier to sell

Demand for a particular VPS plan or location can continue while the capacity needed to supply it is unavailable.

The system coordinates sellable VPS capacity by plan and location. It keeps that availability connected to the customer’s stated need and notifies interested customers when the relevant capacity becomes available again.

The restock notification continues a customer conversation with a service that can now be supplied. The system can recommend a suitable plan and continue the exchange by email, helping the customer move from an enquiry towards a purchase.

Capacity planning also informs placement, provisioning and migration decisions. The provider can use available resources more productively while keeping enough headroom for customer workloads and the services it is offering.

Turning customer conversations into stronger sales campaigns

The sales system brings together customer needs, available plans and locations, and the history of the conversation. It uses this information to recommend relevant services, follow up at the right time and continue handling customer replies.

That work also supports sales promotions: organizing the program around customer interest and capacity the provider can supply, then continuing the interaction as customers ask questions and make purchasing decisions.

Coryntas reports revenue increases of 40–70% across multiple sales campaigns compared with previous campaigns, based on client-held campaign reporting.

Order volume also increased substantially across the reported campaigns.

The deployed AI system measures and generates reports from the client’s data. The reports remain with the client and can be made available to both teams for review.

Improving conversion and bringing inactive customers back

The agents also work on the commercial side of the business.

Not every customer who stops purchasing has permanently left. Some no longer need a particular product, some have moved to another configuration, some stopped after a specific service experience, and others simply have no reason to return until the right product or offer becomes relevant again.

The sales system analyzes available customer history, product usage, buying behavior, previous interactions, and known reasons for stopping or changing a service.

That information is used to decide what kind of follow-up makes sense rather than sending the same promotion to every inactive customer.

The workflow can include:

1. Analyze customer history
2. identify likely need or reason for inactivity
3. select relevant product or follow-up
4. contact the customer
5. continue the sales conversation
6. measure the result

For active prospects, the same approach helps the sales agent identify what the customer is actually trying to accomplish and recommend products based on that context.

Coryntas reports a 56% relative increase in the reactivation rate among customers who had not used the service for more than a year, compared with the previous baseline.

The system uses what is already known about the customer to decide who should be contacted, why they may have stopped buying, and what is most relevant to discuss next. Email conversations continue with that history available, so follow-up can respond to the customer’s actual interest.

That turns the agents into part of the provider’s revenue operation as well as its support and infrastructure operation.

Customer requests can continue into execution

Many customer workflows follow the same principle.

If a customer requests an IP change, the agent can identify the service, check the applicable policy, determine whether a charge applies, explain it to the customer, obtain confirmation, create the billing action, verify payment, perform the infrastructure change, and report the result.

The conversation is only one part of the workflow.

The useful automation happens after the customer has explained what they need.

This applies across account changes, billing operations, technical support, provisioning, and other recurring tasks where the necessary systems have been connected to the agents.

Operating across approximately 100,000 VPS

The deployment extends from individual customer requests to operations across infrastructure serving approximately 100,000 VPS.

Agents can work at both the individual VPS and hypervisor levels.

Their responsibilities include monitoring operational conditions, examining node health, investigating network issues, reviewing resource usage, assisting with workload placement, provisioning services, performing migrations, and carrying out administrative operations.

Some tasks involve one VPS.

Others involve groups of systems or infrastructure-wide decisions.

The same agent environment can therefore support a customer who cannot connect to a single server and an operations team deciding how workloads should be redistributed before an expected period of high demand.

That operational range is an important part of the deployment.

Keeping business rules and system state together

A large portion of support and operations work depends on knowing which rule applies to which situation.

Products can have different policies. Locations can have different conditions. The same customer request may require a different action depending on the service they own.

The agents combine those business rules with information from the provider’s systems.

This allows the decision to be based on the actual account, service, infrastructure state, and applicable policy instead of generating a generic answer that an employee still needs to interpret manually.

When a policy changes, the relevant agent behavior can be updated centrally.

That reduces the amount of operational knowledge every staff member has to maintain individually and makes it easier to apply changes consistently across a large volume of requests.

40–70% higher revenue across sales campaigns

Coryntas reports revenue increases of 40–70% across multiple sales campaigns compared with previous campaigns, based on reporting generated from client data and retained by the client.

The campaigns combine improvements to promotion planning with more relevant recommendations, timely email follow-up and continued customer interaction. Order volume also increased substantially.

The range describes the reported revenue increases across several campaigns. The separate reactivation result concerns customers returning after more than a year of inactivity.

Up to 95% of previously manual task volume

Across the recurring workflows included in the deployment, the provider estimates that the agents now handle up to 95% of task volume that previously required manual staff execution.

Human involvement is concentrated around exceptions, approval-sensitive actions, unusual cases, incomplete information, and situations outside the agents’ current operating scope.

Staff also review agent activity, investigate failures, refine policies, and expand the workflows as new cases appear.

Periodic reports give the team visibility into the work carried out by the agents.

A larger share of routine work can therefore be completed without requiring a staff member to process each request from beginning to end.

Scaling workload without scaling headcount at the same rate

Before the deployment, growth in customers and infrastructure created a relatively direct increase in manual work.

More customers meant more tickets and commercial activity. More VPS meant more provisioning, monitoring, migrations, resource decisions, and infrastructure administration.

That relationship is now less direct.

Once a recurring workflow has been defined and integrated, additional volume can often be absorbed by the agent system rather than becoming the same amount of additional manual work.

The existing team can supervise a larger operation while concentrating its time on exceptions, infrastructure improvements, new products, and work that is harder to standardize.

The agents are also used to support some of that development work, giving the team more capacity to work on what the provider should build next rather than spending most of its time processing what already exists.

The deployment today

The AI system now carries work across sales, support, billing, technical operations, infrastructure management, resource optimization, and administrative execution.

The system can answer questions, investigate actual operating conditions, and perform actions through connected systems.

It helps diagnose network and VPS problems, improve how infrastructure capacity is allocated, automate operational work, connect restock availability with customer demand, support sales promotions, and bring previously inactive customers back into the sales process.

Across supported workflows, the provider estimates that the system handles up to 95% of task volume that was previously performed manually.

The result is a larger amount of customer and operational work being handled without requiring the organization to increase manual execution capacity at the same rate.