In late October, shares in Nvidia, Intel and other publicly traded technology companies fell on the US stock market. This renewed the question: when will the enormous investments in AI infrastructure begin to deliver the expected returns?
On October 30, the Nasdaq Composite index lost nearly 2.9%. Investor concerns regarding the massive spending by tech companies on AI and when that spending would pay off became one of the factors behind the decline. On November 3, the index recovered some of its losses. Such fluctuations do not signal the end of artificial intelligence development, but they do show how much more closely the market is evaluating the economic side of technology investments.
Yuriy Syvytsky, in his column for Intecracy Ventures, explored the question of whether future AI revenues will be able to justify current investments. For me, this discussion has another, more practical dimension: what technologies, products, and services will enterprises be willing to pay for in the coming years?
The answer to this question will also determine the development of Intecracy Group in 2027.
AI is becoming more accessible, but business solutions are not becoming simpler
In the global AI industry, two processes are occurring simultaneously that, at first glance, seem to contradict each other.
On one hand, investments in computing infrastructure continue to grow. According to an S&P Global Ratings forecast, capital expenditures by the largest hyperscalers could exceed $1.3 trillion by 2027. This represents massive funding for data centers, server hardware, energy, and supporting infrastructure.
On the other hand, the use of the models themselves is becoming cheaper. According to the Stanford AI Index 2025, inference costs for models performing at a level comparable to GPT-3.5 on the MMLU benchmark fell more than 280-fold between November 2022 and October 2024.
For application developers, this opens up new opportunities. Technologies that were economically inaccessible to many enterprises until recently are gradually becoming standard tools for automation.
At the same time, the accessibility of AI is intensifying competition. If almost any developer can connect a model, the mere fact of its use no longer creates a lasting competitive advantage.
As Yuriy Syvytsky noted in his column for Software Ukraine, the economics of software development are also changing. Generative AI accelerates the writing of individual code fragments, prototyping, and the execution of routine tasks. However, creating a reliable enterprise system takes more than generating code. Architecture, security, integration, data quality, testing, and maintenance remain complex engineering tasks.
I expect that an increasing portion of the value of IT companies will be linked not to the volume of code written, but to the ability to understand how an enterprise operates, propose solutions, and make sure they are used successfully.
Our priority: practical results for customers
In 2027, we plan to pay more attention to the development of our own products, the integration of AI into business processes, and joint projects among the alliance's companies.
At the same time, I do not think we should add artificial intelligence to every product simply because the technology is attracting market attention.
In corporate automation, something else is more important: what problem the system solves, how much time or how many resources it saves, how the quality of work changes, and how much its operation will cost after launch.
For example, AI can help an employee process a document, prepare a response to a client, find necessary information, or determine the next step in a business process. However, the result does not emerge when the model has generated text, but when the business task has been completed.
Serhiy Balashuk, in his column for Software Ukraine, proposed evaluating the full cost of a correctly completed business operation. We want to apply this approach to enterprise AI solutions. The calculation must include not only tokens or computing resources but also data preparation, integration, human oversight of the result, maintenance, and error correction.
For our teams, this also means changing the approach to the start of a project. Even before development, it is necessary to agree with the customer on what result will be considered successful, how it will be measured, and what will happen after the pilot is completed.
Developing our own products should also change the economics of our companies. We want to make greater use of existing platforms, modules and proven integrations to shorten implementation times and reduce the cost of repeating development work. Each new project should build on what we have already created. That requires continued investment in product development and maintenance of the shared technology behind those products. Custom development will remain important for tasks shaped by each customer’s particular needs.
Corporate AI: from a separate assistant to a business process
The October series of IQusion Days events showed how broad the possibilities for applying AI in enterprise systems have become.
One area is human resources management. In IQusion products, we are developing tools for working with resumes, matching skills with job requirements, and preparing information for employee training and development.
Employee surveys conducted as a conversation with an AI agent attracted particular attention. Instead of a complex questionnaire, an employee can answer questions in a normal conversation, and the system helps structure the information received.
For HR departments, this is an opportunity to reduce routine work and obtain higher-quality information. At the same time, personnel decisions must remain with authorized employees who can verify the system's recommendations.
Another important area is the use of AI in the Scriptum business process automation platform.
The system can help determine the type of document, extract details, prepare a summary, and suggest an assignee or draft a response. Such capabilities are needed in procurement, contract work, document management, and request processing.
However, I consider it fundamentally important not to leave AI as a separate window where an employee manually transfers information.
The result of the model's work should return directly to the document, task, request record, or other element of the business process. At the same time, the access, verification, and approval rules adopted in the organization must be maintained.
A separate area is working with corporate knowledge. At IQusion Days, we examined RAG and AI Governance approaches: using AI based on internal documents, regulations, contracts, and instructions.
A lot of work remains for integrators here. It is not enough to simply connect a model to a document repository. It is necessary to prepare data, keep sources up to date, set access permissions, verify responses, and train users.
This is part of a complete implementation for which an IT company must be prepared to take responsibility together with the customer.
DooxSwitch: communications, AI, and the economics of digital services
Corporate and telecommunication systems remain a strategic part of our business.
At IQusion Days, we presented the capabilities of voice AI agents, including Omilia technologies. They allow for the automation of routine requests, working with corporate information, and transferring complex issues to operators.
We are also developing this area at DooxSwitch, relying on the team's many years of experience in telecommunications.
DooxSwitch combines the development and integration of solutions for call routing, billing, corporate telephony, messaging, and interaction with CRM, ERP, and business process management systems.
In such solutions, a voice AI agent can become part of a standard service process. It accepts a request, clarifies the issue, obtains data through authorized interfaces, creates a task, or transfers the conversation to an employee along with the necessary context.
The customer receives not a separate chatbot, but the ability to organize a full cycle of work with a request — from the first contact to its completion.
I consider the integration of AI with telecommunications billing to be particularly promising.
Traditional billing systems record calls, connection duration, destinations and services used. When AI agents are added to service, a need arises to also account for new resources: agent usage time, the number of tokens used, sessions, and other parameters.
DooxSwitch is developing tools to track this usage, allowing operators and providers of digital services to set rates, control costs and bill their customers.
This is important not only for enterprises that automate their own contact centers. Such infrastructure can become the basis for providers offering voice AI services to other companies.
Another area is video communications. We are exploring the possibilities of meeting transcription, automatic preparation of summaries and tasks, integration with enterprise systems, and the use of digital avatars for specific communication scenarios.
For 2027, our task is to develop integrated solutions in which voice, messaging, video conferencing, and AI work as parts of a single business infrastructure.
At the same time, DooxSwitch is not limited to one model of delivery or deployment. We determine the architecture, set of components, integrations, and operational procedures for each specific customer.
Infrastructure and cloud services: technology needs a reliable foundation
The spread of AI does not reduce the importance of traditional IT infrastructure. On the contrary, new services increase the requirements for computing resources, data storage, networks, security, and business continuity.
Therefore, we continue to develop our companies’ infrastructure services alongside product solutions.
SL Global Service offers cloud migration services, deployment, monitoring, and system support. At the same time, for some enterprises, a hybrid architecture that combines cloud resources with their own infrastructure will remain the optimal solution.
I see no point in moving a system to the cloud just for the sake of moving.
The customer must understand what they are getting: higher reliability, scalability, simpler support, or cost savings. If such advantages are absent, other options should be considered.
Softline's infrastructure expertise covers server systems, networks, data storage, and the integration of hardware and software. This expertise remains essential both for modernizing existing IT environments and for deploying new application systems.
Managing infrastructure costs is becoming particularly important.
At IQusion Days, we examined FinOps approaches: allocating costs between departments and services, selecting resources according to load, eliminating unnecessary expenses, and controlling the use of cloud capacity.
For enterprise AI, this is especially relevant. A small pilot might be cheap, but scaling to thousands of users or millions of operations can significantly change the economics of the solution.
Therefore, we need to work out the running costs before a large-scale rollout, not after receiving the first large bill.
Secure systems: not every task requires general-purpose AI
For banks, government institutions, defense enterprises, and critical infrastructure, information security requirements often determine the very architecture of a solution.
At Softline IT, this work includes Softline Defense and the UnityBaseDefense platform.
We are talking about systems where access control, audit logging, cryptographic protection, operation in closed environments, and the ability to verify user operations are important.
The application of AI in such systems opens up new opportunities but at the same time creates additional risks.
For example, a local AI assistant can help search for information in documents or prepare draft responses. However, its capabilities must be limited to authorized data and actions.
That is precisely why the Zero Trust and AI Governance topics that we examined at IQusion Days have practical significance for the development of corporate platforms.
If an agent works on behalf of a user, it does not mean that it should automatically receive access to all documents or the right to change any records.
Permissions must be checked when each operation is performed.
At the same time, not every task needs to be solved using a large language model. In many cases, standard automation based on clearly defined rules will be cheaper, more reliable, and easier to verify.
I believe that the ability to choose the appropriate technology, and not necessarily the newest one, is also an important professional advantage of an IT company.
AZIOT: AI goes beyond office processes
We see specific opportunities in the automation of buildings, engineering systems, and distributed facilities.
The AZIOT platform, which the Data Management IG division is developing, combines data from sensors, meters, and equipment with tools for monitoring, control, and incident response.
Here, AI can be useful primarily for analyzing anomalies, detecting abnormal resource consumption, finding the causes of recurring malfunctions, and preparing recommendations for operators.
For example, the system can help notice unusual electricity consumption faster or find patterns in equipment operation that are difficult to detect manually.
But the value of such solutions is determined not by the complexity of the algorithm, but by a concrete result: reducing costs, shortening response times, preventing downtime, or improving how facilities are run.
Therefore, we want to link the development of AZIOT in 2027 to practical use cases and testing their effectiveness with real data from facilities.
Intecracy Group’s strength lies in its companies’ specializations
Intecracy Group unites companies with different expertise, products, and experience. I believe that this diversity can become one of our main advantages in the coming years.
Softengi works on applied AI solutions, including computer vision for manufacturing and quality control.
Nectain develops document processing technologies: recognition, classification, extraction, and structuring of information for further automation.
InBase creates enterprise applications and is one of the key developers of the UnityBase platform.
Deals complements our capabilities in the field of electronic document exchange and interaction with counterparties.
IQusion, Data Management IG, Softline, SL Global Service, and other companies add their own products, infrastructure solutions, expertise in integration, data handling, and support.
We want our companies to start working together while discussing a customer’s needs. For example, a business implementing document management may also need cloud migration or secure infrastructure. The company already working with that customer should bring in colleagues with the relevant experience and prepare a joint proposal. This would help us expand our work with existing customers and find new ones together. The customer should understand what the solution includes, what it costs and what each participant is responsible for.
I do not believe that every company in the group needs to create its own general-purpose AI assistant (as an example; the point applies to any solution) or duplicate the technologies of its colleagues.
It is much more promising for each company to deepen its expertise, build on shared technology and develop integrated solutions around the specific needs of enterprises.
For example, in a document management automation project, Nectain's document recognition tools, IQusion's process automation, UnityBase's capabilities, and the infrastructure services of other group companies can be combined.
In a telecommunications project, DooxSwitch, corporate platforms, and integration teams can play a similar role.
Our methodology provides for a lead company for the project. It coordinates the work of other participants and is responsible to the customer for fulfilling the agreed-upon obligations.
From implementation to long-term support
The accessibility of AI simplifies the creation of prototypes but does not eliminate the responsibility for software.
Mykhailo Vyhovskyi, in his column about code created by AI, drew attention to the fact that the speed of generation does not replace engineering responsibility. That applies directly to our work.
Showing an enterprise that the software can perform a task is not enough. It is important to ensure its reliable operation, integration, access control, updates, and recovery after failures.
In my column for Software Ukraine, I wrote about situations where the software technically works but the business does not see the expected benefit.
The reason can be unprepared data, undefined responsibility, lack of training, resistance to change, or the mismatch of the automated process with the actual work of the enterprise.
Therefore, for us, the development of enterprise IT is also the ability to support the customer at every stage: from needs analysis and design to implementation, user training, and subsequent operation.
After launch, we want to agree with the customer on how the system will develop: which features are needed next, when to review costs and how to assess the results. That plan should evolve as the business changes. For our companies, it provides a basis for lasting service relationships with a defined scope of work and clear fees. For the customer, it means working with a team that knows the system and is responsible for its support and agreed changes.
New products, teams, and partnerships
The development of Intecracy Group is not limited to the companies and products that are already part of the alliance.
We are open to cooperation with technology teams, developers, and entrepreneurs who are creating new solutions or have experience in fields that complement our expertise.
For early-stage technology projects, Intecrator — an accelerator based on the Venture Builder model — is part of Intecracy Ventures.
Its task is to help teams test market demand, develop products, determine architecture, find their first customers, and prepare for scaling.
For me, the key question when considering a new idea is not how modern the technologies the team uses are.
First and foremost, it is necessary to understand what problem the product solves, who is willing to pay for it, why the chosen solution will be better than existing alternatives, and how to ensure its further development.
Not all promising technologies should be created inside Intecracy Group. In many cases, it is more effective to cooperate with teams that already have the necessary experience and their own developments.
If you have a special proposal that does not fit the Intecrator format, you can send an application to me personally through my website.
What 2027 will mean for us
In 2027, we want each new implementation to build on the products and experience our companies have already developed. Customers should be able to bring in the right teams through a clear proposal and have partners to help develop their systems after launch. That is where I see the practical value of the Intecracy Group alliance.
If you have a specific task, a promising product, or a partnership idea that we could pursue with our teams, we will be happy to discuss it.